/[pcre]/code/trunk/pcre_jit_compile.c
ViewVC logotype

Diff of /code/trunk/pcre_jit_compile.c

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 993 by zherczeg, Tue Jul 10 04:33:00 2012 UTC revision 1423 by zherczeg, Tue Dec 31 07:57:56 2013 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2012 University of Cambridge             Copyright (c) 1997-2013 University of Cambridge
10    
11    The machine code generator part (this module) was written by Zoltan Herczeg    The machine code generator part (this module) was written by Zoltan Herczeg
12                        Copyright (c) 2010-2012                        Copyright (c) 2010-2013
13    
14  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
15  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 46  POSSIBILITY OF SUCH DAMAGE. Line 46  POSSIBILITY OF SUCH DAMAGE.
46    
47  #include "pcre_internal.h"  #include "pcre_internal.h"
48    
49  #ifdef SUPPORT_JIT  #if defined SUPPORT_JIT
50    
51  /* All-in-one: Since we use the JIT compiler only from here,  /* All-in-one: Since we use the JIT compiler only from here,
52  we just include it. This way we don't need to touch the build  we just include it. This way we don't need to touch the build
# Line 65  system files. */ Line 65  system files. */
65  #error Unsupported architecture  #error Unsupported architecture
66  #endif  #endif
67    
68  /* Allocate memory on the stack. Fast, but limited size. */  /* Defines for debugging purposes. */
 #define LOCAL_SPACE_SIZE 32768  
69    
70    /* 1 - Use unoptimized capturing brackets.
71       2 - Enable capture_last_ptr (includes option 1). */
72    /* #define DEBUG_FORCE_UNOPTIMIZED_CBRAS 2 */
73    
74    /* 1 - Always have a control head. */
75    /* #define DEBUG_FORCE_CONTROL_HEAD 1 */
76    
77    /* Allocate memory for the regex stack on the real machine stack.
78    Fast, but limited size. */
79    #define MACHINE_STACK_SIZE 32768
80    
81    /* Growth rate for stack allocated by the OS. Should be the multiply
82    of page size. */
83  #define STACK_GROWTH_RATE 8192  #define STACK_GROWTH_RATE 8192
84    
85  /* Enable to check that the allocation could destroy temporaries. */  /* Enable to check that the allocation could destroy temporaries. */
# Line 89  vertical (sub-expression) (See struct ba Line 101  vertical (sub-expression) (See struct ba
101    
102  The condition checkers are boolean (true/false) checkers. Machine code is generated  The condition checkers are boolean (true/false) checkers. Machine code is generated
103  for the checker itself and for the actions depending on the result of the checker.  for the checker itself and for the actions depending on the result of the checker.
104  The 'true' case is called as the try path (expected path), and the other is called as  The 'true' case is called as the matching path (expected path), and the other is called as
105  the 'backtrack' path. Branch instructions are expesive for all CPUs, so we avoid taken  the 'backtrack' path. Branch instructions are expesive for all CPUs, so we avoid taken
106  branches on the try path.  branches on the matching path.
107    
108   Greedy star operator (*) :   Greedy star operator (*) :
109     Try path: match happens.     Matching path: match happens.
110     Backtrack path: match failed.     Backtrack path: match failed.
111   Non-greedy star operator (*?) :   Non-greedy star operator (*?) :
112     Try path: no need to perform a match.     Matching path: no need to perform a match.
113     Backtrack path: match is required.     Backtrack path: match is required.
114    
115  The following example shows how the code generated for a capturing bracket  The following example shows how the code generated for a capturing bracket
# Line 108  we have the following regular expression Line 120  we have the following regular expression
120    
121  The generated code will be the following:  The generated code will be the following:
122    
123   A try path   A matching path
124   '(' try path (pushing arguments to the stack)   '(' matching path (pushing arguments to the stack)
125   B try path   B matching path
126   ')' try path (pushing arguments to the stack)   ')' matching path (pushing arguments to the stack)
127   D try path   D matching path
128   return with successful match   return with successful match
129    
130   D backtrack path   D backtrack path
131   ')' backtrack path (If we arrived from "C" jump to the backtrack of "C")   ')' backtrack path (If we arrived from "C" jump to the backtrack of "C")
132   B backtrack path   B backtrack path
133   C expected path   C expected path
134   jump to D try path   jump to D matching path
135   C backtrack path   C backtrack path
136   A backtrack path   A backtrack path
137    
# Line 127  The generated code will be the following Line 139  The generated code will be the following
139   code paths. In this way the topmost value on the stack is always belong   code paths. In this way the topmost value on the stack is always belong
140   to the current backtrack code path. The backtrack path must check   to the current backtrack code path. The backtrack path must check
141   whether there is a next alternative. If so, it needs to jump back to   whether there is a next alternative. If so, it needs to jump back to
142   the try path eventually. Otherwise it needs to clear out its own stack   the matching path eventually. Otherwise it needs to clear out its own stack
143   frame and continue the execution on the backtrack code paths.   frame and continue the execution on the backtrack code paths.
144  */  */
145    
146  /*  /*
147  Saved stack frames:  Saved stack frames:
148    
149  Atomic blocks and asserts require reloading the values of local variables  Atomic blocks and asserts require reloading the values of private data
150  when the backtrack mechanism performed. Because of OP_RECURSE, the locals  when the backtrack mechanism performed. Because of OP_RECURSE, the data
151  are not necessarly known in compile time, thus we need a dynamic restore  are not necessarly known in compile time, thus we need a dynamic restore
152  mechanism.  mechanism.
153    
154  The stack frames are stored in a chain list, and have the following format:  The stack frames are stored in a chain list, and have the following format:
155  ([ capturing bracket offset ][ start value ][ end value ])+ ... [ 0 ] [ previous head ]  ([ capturing bracket offset ][ start value ][ end value ])+ ... [ 0 ] [ previous head ]
156    
157  Thus we can restore the locals to a particular point in the stack.  Thus we can restore the private data to a particular point in the stack.
158  */  */
159    
160  typedef struct jit_arguments {  typedef struct jit_arguments {
# Line 154  typedef struct jit_arguments { Line 166  typedef struct jit_arguments {
166    int *offsets;    int *offsets;
167    pcre_uchar *uchar_ptr;    pcre_uchar *uchar_ptr;
168    pcre_uchar *mark_ptr;    pcre_uchar *mark_ptr;
169      void *callout_data;
170    /* Everything else after. */    /* Everything else after. */
171    int offsetcount;    pcre_uint32 limit_match;
172    int calllimit;    int real_offset_count;
173      int offset_count;
174    pcre_uint8 notbol;    pcre_uint8 notbol;
175    pcre_uint8 noteol;    pcre_uint8 noteol;
176    pcre_uint8 notempty;    pcre_uint8 notempty;
# Line 167  typedef struct executable_functions { Line 181  typedef struct executable_functions {
181    void *executable_funcs[JIT_NUMBER_OF_COMPILE_MODES];    void *executable_funcs[JIT_NUMBER_OF_COMPILE_MODES];
182    PUBL(jit_callback) callback;    PUBL(jit_callback) callback;
183    void *userdata;    void *userdata;
184      pcre_uint32 top_bracket;
185      pcre_uint32 limit_match;
186    sljit_uw executable_sizes[JIT_NUMBER_OF_COMPILE_MODES];    sljit_uw executable_sizes[JIT_NUMBER_OF_COMPILE_MODES];
187  } executable_functions;  } executable_functions;
188    
# Line 175  typedef struct jump_list { Line 191  typedef struct jump_list {
191    struct jump_list *next;    struct jump_list *next;
192  } jump_list;  } jump_list;
193    
 enum stub_types { stack_alloc };  
   
194  typedef struct stub_list {  typedef struct stub_list {
   enum stub_types type;  
   int data;  
195    struct sljit_jump *start;    struct sljit_jump *start;
196    struct sljit_label *quit;    struct sljit_label *quit;
197    struct stub_list *next;    struct stub_list *next;
198  } stub_list;  } stub_list;
199    
200    enum frame_types {
201      no_frame = -1,
202      no_stack = -2
203    };
204    
205    enum control_types {
206      type_mark = 0,
207      type_then_trap = 1
208    };
209    
210  typedef int (SLJIT_CALL *jit_function)(jit_arguments *args);  typedef int (SLJIT_CALL *jit_function)(jit_arguments *args);
211    
212  /* The following structure is the key data type for the recursive  /* The following structure is the key data type for the recursive
213  code generator. It is allocated by compile_trypath, and contains  code generator. It is allocated by compile_matchingpath, and contains
214  the aguments for compile_backtrackpath. Must be the first member  the arguments for compile_backtrackingpath. Must be the first member
215  of its descendants. */  of its descendants. */
216  typedef struct backtrack_common {  typedef struct backtrack_common {
217    /* Concatenation stack. */    /* Concatenation stack. */
# Line 205  typedef struct backtrack_common { Line 227  typedef struct backtrack_common {
227  typedef struct assert_backtrack {  typedef struct assert_backtrack {
228    backtrack_common common;    backtrack_common common;
229    jump_list *condfailed;    jump_list *condfailed;
230    /* Less than 0 (-1) if a frame is not needed. */    /* Less than 0 if a frame is not needed. */
231    int framesize;    int framesize;
232    /* Points to our private memory word on the stack. */    /* Points to our private memory word on the stack. */
233    int localptr;    int private_data_ptr;
234    /* For iterators. */    /* For iterators. */
235    struct sljit_label *trypath;    struct sljit_label *matchingpath;
236  } assert_backtrack;  } assert_backtrack;
237    
238  typedef struct bracket_backtrack {  typedef struct bracket_backtrack {
239    backtrack_common common;    backtrack_common common;
240    /* Where to coninue if an alternative is successfully matched. */    /* Where to coninue if an alternative is successfully matched. */
241    struct sljit_label *alttrypath;    struct sljit_label *alternative_matchingpath;
242    /* For rmin and rmax iterators. */    /* For rmin and rmax iterators. */
243    struct sljit_label *recursivetrypath;    struct sljit_label *recursive_matchingpath;
244    /* For greedy ? operator. */    /* For greedy ? operator. */
245    struct sljit_label *zerotrypath;    struct sljit_label *zero_matchingpath;
246    /* Contains the branches of a failed condition. */    /* Contains the branches of a failed condition. */
247    union {    union {
248      /* Both for OP_COND, OP_SCOND. */      /* Both for OP_COND, OP_SCOND. */
249      jump_list *condfailed;      jump_list *condfailed;
250      assert_backtrack *assert;      assert_backtrack *assert;
251      /* For OP_ONCE. -1 if not needed. */      /* For OP_ONCE. Less than 0 if not needed. */
252      int framesize;      int framesize;
253    } u;    } u;
254    /* Points to our private memory word on the stack. */    /* Points to our private memory word on the stack. */
255    int localptr;    int private_data_ptr;
256  } bracket_backtrack;  } bracket_backtrack;
257    
258  typedef struct bracketpos_backtrack {  typedef struct bracketpos_backtrack {
259    backtrack_common common;    backtrack_common common;
260    /* Points to our private memory word on the stack. */    /* Points to our private memory word on the stack. */
261    int localptr;    int private_data_ptr;
262    /* Reverting stack is needed. */    /* Reverting stack is needed. */
263    int framesize;    int framesize;
264    /* Allocated stack size. */    /* Allocated stack size. */
# Line 245  typedef struct bracketpos_backtrack { Line 267  typedef struct bracketpos_backtrack {
267    
268  typedef struct braminzero_backtrack {  typedef struct braminzero_backtrack {
269    backtrack_common common;    backtrack_common common;
270    struct sljit_label *trypath;    struct sljit_label *matchingpath;
271  } braminzero_backtrack;  } braminzero_backtrack;
272    
273  typedef struct iterator_backtrack {  typedef struct iterator_backtrack {
274    backtrack_common common;    backtrack_common common;
275    /* Next iteration. */    /* Next iteration. */
276    struct sljit_label *trypath;    struct sljit_label *matchingpath;
277  } iterator_backtrack;  } iterator_backtrack;
278    
279  typedef struct recurse_entry {  typedef struct recurse_entry {
# Line 261  typedef struct recurse_entry { Line 283  typedef struct recurse_entry {
283    /* Collects the calls until the function is not created. */    /* Collects the calls until the function is not created. */
284    jump_list *calls;    jump_list *calls;
285    /* Points to the starting opcode. */    /* Points to the starting opcode. */
286    int start;    sljit_sw start;
287  } recurse_entry;  } recurse_entry;
288    
289  typedef struct recurse_backtrack {  typedef struct recurse_backtrack {
290    backtrack_common common;    backtrack_common common;
291      BOOL inlined_pattern;
292  } recurse_backtrack;  } recurse_backtrack;
293    
294  #define MAX_RANGE_SIZE 6  #define OP_THEN_TRAP OP_TABLE_LENGTH
295    
296    typedef struct then_trap_backtrack {
297      backtrack_common common;
298      /* If then_trap is not NULL, this structure contains the real
299      then_trap for the backtracking path. */
300      struct then_trap_backtrack *then_trap;
301      /* Points to the starting opcode. */
302      sljit_sw start;
303      /* Exit point for the then opcodes of this alternative. */
304      jump_list *quit;
305      /* Frame size of the current alternative. */
306      int framesize;
307    } then_trap_backtrack;
308    
309    #define MAX_RANGE_SIZE 4
310    
311  typedef struct compiler_common {  typedef struct compiler_common {
312      /* The sljit ceneric compiler. */
313    struct sljit_compiler *compiler;    struct sljit_compiler *compiler;
314      /* First byte code. */
315    pcre_uchar *start;    pcre_uchar *start;
316      /* Maps private data offset to each opcode. */
317    /* Opcode local area direct map. */    sljit_si *private_data_ptrs;
318    int *localptrs;    /* Tells whether the capturing bracket is optimized. */
319    int cbraptr;    pcre_uint8 *optimized_cbracket;
320    /* OVector starting point. Must be divisible by 2. */    /* Tells whether the starting offset is a target of then. */
321      pcre_uint8 *then_offsets;
322      /* Current position where a THEN must jump. */
323      then_trap_backtrack *then_trap;
324      /* Starting offset of private data for capturing brackets. */
325      int cbra_ptr;
326      /* Output vector starting point. Must be divisible by 2. */
327    int ovector_start;    int ovector_start;
328    /* Last known position of the requested byte. */    /* Last known position of the requested byte. */
329    int req_char_ptr;    int req_char_ptr;
330    /* Head of the last recursion. */    /* Head of the last recursion. */
331    int recursive_head;    int recursive_head_ptr;
332    /* First inspected character for partial matching. */    /* First inspected character for partial matching. */
333    int start_used_ptr;    int start_used_ptr;
334    /* Starting pointer for partial soft matches. */    /* Starting pointer for partial soft matches. */
# Line 291  typedef struct compiler_common { Line 337  typedef struct compiler_common {
337    int first_line_end;    int first_line_end;
338    /* Points to the marked string. */    /* Points to the marked string. */
339    int mark_ptr;    int mark_ptr;
340      /* Recursive control verb management chain. */
341      int control_head_ptr;
342      /* Points to the last matched capture block index. */
343      int capture_last_ptr;
344      /* Points to the starting position of the current match. */
345      int start_ptr;
346    
347    /* Flipped and lower case tables. */    /* Flipped and lower case tables. */
348    const pcre_uint8 *fcc;    const pcre_uint8 *fcc;
349    sljit_w lcc;    sljit_sw lcc;
350    /* Mode can be PCRE_STUDY_JIT_COMPILE and others. */    /* Mode can be PCRE_STUDY_JIT_COMPILE and others. */
351    int mode;    int mode;
352      /* \K is found in the pattern. */
353      BOOL has_set_som;
354      /* (*SKIP:arg) is found in the pattern. */
355      BOOL has_skip_arg;
356      /* (*THEN) is found in the pattern. */
357      BOOL has_then;
358      /* Needs to know the start position anytime. */
359      BOOL needs_start_ptr;
360      /* Currently in recurse or negative assert. */
361      BOOL local_exit;
362      /* Currently in a positive assert. */
363      BOOL positive_assert;
364    /* Newline control. */    /* Newline control. */
365    int nltype;    int nltype;
366      pcre_uint32 nlmax;
367    int newline;    int newline;
368    int bsr_nltype;    int bsr_nltype;
369      pcre_uint32 bsr_nlmax;
370    /* Dollar endonly. */    /* Dollar endonly. */
371    int endonly;    int endonly;
   BOOL has_set_som;  
372    /* Tables. */    /* Tables. */
373    sljit_w ctypes;    sljit_sw ctypes;
   int digits[2 + MAX_RANGE_SIZE];  
374    /* Named capturing brackets. */    /* Named capturing brackets. */
375    sljit_uw name_table;    pcre_uchar *name_table;
376    sljit_w name_count;    sljit_sw name_count;
377    sljit_w name_entry_size;    sljit_sw name_entry_size;
378    
379    /* Labels and jump lists. */    /* Labels and jump lists. */
380    struct sljit_label *partialmatchlabel;    struct sljit_label *partialmatchlabel;
381    struct sljit_label *quitlabel;    struct sljit_label *quit_label;
382    struct sljit_label *acceptlabel;    struct sljit_label *forced_quit_label;
383      struct sljit_label *accept_label;
384    stub_list *stubs;    stub_list *stubs;
385    recurse_entry *entries;    recurse_entry *entries;
386    recurse_entry *currententry;    recurse_entry *currententry;
387    jump_list *partialmatch;    jump_list *partialmatch;
388    jump_list *quit;    jump_list *quit;
389      jump_list *positive_assert_quit;
390      jump_list *forced_quit;
391    jump_list *accept;    jump_list *accept;
392    jump_list *calllimit;    jump_list *calllimit;
393    jump_list *stackalloc;    jump_list *stackalloc;
# Line 331  typedef struct compiler_common { Line 398  typedef struct compiler_common {
398    jump_list *vspace;    jump_list *vspace;
399    jump_list *casefulcmp;    jump_list *casefulcmp;
400    jump_list *caselesscmp;    jump_list *caselesscmp;
401      jump_list *reset_match;
402    BOOL jscript_compat;    BOOL jscript_compat;
403  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
404    BOOL utf;    BOOL utf;
405  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
406    BOOL use_ucp;    BOOL use_ucp;
407  #endif  #endif
   jump_list *utfreadchar;  
408  #ifdef COMPILE_PCRE8  #ifdef COMPILE_PCRE8
409      jump_list *utfreadchar;
410      jump_list *utfreadchar16;
411    jump_list *utfreadtype8;    jump_list *utfreadtype8;
412  #endif  #endif
413  #endif /* SUPPORT_UTF */  #endif /* SUPPORT_UTF */
# Line 355  typedef struct compare_context { Line 424  typedef struct compare_context {
424  #if defined SLJIT_UNALIGNED && SLJIT_UNALIGNED  #if defined SLJIT_UNALIGNED && SLJIT_UNALIGNED
425    int ucharptr;    int ucharptr;
426    union {    union {
427      sljit_i asint;      sljit_si asint;
428      sljit_uh asushort;      sljit_uh asushort;
429  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
430      sljit_ub asbyte;      sljit_ub asbyte;
431      sljit_ub asuchars[4];      sljit_ub asuchars[4];
432  #else  #elif defined COMPILE_PCRE16
 #ifdef COMPILE_PCRE16  
433      sljit_uh asuchars[2];      sljit_uh asuchars[2];
434  #endif  #elif defined COMPILE_PCRE32
435        sljit_ui asuchars[1];
436  #endif  #endif
437    } c;    } c;
438    union {    union {
439      sljit_i asint;      sljit_si asint;
440      sljit_uh asushort;      sljit_uh asushort;
441  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
442      sljit_ub asbyte;      sljit_ub asbyte;
443      sljit_ub asuchars[4];      sljit_ub asuchars[4];
444  #else  #elif defined COMPILE_PCRE16
 #ifdef COMPILE_PCRE16  
445      sljit_uh asuchars[2];      sljit_uh asuchars[2];
446  #endif  #elif defined COMPILE_PCRE32
447        sljit_ui asuchars[1];
448  #endif  #endif
449    } oc;    } oc;
450  #endif  #endif
451  } compare_context;  } compare_context;
452    
 enum {  
   frame_end = 0,  
   frame_setstrbegin = -1,  
   frame_setmark = -2  
 };  
   
453  /* Undefine sljit macros. */  /* Undefine sljit macros. */
454  #undef CMP  #undef CMP
455    
456  /* Used for accessing the elements of the stack. */  /* Used for accessing the elements of the stack. */
457  #define STACK(i)      ((-(i) - 1) * (int)sizeof(sljit_w))  #define STACK(i)      ((-(i) - 1) * (int)sizeof(sljit_sw))
458    
459  #define TMP1          SLJIT_TEMPORARY_REG1  #define TMP1          SLJIT_SCRATCH_REG1
460  #define TMP2          SLJIT_TEMPORARY_REG3  #define TMP2          SLJIT_SCRATCH_REG3
461  #define TMP3          SLJIT_TEMPORARY_EREG2  #define TMP3          SLJIT_TEMPORARY_EREG2
462  #define STR_PTR       SLJIT_SAVED_REG1  #define STR_PTR       SLJIT_SAVED_REG1
463  #define STR_END       SLJIT_SAVED_REG2  #define STR_END       SLJIT_SAVED_REG2
464  #define STACK_TOP     SLJIT_TEMPORARY_REG2  #define STACK_TOP     SLJIT_SCRATCH_REG2
465  #define STACK_LIMIT   SLJIT_SAVED_REG3  #define STACK_LIMIT   SLJIT_SAVED_REG3
466  #define ARGUMENTS     SLJIT_SAVED_EREG1  #define ARGUMENTS     SLJIT_SAVED_EREG1
467  #define CALL_COUNT    SLJIT_SAVED_EREG2  #define COUNT_MATCH   SLJIT_SAVED_EREG2
468  #define RETURN_ADDR   SLJIT_TEMPORARY_EREG1  #define RETURN_ADDR   SLJIT_TEMPORARY_EREG1
469    
470  /* Locals layout. */  /* Local space layout. */
471  /* These two locals can be used by the current opcode. */  /* These two locals can be used by the current opcode. */
472  #define LOCALS0          (0 * sizeof(sljit_w))  #define LOCALS0          (0 * sizeof(sljit_sw))
473  #define LOCALS1          (1 * sizeof(sljit_w))  #define LOCALS1          (1 * sizeof(sljit_sw))
474  /* Two local variables for possessive quantifiers (char1 cannot use them). */  /* Two local variables for possessive quantifiers (char1 cannot use them). */
475  #define POSSESSIVE0      (2 * sizeof(sljit_w))  #define POSSESSIVE0      (2 * sizeof(sljit_sw))
476  #define POSSESSIVE1      (3 * sizeof(sljit_w))  #define POSSESSIVE1      (3 * sizeof(sljit_sw))
477  /* Max limit of recursions. */  /* Max limit of recursions. */
478  #define CALL_LIMIT       (4 * sizeof(sljit_w))  #define LIMIT_MATCH      (4 * sizeof(sljit_sw))
479  /* The output vector is stored on the stack, and contains pointers  /* The output vector is stored on the stack, and contains pointers
480  to characters. The vector data is divided into two groups: the first  to characters. The vector data is divided into two groups: the first
481  group contains the start / end character pointers, and the second is  group contains the start / end character pointers, and the second is
482  the start pointers when the end of the capturing group has not yet reached. */  the start pointers when the end of the capturing group has not yet reached. */
483  #define OVECTOR_START    (common->ovector_start)  #define OVECTOR_START    (common->ovector_start)
484  #define OVECTOR(i)       (OVECTOR_START + (i) * sizeof(sljit_w))  #define OVECTOR(i)       (OVECTOR_START + (i) * (sljit_sw)sizeof(sljit_sw))
485  #define OVECTOR_PRIV(i)  (common->cbraptr + (i) * sizeof(sljit_w))  #define OVECTOR_PRIV(i)  (common->cbra_ptr + (i) * (sljit_sw)sizeof(sljit_sw))
486  #define PRIV_DATA(cc)    (common->localptrs[(cc) - common->start])  #define PRIVATE_DATA(cc) (common->private_data_ptrs[(cc) - common->start])
487    
488  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
489  #define MOV_UCHAR  SLJIT_MOV_UB  #define MOV_UCHAR  SLJIT_MOV_UB
490  #define MOVU_UCHAR SLJIT_MOVU_UB  #define MOVU_UCHAR SLJIT_MOVU_UB
491  #else  #elif defined COMPILE_PCRE16
 #ifdef COMPILE_PCRE16  
492  #define MOV_UCHAR  SLJIT_MOV_UH  #define MOV_UCHAR  SLJIT_MOV_UH
493  #define MOVU_UCHAR SLJIT_MOVU_UH  #define MOVU_UCHAR SLJIT_MOVU_UH
494    #elif defined COMPILE_PCRE32
495    #define MOV_UCHAR  SLJIT_MOV_UI
496    #define MOVU_UCHAR SLJIT_MOVU_UI
497  #else  #else
498  #error Unsupported compiling mode  #error Unsupported compiling mode
499  #endif  #endif
 #endif  
500    
501  /* Shortcuts. */  /* Shortcuts. */
502  #define DEFINE_COMPILER \  #define DEFINE_COMPILER \
# Line 449  the start pointers when the end of the c Line 513  the start pointers when the end of the c
513    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))
514  #define JUMPHERE(jump) \  #define JUMPHERE(jump) \
515    sljit_set_label((jump), sljit_emit_label(compiler))    sljit_set_label((jump), sljit_emit_label(compiler))
516    #define SET_LABEL(jump, label) \
517      sljit_set_label((jump), (label))
518  #define CMP(type, src1, src1w, src2, src2w) \  #define CMP(type, src1, src1w, src2, src2w) \
519    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))
520  #define CMPTO(type, src1, src1w, src2, src2w, label) \  #define CMPTO(type, src1, src1w, src2, src2w, label) \
521    sljit_set_label(sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w)), (label))    sljit_set_label(sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w)), (label))
522  #define COND_VALUE(op, dst, dstw, type) \  #define OP_FLAGS(op, dst, dstw, src, srcw, type) \
523    sljit_emit_cond_value(compiler, (op), (dst), (dstw), (type))    sljit_emit_op_flags(compiler, (op), (dst), (dstw), (src), (srcw), (type))
524  #define GET_LOCAL_BASE(dst, dstw, offset) \  #define GET_LOCAL_BASE(dst, dstw, offset) \
525    sljit_get_local_base(compiler, (dst), (dstw), (offset))    sljit_get_local_base(compiler, (dst), (dstw), (offset))
526    
527    #define READ_CHAR_ANY 0x7fffffff
528    
529  static pcre_uchar* bracketend(pcre_uchar* cc)  static pcre_uchar* bracketend(pcre_uchar* cc)
530  {  {
531  SLJIT_ASSERT((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND));  SLJIT_ASSERT((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND));
# Line 467  cc += 1 + LINK_SIZE; Line 535  cc += 1 + LINK_SIZE;
535  return cc;  return cc;
536  }  }
537    
538    static int ones_in_half_byte[16] = {
539      /* 0 */ 0, 1, 1, 2, /* 4 */ 1, 2, 2, 3,
540      /* 8 */ 1, 2, 2, 3, /* 12 */ 2, 3, 3, 4
541    };
542    
543  /* Functions whose might need modification for all new supported opcodes:  /* Functions whose might need modification for all new supported opcodes:
544   next_opcode   next_opcode
545   get_localspace   check_opcode_types
546   set_localptrs   set_private_data_ptrs
547   get_framesize   get_framesize
548   init_frame   init_frame
549   get_localsize   get_private_data_copy_length
550   copy_locals   copy_private_data
551   compile_trypath   compile_matchingpath
552   compile_backtrackpath   compile_backtrackingpath
553  */  */
554    
555  static pcre_uchar *next_opcode(compiler_common *common, pcre_uchar *cc)  static pcre_uchar *next_opcode(compiler_common *common, pcre_uchar *cc)
# Line 497  switch(*cc) Line 570  switch(*cc)
570    case OP_WORDCHAR:    case OP_WORDCHAR:
571    case OP_ANY:    case OP_ANY:
572    case OP_ALLANY:    case OP_ALLANY:
573      case OP_NOTPROP:
574      case OP_PROP:
575    case OP_ANYNL:    case OP_ANYNL:
576    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
577    case OP_HSPACE:    case OP_HSPACE:
# Line 509  switch(*cc) Line 584  switch(*cc)
584    case OP_CIRCM:    case OP_CIRCM:
585    case OP_DOLL:    case OP_DOLL:
586    case OP_DOLLM:    case OP_DOLLM:
   case OP_TYPESTAR:  
   case OP_TYPEMINSTAR:  
   case OP_TYPEPLUS:  
   case OP_TYPEMINPLUS:  
   case OP_TYPEQUERY:  
   case OP_TYPEMINQUERY:  
   case OP_TYPEPOSSTAR:  
   case OP_TYPEPOSPLUS:  
   case OP_TYPEPOSQUERY:  
587    case OP_CRSTAR:    case OP_CRSTAR:
588    case OP_CRMINSTAR:    case OP_CRMINSTAR:
589    case OP_CRPLUS:    case OP_CRPLUS:
590    case OP_CRMINPLUS:    case OP_CRMINPLUS:
591    case OP_CRQUERY:    case OP_CRQUERY:
592    case OP_CRMINQUERY:    case OP_CRMINQUERY:
593      case OP_CRRANGE:
594      case OP_CRMINRANGE:
595      case OP_CRPOSSTAR:
596      case OP_CRPOSPLUS:
597      case OP_CRPOSQUERY:
598      case OP_CRPOSRANGE:
599      case OP_CLASS:
600      case OP_NCLASS:
601      case OP_REF:
602      case OP_REFI:
603      case OP_DNREF:
604      case OP_DNREFI:
605      case OP_RECURSE:
606      case OP_CALLOUT:
607      case OP_ALT:
608      case OP_KET:
609      case OP_KETRMAX:
610      case OP_KETRMIN:
611      case OP_KETRPOS:
612      case OP_REVERSE:
613      case OP_ASSERT:
614      case OP_ASSERT_NOT:
615      case OP_ASSERTBACK:
616      case OP_ASSERTBACK_NOT:
617      case OP_ONCE:
618      case OP_ONCE_NC:
619      case OP_BRA:
620      case OP_BRAPOS:
621      case OP_CBRA:
622      case OP_CBRAPOS:
623      case OP_COND:
624      case OP_SBRA:
625      case OP_SBRAPOS:
626      case OP_SCBRA:
627      case OP_SCBRAPOS:
628      case OP_SCOND:
629      case OP_CREF:
630      case OP_DNCREF:
631      case OP_RREF:
632      case OP_DNRREF:
633    case OP_DEF:    case OP_DEF:
634    case OP_BRAZERO:    case OP_BRAZERO:
635    case OP_BRAMINZERO:    case OP_BRAMINZERO:
636    case OP_BRAPOSZERO:    case OP_BRAPOSZERO:
637      case OP_PRUNE:
638      case OP_SKIP:
639      case OP_THEN:
640    case OP_COMMIT:    case OP_COMMIT:
641    case OP_FAIL:    case OP_FAIL:
642    case OP_ACCEPT:    case OP_ACCEPT:
643    case OP_ASSERT_ACCEPT:    case OP_ASSERT_ACCEPT:
644      case OP_CLOSE:
645    case OP_SKIPZERO:    case OP_SKIPZERO:
646    return cc + 1;    return cc + PRIV(OP_lengths)[*cc];
   
   case OP_ANYBYTE:  
 #ifdef SUPPORT_UTF  
   if (common->utf) return NULL;  
 #endif  
   return cc + 1;  
647    
648    case OP_CHAR:    case OP_CHAR:
649    case OP_CHARI:    case OP_CHARI:
# Line 551  switch(*cc) Line 655  switch(*cc)
655    case OP_MINPLUS:    case OP_MINPLUS:
656    case OP_QUERY:    case OP_QUERY:
657    case OP_MINQUERY:    case OP_MINQUERY:
658      case OP_UPTO:
659      case OP_MINUPTO:
660      case OP_EXACT:
661    case OP_POSSTAR:    case OP_POSSTAR:
662    case OP_POSPLUS:    case OP_POSPLUS:
663    case OP_POSQUERY:    case OP_POSQUERY:
664      case OP_POSUPTO:
665    case OP_STARI:    case OP_STARI:
666    case OP_MINSTARI:    case OP_MINSTARI:
667    case OP_PLUSI:    case OP_PLUSI:
668    case OP_MINPLUSI:    case OP_MINPLUSI:
669    case OP_QUERYI:    case OP_QUERYI:
670    case OP_MINQUERYI:    case OP_MINQUERYI:
671      case OP_UPTOI:
672      case OP_MINUPTOI:
673      case OP_EXACTI:
674    case OP_POSSTARI:    case OP_POSSTARI:
675    case OP_POSPLUSI:    case OP_POSPLUSI:
676    case OP_POSQUERYI:    case OP_POSQUERYI:
677      case OP_POSUPTOI:
678    case OP_NOTSTAR:    case OP_NOTSTAR:
679    case OP_NOTMINSTAR:    case OP_NOTMINSTAR:
680    case OP_NOTPLUS:    case OP_NOTPLUS:
681    case OP_NOTMINPLUS:    case OP_NOTMINPLUS:
682    case OP_NOTQUERY:    case OP_NOTQUERY:
683    case OP_NOTMINQUERY:    case OP_NOTMINQUERY:
684      case OP_NOTUPTO:
685      case OP_NOTMINUPTO:
686      case OP_NOTEXACT:
687    case OP_NOTPOSSTAR:    case OP_NOTPOSSTAR:
688    case OP_NOTPOSPLUS:    case OP_NOTPOSPLUS:
689    case OP_NOTPOSQUERY:    case OP_NOTPOSQUERY:
690      case OP_NOTPOSUPTO:
691    case OP_NOTSTARI:    case OP_NOTSTARI:
692    case OP_NOTMINSTARI:    case OP_NOTMINSTARI:
693    case OP_NOTPLUSI:    case OP_NOTPLUSI:
694    case OP_NOTMINPLUSI:    case OP_NOTMINPLUSI:
695    case OP_NOTQUERYI:    case OP_NOTQUERYI:
696    case OP_NOTMINQUERYI:    case OP_NOTMINQUERYI:
   case OP_NOTPOSSTARI:  
   case OP_NOTPOSPLUSI:  
   case OP_NOTPOSQUERYI:  
   cc += 2;  
 #ifdef SUPPORT_UTF  
   if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);  
 #endif  
   return cc;  
   
   case OP_UPTO:  
   case OP_MINUPTO:  
   case OP_EXACT:  
   case OP_POSUPTO:  
   case OP_UPTOI:  
   case OP_MINUPTOI:  
   case OP_EXACTI:  
   case OP_POSUPTOI:  
   case OP_NOTUPTO:  
   case OP_NOTMINUPTO:  
   case OP_NOTEXACT:  
   case OP_NOTPOSUPTO:  
697    case OP_NOTUPTOI:    case OP_NOTUPTOI:
698    case OP_NOTMINUPTOI:    case OP_NOTMINUPTOI:
699    case OP_NOTEXACTI:    case OP_NOTEXACTI:
700      case OP_NOTPOSSTARI:
701      case OP_NOTPOSPLUSI:
702      case OP_NOTPOSQUERYI:
703    case OP_NOTPOSUPTOI:    case OP_NOTPOSUPTOI:
704    cc += 2 + IMM2_SIZE;    cc += PRIV(OP_lengths)[*cc];
705  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
706    if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);    if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
707  #endif  #endif
708    return cc;    return cc;
709    
710    case OP_NOTPROP:    /* Special cases. */
711    case OP_PROP:    case OP_TYPESTAR:
712    return cc + 1 + 2;    case OP_TYPEMINSTAR:
713      case OP_TYPEPLUS:
714      case OP_TYPEMINPLUS:
715      case OP_TYPEQUERY:
716      case OP_TYPEMINQUERY:
717    case OP_TYPEUPTO:    case OP_TYPEUPTO:
718    case OP_TYPEMINUPTO:    case OP_TYPEMINUPTO:
719    case OP_TYPEEXACT:    case OP_TYPEEXACT:
720      case OP_TYPEPOSSTAR:
721      case OP_TYPEPOSPLUS:
722      case OP_TYPEPOSQUERY:
723    case OP_TYPEPOSUPTO:    case OP_TYPEPOSUPTO:
724    case OP_REF:    return cc + PRIV(OP_lengths)[*cc] - 1;
   case OP_REFI:  
   case OP_CREF:  
   case OP_NCREF:  
   case OP_RREF:  
   case OP_NRREF:  
   case OP_CLOSE:  
   cc += 1 + IMM2_SIZE;  
   return cc;  
   
   case OP_CRRANGE:  
   case OP_CRMINRANGE:  
   return cc + 1 + 2 * IMM2_SIZE;  
725    
726    case OP_CLASS:    case OP_ANYBYTE:
727    case OP_NCLASS:  #ifdef SUPPORT_UTF
728    return cc + 1 + 32 / sizeof(pcre_uchar);    if (common->utf) return NULL;
729    #endif
730      return cc + 1;
731    
732  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
733    case OP_XCLASS:    case OP_XCLASS:
734    return cc + GET(cc, 1);    return cc + GET(cc, 1);
735  #endif  #endif
736    
   case OP_RECURSE:  
   case OP_ASSERT:  
   case OP_ASSERT_NOT:  
   case OP_ASSERTBACK:  
   case OP_ASSERTBACK_NOT:  
   case OP_REVERSE:  
   case OP_ONCE:  
   case OP_ONCE_NC:  
   case OP_BRA:  
   case OP_BRAPOS:  
   case OP_COND:  
   case OP_SBRA:  
   case OP_SBRAPOS:  
   case OP_SCOND:  
   case OP_ALT:  
   case OP_KET:  
   case OP_KETRMAX:  
   case OP_KETRMIN:  
   case OP_KETRPOS:  
   return cc + 1 + LINK_SIZE;  
   
   case OP_CBRA:  
   case OP_CBRAPOS:  
   case OP_SCBRA:  
   case OP_SCBRAPOS:  
   return cc + 1 + LINK_SIZE + IMM2_SIZE;  
   
737    case OP_MARK:    case OP_MARK:
738      case OP_PRUNE_ARG:
739      case OP_SKIP_ARG:
740      case OP_THEN_ARG:
741    return cc + 1 + 2 + cc[1];    return cc + 1 + 2 + cc[1];
742    
743    default:    default:
744      /* All opcodes are supported now! */
745      SLJIT_ASSERT_STOP();
746    return NULL;    return NULL;
747    }    }
748  }  }
749    
750  #define CASE_ITERATOR_LOCAL1 \  static BOOL check_opcode_types(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend)
     case OP_MINSTAR: \  
     case OP_MINPLUS: \  
     case OP_QUERY: \  
     case OP_MINQUERY: \  
     case OP_MINSTARI: \  
     case OP_MINPLUSI: \  
     case OP_QUERYI: \  
     case OP_MINQUERYI: \  
     case OP_NOTMINSTAR: \  
     case OP_NOTMINPLUS: \  
     case OP_NOTQUERY: \  
     case OP_NOTMINQUERY: \  
     case OP_NOTMINSTARI: \  
     case OP_NOTMINPLUSI: \  
     case OP_NOTQUERYI: \  
     case OP_NOTMINQUERYI:  
   
 #define CASE_ITERATOR_LOCAL2A \  
     case OP_STAR: \  
     case OP_PLUS: \  
     case OP_STARI: \  
     case OP_PLUSI: \  
     case OP_NOTSTAR: \  
     case OP_NOTPLUS: \  
     case OP_NOTSTARI: \  
     case OP_NOTPLUSI:  
   
 #define CASE_ITERATOR_LOCAL2B \  
     case OP_UPTO: \  
     case OP_MINUPTO: \  
     case OP_UPTOI: \  
     case OP_MINUPTOI: \  
     case OP_NOTUPTO: \  
     case OP_NOTMINUPTO: \  
     case OP_NOTUPTOI: \  
     case OP_NOTMINUPTOI:  
   
 #define CASE_ITERATOR_TYPE_LOCAL1 \  
     case OP_TYPEMINSTAR: \  
     case OP_TYPEMINPLUS: \  
     case OP_TYPEQUERY: \  
     case OP_TYPEMINQUERY:  
   
 #define CASE_ITERATOR_TYPE_LOCAL2A \  
     case OP_TYPESTAR: \  
     case OP_TYPEPLUS:  
   
 #define CASE_ITERATOR_TYPE_LOCAL2B \  
     case OP_TYPEUPTO: \  
     case OP_TYPEMINUPTO:  
   
 static int get_class_iterator_size(pcre_uchar *cc)  
 {  
 switch(*cc)  
   {  
   case OP_CRSTAR:  
   case OP_CRPLUS:  
   return 2;  
   
   case OP_CRMINSTAR:  
   case OP_CRMINPLUS:  
   case OP_CRQUERY:  
   case OP_CRMINQUERY:  
   return 1;  
   
   case OP_CRRANGE:  
   case OP_CRMINRANGE:  
   if (GET2(cc, 1) == GET2(cc, 1 + IMM2_SIZE))  
     return 0;  
   return 2;  
   
   default:  
   return 0;  
   }  
 }  
   
 static int get_localspace(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend)  
751  {  {
752  int localspace = 0;  int count;
753  pcre_uchar *alternative;  pcre_uchar *slot;
 pcre_uchar *end = NULL;  
 int space, size, bracketlen;  
754    
755  /* Calculate important variables (like stack size) and checks whether all opcodes are supported. */  /* Calculate important variables (like stack size) and checks whether all opcodes are supported. */
756  while (cc < ccend)  while (cc < ccend)
757    {    {
   space = 0;  
   size = 0;  
   bracketlen = 0;  
758    switch(*cc)    switch(*cc)
759      {      {
760      case OP_SET_SOM:      case OP_SET_SOM:
# Line 772  while (cc < ccend) Line 762  while (cc < ccend)
762      cc += 1;      cc += 1;
763      break;      break;
764    
765      case OP_ASSERT:      case OP_REF:
766      case OP_ASSERT_NOT:      case OP_REFI:
767      case OP_ASSERTBACK:      common->optimized_cbracket[GET2(cc, 1)] = 0;
768      case OP_ASSERTBACK_NOT:      cc += 1 + IMM2_SIZE;
     case OP_ONCE:  
     case OP_ONCE_NC:  
     case OP_BRAPOS:  
     case OP_SBRA:  
     case OP_SBRAPOS:  
     case OP_SCOND:  
     localspace += sizeof(sljit_w);  
     bracketlen = 1 + LINK_SIZE;  
769      break;      break;
770    
771      case OP_CBRAPOS:      case OP_CBRAPOS:
772      case OP_SCBRAPOS:      case OP_SCBRAPOS:
773      localspace += sizeof(sljit_w);      common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] = 0;
774      bracketlen = 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
775      break;      break;
776    
777      case OP_COND:      case OP_COND:
778      /* Might be a hidden SCOND. */      case OP_SCOND:
779      alternative = cc + GET(cc, 1);      /* Only AUTO_CALLOUT can insert this opcode. We do
780      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)         not intend to support this case. */
781        localspace += sizeof(sljit_w);      if (cc[1 + LINK_SIZE] == OP_CALLOUT)
782      bracketlen = 1 + LINK_SIZE;        return FALSE;
783      break;      cc += 1 + LINK_SIZE;
   
     case OP_BRA:  
     bracketlen = 1 + LINK_SIZE;  
     break;  
   
     case OP_CBRA:  
     case OP_SCBRA:  
     bracketlen = 1 + LINK_SIZE + IMM2_SIZE;  
     break;  
   
     CASE_ITERATOR_LOCAL1  
     space = 1;  
     size = -2;  
     break;  
   
     CASE_ITERATOR_LOCAL2A  
     space = 2;  
     size = -2;  
     break;  
   
     CASE_ITERATOR_LOCAL2B  
     space = 2;  
     size = -(2 + IMM2_SIZE);  
     break;  
   
     CASE_ITERATOR_TYPE_LOCAL1  
     space = 1;  
     size = 1;  
     break;  
   
     CASE_ITERATOR_TYPE_LOCAL2A  
     if (cc[1] != OP_ANYNL && cc[1] != OP_EXTUNI)  
       space = 2;  
     size = 1;  
     break;  
   
     CASE_ITERATOR_TYPE_LOCAL2B  
     if (cc[1 + IMM2_SIZE] != OP_ANYNL && cc[1 + IMM2_SIZE] != OP_EXTUNI)  
       space = 2;  
     size = 1 + IMM2_SIZE;  
784      break;      break;
785    
786      case OP_CLASS:      case OP_CREF:
787      case OP_NCLASS:      common->optimized_cbracket[GET2(cc, 1)] = 0;
788      size += 1 + 32 / sizeof(pcre_uchar);      cc += 1 + IMM2_SIZE;
     space = get_class_iterator_size(cc + size);  
789      break;      break;
790    
791  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8      case OP_DNREF:
792      case OP_XCLASS:      case OP_DNREFI:
793      size = GET(cc, 1);      case OP_DNCREF:
794      space = get_class_iterator_size(cc + size);      count = GET2(cc, 1 + IMM2_SIZE);
795        slot = common->name_table + GET2(cc, 1) * common->name_entry_size;
796        while (count-- > 0)
797          {
798          common->optimized_cbracket[GET2(slot, 0)] = 0;
799          slot += common->name_entry_size;
800          }
801        cc += 1 + 2 * IMM2_SIZE;
802      break;      break;
 #endif  
803    
804      case OP_RECURSE:      case OP_RECURSE:
805      /* Set its value only once. */      /* Set its value only once. */
806      if (common->recursive_head == 0)      if (common->recursive_head_ptr == 0)
807        {        {
808        common->recursive_head = common->ovector_start;        common->recursive_head_ptr = common->ovector_start;
809        common->ovector_start += sizeof(sljit_w);        common->ovector_start += sizeof(sljit_sw);
810        }        }
811      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
812      break;      break;
813    
814        case OP_CALLOUT:
815        if (common->capture_last_ptr == 0)
816          {
817          common->capture_last_ptr = common->ovector_start;
818          common->ovector_start += sizeof(sljit_sw);
819          }
820        cc += 2 + 2 * LINK_SIZE;
821        break;
822    
823        case OP_THEN_ARG:
824        common->has_then = TRUE;
825        common->control_head_ptr = 1;
826        /* Fall through. */
827    
828        case OP_PRUNE_ARG:
829        common->needs_start_ptr = TRUE;
830        /* Fall through. */
831    
832      case OP_MARK:      case OP_MARK:
833      if (common->mark_ptr == 0)      if (common->mark_ptr == 0)
834        {        {
835        common->mark_ptr = common->ovector_start;        common->mark_ptr = common->ovector_start;
836        common->ovector_start += sizeof(sljit_w);        common->ovector_start += sizeof(sljit_sw);
837        }        }
838      cc += 1 + 2 + cc[1];      cc += 1 + 2 + cc[1];
839      break;      break;
840    
841        case OP_THEN:
842        common->has_then = TRUE;
843        common->control_head_ptr = 1;
844        /* Fall through. */
845    
846        case OP_PRUNE:
847        case OP_SKIP:
848        common->needs_start_ptr = TRUE;
849        cc += 1;
850        break;
851    
852        case OP_SKIP_ARG:
853        common->control_head_ptr = 1;
854        common->has_skip_arg = TRUE;
855        cc += 1 + 2 + cc[1];
856        break;
857    
858      default:      default:
859      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
860      if (cc == NULL)      if (cc == NULL)
861        return -1;        return FALSE;
862      break;      break;
863      }      }
864      }
865    return TRUE;
866    }
867    
868    if (space > 0 && cc >= end)  static int get_class_iterator_size(pcre_uchar *cc)
869      localspace += sizeof(sljit_w) * space;  {
870    switch(*cc)
871      {
872      case OP_CRSTAR:
873      case OP_CRPLUS:
874      return 2;
875    
876    if (size != 0)    case OP_CRMINSTAR:
877      case OP_CRMINPLUS:
878      case OP_CRQUERY:
879      case OP_CRMINQUERY:
880      return 1;
881    
882      case OP_CRRANGE:
883      case OP_CRMINRANGE:
884      if (GET2(cc, 1) == GET2(cc, 1 + IMM2_SIZE))
885        return 0;
886      return 2;
887    
888      default:
889      return 0;
890      }
891    }
892    
893    static BOOL detect_repeat(compiler_common *common, pcre_uchar *begin)
894    {
895    pcre_uchar *end = bracketend(begin);
896    pcre_uchar *next;
897    pcre_uchar *next_end;
898    pcre_uchar *max_end;
899    pcre_uchar type;
900    sljit_sw length = end - begin;
901    int min, max, i;
902    
903    /* Detect fixed iterations first. */
904    if (end[-(1 + LINK_SIZE)] != OP_KET)
905      return FALSE;
906    
907    /* Already detected repeat. */
908    if (common->private_data_ptrs[end - common->start - LINK_SIZE] != 0)
909      return TRUE;
910    
911    next = end;
912    min = 1;
913    while (1)
914      {
915      if (*next != *begin)
916        break;
917      next_end = bracketend(next);
918      if (next_end - next != length || memcmp(begin, next, IN_UCHARS(length)) != 0)
919        break;
920      next = next_end;
921      min++;
922      }
923    
924    if (min == 2)
925      return FALSE;
926    
927    max = 0;
928    max_end = next;
929    if (*next == OP_BRAZERO || *next == OP_BRAMINZERO)
930      {
931      type = *next;
932      while (1)
933      {      {
934      if (size < 0)      if (next[0] != type || next[1] != OP_BRA || next[2 + LINK_SIZE] != *begin)
935        {        break;
936        cc += -size;      next_end = bracketend(next + 2 + LINK_SIZE);
937  #ifdef SUPPORT_UTF      if (next_end - next != (length + 2 + LINK_SIZE) || memcmp(begin, next + 2 + LINK_SIZE, IN_UCHARS(length)) != 0)
938        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);        break;
939  #endif      next = next_end;
940        }      max++;
     else  
       cc += size;  
941      }      }
942    
943    if (bracketlen > 0)    if (next[0] == type && next[1] == *begin && max >= 1)
944      {      {
945      if (cc >= end)      next_end = bracketend(next + 1);
946        if (next_end - next == (length + 1) && memcmp(begin, next + 1, IN_UCHARS(length)) == 0)
947        {        {
948        end = bracketend(cc);        for (i = 0; i < max; i++, next_end += 1 + LINK_SIZE)
949        if (end[-1 - LINK_SIZE] == OP_KET)          if (*next_end != OP_KET)
950          end = NULL;            break;
951    
952          if (i == max)
953            {
954            common->private_data_ptrs[max_end - common->start - LINK_SIZE] = next_end - max_end;
955            common->private_data_ptrs[max_end - common->start - LINK_SIZE + 1] = (type == OP_BRAZERO) ? OP_UPTO : OP_MINUPTO;
956            /* +2 the original and the last. */
957            common->private_data_ptrs[max_end - common->start - LINK_SIZE + 2] = max + 2;
958            if (min == 1)
959              return TRUE;
960            min--;
961            max_end -= (1 + LINK_SIZE) + GET(max_end, -LINK_SIZE);
962            }
963        }        }
     cc += bracketlen;  
964      }      }
965    }    }
966  return localspace;  
967    if (min >= 3)
968      {
969      common->private_data_ptrs[end - common->start - LINK_SIZE] = max_end - end;
970      common->private_data_ptrs[end - common->start - LINK_SIZE + 1] = OP_EXACT;
971      common->private_data_ptrs[end - common->start - LINK_SIZE + 2] = min;
972      return TRUE;
973      }
974    
975    return FALSE;
976  }  }
977    
978  static void set_localptrs(compiler_common *common, int localptr, pcre_uchar *ccend)  #define CASE_ITERATOR_PRIVATE_DATA_1 \
979        case OP_MINSTAR: \
980        case OP_MINPLUS: \
981        case OP_QUERY: \
982        case OP_MINQUERY: \
983        case OP_MINSTARI: \
984        case OP_MINPLUSI: \
985        case OP_QUERYI: \
986        case OP_MINQUERYI: \
987        case OP_NOTMINSTAR: \
988        case OP_NOTMINPLUS: \
989        case OP_NOTQUERY: \
990        case OP_NOTMINQUERY: \
991        case OP_NOTMINSTARI: \
992        case OP_NOTMINPLUSI: \
993        case OP_NOTQUERYI: \
994        case OP_NOTMINQUERYI:
995    
996    #define CASE_ITERATOR_PRIVATE_DATA_2A \
997        case OP_STAR: \
998        case OP_PLUS: \
999        case OP_STARI: \
1000        case OP_PLUSI: \
1001        case OP_NOTSTAR: \
1002        case OP_NOTPLUS: \
1003        case OP_NOTSTARI: \
1004        case OP_NOTPLUSI:
1005    
1006    #define CASE_ITERATOR_PRIVATE_DATA_2B \
1007        case OP_UPTO: \
1008        case OP_MINUPTO: \
1009        case OP_UPTOI: \
1010        case OP_MINUPTOI: \
1011        case OP_NOTUPTO: \
1012        case OP_NOTMINUPTO: \
1013        case OP_NOTUPTOI: \
1014        case OP_NOTMINUPTOI:
1015    
1016    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_1 \
1017        case OP_TYPEMINSTAR: \
1018        case OP_TYPEMINPLUS: \
1019        case OP_TYPEQUERY: \
1020        case OP_TYPEMINQUERY:
1021    
1022    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2A \
1023        case OP_TYPESTAR: \
1024        case OP_TYPEPLUS:
1025    
1026    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2B \
1027        case OP_TYPEUPTO: \
1028        case OP_TYPEMINUPTO:
1029    
1030    static void set_private_data_ptrs(compiler_common *common, int *private_data_start, pcre_uchar *ccend)
1031  {  {
1032  pcre_uchar *cc = common->start;  pcre_uchar *cc = common->start;
1033  pcre_uchar *alternative;  pcre_uchar *alternative;
1034  pcre_uchar *end = NULL;  pcre_uchar *end = NULL;
1035    int private_data_ptr = *private_data_start;
1036  int space, size, bracketlen;  int space, size, bracketlen;
1037    
1038  while (cc < ccend)  while (cc < ccend)
# Line 922  while (cc < ccend) Line 1040  while (cc < ccend)
1040    space = 0;    space = 0;
1041    size = 0;    size = 0;
1042    bracketlen = 0;    bracketlen = 0;
1043      if (private_data_ptr > SLJIT_MAX_LOCAL_SIZE)
1044        return;
1045    
1046      if (*cc == OP_ONCE || *cc == OP_ONCE_NC || *cc == OP_BRA || *cc == OP_CBRA || *cc == OP_COND)
1047        if (detect_repeat(common, cc))
1048          {
1049          /* These brackets are converted to repeats, so no global
1050          based single character repeat is allowed. */
1051          if (cc >= end)
1052            end = bracketend(cc);
1053          }
1054    
1055    switch(*cc)    switch(*cc)
1056      {      {
1057        case OP_KET:
1058        if (common->private_data_ptrs[cc + 1 - common->start] != 0)
1059          {
1060          common->private_data_ptrs[cc - common->start] = private_data_ptr;
1061          private_data_ptr += sizeof(sljit_sw);
1062          cc += common->private_data_ptrs[cc + 1 - common->start];
1063          }
1064        cc += 1 + LINK_SIZE;
1065        break;
1066    
1067      case OP_ASSERT:      case OP_ASSERT:
1068      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1069      case OP_ASSERTBACK:      case OP_ASSERTBACK:
# Line 934  while (cc < ccend) Line 1074  while (cc < ccend)
1074      case OP_SBRA:      case OP_SBRA:
1075      case OP_SBRAPOS:      case OP_SBRAPOS:
1076      case OP_SCOND:      case OP_SCOND:
1077      common->localptrs[cc - common->start] = localptr;      common->private_data_ptrs[cc - common->start] = private_data_ptr;
1078      localptr += sizeof(sljit_w);      private_data_ptr += sizeof(sljit_sw);
1079      bracketlen = 1 + LINK_SIZE;      bracketlen = 1 + LINK_SIZE;
1080      break;      break;
1081    
1082      case OP_CBRAPOS:      case OP_CBRAPOS:
1083      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1084      common->localptrs[cc - common->start] = localptr;      common->private_data_ptrs[cc - common->start] = private_data_ptr;
1085      localptr += sizeof(sljit_w);      private_data_ptr += sizeof(sljit_sw);
1086      bracketlen = 1 + LINK_SIZE + IMM2_SIZE;      bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
1087      break;      break;
1088    
# Line 951  while (cc < ccend) Line 1091  while (cc < ccend)
1091      alternative = cc + GET(cc, 1);      alternative = cc + GET(cc, 1);
1092      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1093        {        {
1094        common->localptrs[cc - common->start] = localptr;        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1095        localptr += sizeof(sljit_w);        private_data_ptr += sizeof(sljit_sw);
1096        }        }
1097      bracketlen = 1 + LINK_SIZE;      bracketlen = 1 + LINK_SIZE;
1098      break;      break;
# Line 966  while (cc < ccend) Line 1106  while (cc < ccend)
1106      bracketlen = 1 + LINK_SIZE + IMM2_SIZE;      bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
1107      break;      break;
1108    
1109      CASE_ITERATOR_LOCAL1      CASE_ITERATOR_PRIVATE_DATA_1
1110      space = 1;      space = 1;
1111      size = -2;      size = -2;
1112      break;      break;
1113    
1114      CASE_ITERATOR_LOCAL2A      CASE_ITERATOR_PRIVATE_DATA_2A
1115      space = 2;      space = 2;
1116      size = -2;      size = -2;
1117      break;      break;
1118    
1119      CASE_ITERATOR_LOCAL2B      CASE_ITERATOR_PRIVATE_DATA_2B
1120      space = 2;      space = 2;
1121      size = -(2 + IMM2_SIZE);      size = -(2 + IMM2_SIZE);
1122      break;      break;
1123    
1124      CASE_ITERATOR_TYPE_LOCAL1      CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1125      space = 1;      space = 1;
1126      size = 1;      size = 1;
1127      break;      break;
1128    
1129      CASE_ITERATOR_TYPE_LOCAL2A      CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1130      if (cc[1] != OP_ANYNL && cc[1] != OP_EXTUNI)      if (cc[1] != OP_ANYNL && cc[1] != OP_EXTUNI)
1131        space = 2;        space = 2;
1132      size = 1;      size = 1;
1133      break;      break;
1134    
1135      CASE_ITERATOR_TYPE_LOCAL2B      CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1136      if (cc[1 + IMM2_SIZE] != OP_ANYNL && cc[1 + IMM2_SIZE] != OP_EXTUNI)      if (cc[1 + IMM2_SIZE] != OP_ANYNL && cc[1 + IMM2_SIZE] != OP_EXTUNI)
1137        space = 2;        space = 2;
1138      size = 1 + IMM2_SIZE;      size = 1 + IMM2_SIZE;
# Line 1017  while (cc < ccend) Line 1157  while (cc < ccend)
1157      break;      break;
1158      }      }
1159    
1160      /* Character iterators, which are not inside a repeated bracket,
1161         gets a private slot instead of allocating it on the stack. */
1162    if (space > 0 && cc >= end)    if (space > 0 && cc >= end)
1163      {      {
1164      common->localptrs[cc - common->start] = localptr;      common->private_data_ptrs[cc - common->start] = private_data_ptr;
1165      localptr += sizeof(sljit_w) * space;      private_data_ptr += sizeof(sljit_sw) * space;
1166      }      }
1167    
1168    if (size != 0)    if (size != 0)
# Line 1047  while (cc < ccend) Line 1189  while (cc < ccend)
1189      cc += bracketlen;      cc += bracketlen;
1190      }      }
1191    }    }
1192    *private_data_start = private_data_ptr;
1193  }  }
1194    
1195  /* Returns with -1 if no need for frame. */  /* Returns with a frame_types (always < 0) if no need for frame. */
1196  static int get_framesize(compiler_common *common, pcre_uchar *cc, BOOL recursive)  static int get_framesize(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, BOOL recursive, BOOL* needs_control_head)
1197  {  {
 pcre_uchar *ccend = bracketend(cc);  
1198  int length = 0;  int length = 0;
1199  BOOL possessive = FALSE;  int possessive = 0;
1200    BOOL stack_restore = FALSE;
1201  BOOL setsom_found = recursive;  BOOL setsom_found = recursive;
1202  BOOL setmark_found = recursive;  BOOL setmark_found = recursive;
1203    /* The last capture is a local variable even for recursions. */
1204    BOOL capture_last_found = FALSE;
1205    
1206  if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))  #if defined DEBUG_FORCE_CONTROL_HEAD && DEBUG_FORCE_CONTROL_HEAD
1207    SLJIT_ASSERT(common->control_head_ptr != 0);
1208    *needs_control_head = TRUE;
1209    #else
1210    *needs_control_head = FALSE;
1211    #endif
1212    
1213    if (ccend == NULL)
1214    {    {
1215    length = 3;    ccend = bracketend(cc) - (1 + LINK_SIZE);
1216    possessive = TRUE;    if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))
1217        {
1218        possessive = length = (common->capture_last_ptr != 0) ? 5 : 3;
1219        /* This is correct regardless of common->capture_last_ptr. */
1220        capture_last_found = TRUE;
1221        }
1222      cc = next_opcode(common, cc);
1223    }    }
1224    
 cc = next_opcode(common, cc);  
1225  SLJIT_ASSERT(cc != NULL);  SLJIT_ASSERT(cc != NULL);
1226  while (cc < ccend)  while (cc < ccend)
1227    switch(*cc)    switch(*cc)
1228      {      {
1229      case OP_SET_SOM:      case OP_SET_SOM:
1230      SLJIT_ASSERT(common->has_set_som);      SLJIT_ASSERT(common->has_set_som);
1231        stack_restore = TRUE;
1232      if (!setsom_found)      if (!setsom_found)
1233        {        {
1234        length += 2;        length += 2;
# Line 1080  while (cc < ccend) Line 1238  while (cc < ccend)
1238      break;      break;
1239    
1240      case OP_MARK:      case OP_MARK:
1241        case OP_PRUNE_ARG:
1242        case OP_THEN_ARG:
1243      SLJIT_ASSERT(common->mark_ptr != 0);      SLJIT_ASSERT(common->mark_ptr != 0);
1244        stack_restore = TRUE;
1245      if (!setmark_found)      if (!setmark_found)
1246        {        {
1247        length += 2;        length += 2;
1248        setmark_found = TRUE;        setmark_found = TRUE;
1249        }        }
1250        if (common->control_head_ptr != 0)
1251          *needs_control_head = TRUE;
1252      cc += 1 + 2 + cc[1];      cc += 1 + 2 + cc[1];
1253      break;      break;
1254    
1255      case OP_RECURSE:      case OP_RECURSE:
1256        stack_restore = TRUE;
1257      if (common->has_set_som && !setsom_found)      if (common->has_set_som && !setsom_found)
1258        {        {
1259        length += 2;        length += 2;
# Line 1100  while (cc < ccend) Line 1264  while (cc < ccend)
1264        length += 2;        length += 2;
1265        setmark_found = TRUE;        setmark_found = TRUE;
1266        }        }
1267        if (common->capture_last_ptr != 0 && !capture_last_found)
1268          {
1269          length += 2;
1270          capture_last_found = TRUE;
1271          }
1272      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1273      break;      break;
1274    
# Line 1107  while (cc < ccend) Line 1276  while (cc < ccend)
1276      case OP_CBRAPOS:      case OP_CBRAPOS:
1277      case OP_SCBRA:      case OP_SCBRA:
1278      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1279        stack_restore = TRUE;
1280        if (common->capture_last_ptr != 0 && !capture_last_found)
1281          {
1282          length += 2;
1283          capture_last_found = TRUE;
1284          }
1285      length += 3;      length += 3;
1286      cc += 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1287      break;      break;
1288    
1289      default:      default:
1290        stack_restore = TRUE;
1291        /* Fall through. */
1292    
1293        case OP_NOT_WORD_BOUNDARY:
1294        case OP_WORD_BOUNDARY:
1295        case OP_NOT_DIGIT:
1296        case OP_DIGIT:
1297        case OP_NOT_WHITESPACE:
1298        case OP_WHITESPACE:
1299        case OP_NOT_WORDCHAR:
1300        case OP_WORDCHAR:
1301        case OP_ANY:
1302        case OP_ALLANY:
1303        case OP_ANYBYTE:
1304        case OP_NOTPROP:
1305        case OP_PROP:
1306        case OP_ANYNL:
1307        case OP_NOT_HSPACE:
1308        case OP_HSPACE:
1309        case OP_NOT_VSPACE:
1310        case OP_VSPACE:
1311        case OP_EXTUNI:
1312        case OP_EODN:
1313        case OP_EOD:
1314        case OP_CIRC:
1315        case OP_CIRCM:
1316        case OP_DOLL:
1317        case OP_DOLLM:
1318        case OP_CHAR:
1319        case OP_CHARI:
1320        case OP_NOT:
1321        case OP_NOTI:
1322    
1323        case OP_EXACT:
1324        case OP_POSSTAR:
1325        case OP_POSPLUS:
1326        case OP_POSQUERY:
1327        case OP_POSUPTO:
1328    
1329        case OP_EXACTI:
1330        case OP_POSSTARI:
1331        case OP_POSPLUSI:
1332        case OP_POSQUERYI:
1333        case OP_POSUPTOI:
1334    
1335        case OP_NOTEXACT:
1336        case OP_NOTPOSSTAR:
1337        case OP_NOTPOSPLUS:
1338        case OP_NOTPOSQUERY:
1339        case OP_NOTPOSUPTO:
1340    
1341        case OP_NOTEXACTI:
1342        case OP_NOTPOSSTARI:
1343        case OP_NOTPOSPLUSI:
1344        case OP_NOTPOSQUERYI:
1345        case OP_NOTPOSUPTOI:
1346    
1347        case OP_TYPEEXACT:
1348        case OP_TYPEPOSSTAR:
1349        case OP_TYPEPOSPLUS:
1350        case OP_TYPEPOSQUERY:
1351        case OP_TYPEPOSUPTO:
1352    
1353        case OP_CLASS:
1354        case OP_NCLASS:
1355        case OP_XCLASS:
1356    
1357      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
1358      SLJIT_ASSERT(cc != NULL);      SLJIT_ASSERT(cc != NULL);
1359      break;      break;
1360      }      }
1361    
1362  /* Possessive quantifiers can use a special case. */  /* Possessive quantifiers can use a special case. */
1363  if (SLJIT_UNLIKELY(possessive) && length == 3)  if (SLJIT_UNLIKELY(possessive == length))
1364    return -1;    return stack_restore ? no_frame : no_stack;
1365    
1366  if (length > 0)  if (length > 0)
1367    return length + 1;    return length + 1;
1368  return -1;  return stack_restore ? no_frame : no_stack;
1369  }  }
1370    
1371  static void init_frame(compiler_common *common, pcre_uchar *cc, int stackpos, int stacktop, BOOL recursive)  static void init_frame(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, int stackpos, int stacktop, BOOL recursive)
1372  {  {
1373  DEFINE_COMPILER;  DEFINE_COMPILER;
 pcre_uchar *ccend = bracketend(cc);  
1374  BOOL setsom_found = recursive;  BOOL setsom_found = recursive;
1375  BOOL setmark_found = recursive;  BOOL setmark_found = recursive;
1376    /* The last capture is a local variable even for recursions. */
1377    BOOL capture_last_found = FALSE;
1378  int offset;  int offset;
1379    
1380  /* >= 1 + shortest item size (2) */  /* >= 1 + shortest item size (2) */
# Line 1139  SLJIT_UNUSED_ARG(stacktop); Line 1382  SLJIT_UNUSED_ARG(stacktop);
1382  SLJIT_ASSERT(stackpos >= stacktop + 2);  SLJIT_ASSERT(stackpos >= stacktop + 2);
1383    
1384  stackpos = STACK(stackpos);  stackpos = STACK(stackpos);
1385  if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))  if (ccend == NULL)
1386    cc = next_opcode(common, cc);    {
1387      ccend = bracketend(cc) - (1 + LINK_SIZE);
1388      if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))
1389        cc = next_opcode(common, cc);
1390      }
1391    
1392  SLJIT_ASSERT(cc != NULL);  SLJIT_ASSERT(cc != NULL);
1393  while (cc < ccend)  while (cc < ccend)
1394    switch(*cc)    switch(*cc)
# Line 1150  while (cc < ccend) Line 1398  while (cc < ccend)
1398      if (!setsom_found)      if (!setsom_found)
1399        {        {
1400        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));
1401        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setstrbegin);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1402        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1403        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1404        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1405        setsom_found = TRUE;        setsom_found = TRUE;
1406        }        }
1407      cc += 1;      cc += 1;
1408      break;      break;
1409    
1410      case OP_MARK:      case OP_MARK:
1411        case OP_PRUNE_ARG:
1412        case OP_THEN_ARG:
1413      SLJIT_ASSERT(common->mark_ptr != 0);      SLJIT_ASSERT(common->mark_ptr != 0);
1414      if (!setmark_found)      if (!setmark_found)
1415        {        {
1416        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);
1417        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setmark);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
1418        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1419        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1420        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1421        setmark_found = TRUE;        setmark_found = TRUE;
1422        }        }
1423      cc += 1 + 2 + cc[1];      cc += 1 + 2 + cc[1];
# Line 1177  while (cc < ccend) Line 1427  while (cc < ccend)
1427      if (common->has_set_som && !setsom_found)      if (common->has_set_som && !setsom_found)
1428        {        {
1429        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));
1430        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setstrbegin);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1431        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1432        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1433        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1434        setsom_found = TRUE;        setsom_found = TRUE;
1435        }        }
1436      if (common->mark_ptr != 0 && !setmark_found)      if (common->mark_ptr != 0 && !setmark_found)
1437        {        {
1438        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);
1439        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setmark);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
1440        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1441        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1442        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1443        setmark_found = TRUE;        setmark_found = TRUE;
1444        }        }
1445        if (common->capture_last_ptr != 0 && !capture_last_found)
1446          {
1447          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->capture_last_ptr);
1448          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1449          stackpos += (int)sizeof(sljit_sw);
1450          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1451          stackpos += (int)sizeof(sljit_sw);
1452          capture_last_found = TRUE;
1453          }
1454      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1455      break;      break;
1456    
# Line 1199  while (cc < ccend) Line 1458  while (cc < ccend)
1458      case OP_CBRAPOS:      case OP_CBRAPOS:
1459      case OP_SCBRA:      case OP_SCBRA:
1460      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1461        if (common->capture_last_ptr != 0 && !capture_last_found)
1462          {
1463          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->capture_last_ptr);
1464          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1465          stackpos += (int)sizeof(sljit_sw);
1466          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1467          stackpos += (int)sizeof(sljit_sw);
1468          capture_last_found = TRUE;
1469          }
1470      offset = (GET2(cc, 1 + LINK_SIZE)) << 1;      offset = (GET2(cc, 1 + LINK_SIZE)) << 1;
1471      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));
1472      stackpos += (int)sizeof(sljit_w);      stackpos += (int)sizeof(sljit_sw);
1473      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset));      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset));
1474      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset + 1));      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset + 1));
1475      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1476      stackpos += (int)sizeof(sljit_w);      stackpos += (int)sizeof(sljit_sw);
1477      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);
1478      stackpos += (int)sizeof(sljit_w);      stackpos += (int)sizeof(sljit_sw);
1479    
1480      cc += 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1481      break;      break;
# Line 1218  while (cc < ccend) Line 1486  while (cc < ccend)
1486      break;      break;
1487      }      }
1488    
1489  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_end);  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, 0);
1490  SLJIT_ASSERT(stackpos == STACK(stacktop));  SLJIT_ASSERT(stackpos == STACK(stacktop));
1491  }  }
1492    
1493  static SLJIT_INLINE int get_localsize(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend)  static SLJIT_INLINE int get_private_data_copy_length(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, BOOL needs_control_head)
1494  {  {
1495  int localsize = 2;  int private_data_length = needs_control_head ? 3 : 2;
1496  int size;  int size;
1497  pcre_uchar *alternative;  pcre_uchar *alternative;
1498  /* Calculate the sum of the local variables. */  /* Calculate the sum of the private machine words. */
1499  while (cc < ccend)  while (cc < ccend)
1500    {    {
1501    size = 0;    size = 0;
1502    switch(*cc)    switch(*cc)
1503      {      {
1504        case OP_KET:
1505        if (PRIVATE_DATA(cc) != 0)
1506          private_data_length++;
1507        cc += 1 + LINK_SIZE;
1508        break;
1509    
1510      case OP_ASSERT:      case OP_ASSERT:
1511      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1512      case OP_ASSERTBACK:      case OP_ASSERTBACK:
# Line 1243  while (cc < ccend) Line 1517  while (cc < ccend)
1517      case OP_SBRA:      case OP_SBRA:
1518      case OP_SBRAPOS:      case OP_SBRAPOS:
1519      case OP_SCOND:      case OP_SCOND:
1520      localsize++;      private_data_length++;
1521      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1522      break;      break;
1523    
1524      case OP_CBRA:      case OP_CBRA:
1525      case OP_SCBRA:      case OP_SCBRA:
1526      localsize++;      if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1527          private_data_length++;
1528      cc += 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1529      break;      break;
1530    
1531      case OP_CBRAPOS:      case OP_CBRAPOS:
1532      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1533      localsize += 2;      private_data_length += 2;
1534      cc += 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1535      break;      break;
1536    
# Line 1263  while (cc < ccend) Line 1538  while (cc < ccend)
1538      /* Might be a hidden SCOND. */      /* Might be a hidden SCOND. */
1539      alternative = cc + GET(cc, 1);      alternative = cc + GET(cc, 1);
1540      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1541        localsize++;        private_data_length++;
1542      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1543      break;      break;
1544    
1545      CASE_ITERATOR_LOCAL1      CASE_ITERATOR_PRIVATE_DATA_1
1546      if (PRIV_DATA(cc))      if (PRIVATE_DATA(cc))
1547        localsize++;        private_data_length++;
1548      cc += 2;      cc += 2;
1549  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
1550      if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1551  #endif  #endif
1552      break;      break;
1553    
1554      CASE_ITERATOR_LOCAL2A      CASE_ITERATOR_PRIVATE_DATA_2A
1555      if (PRIV_DATA(cc))      if (PRIVATE_DATA(cc))
1556        localsize += 2;        private_data_length += 2;
1557      cc += 2;      cc += 2;
1558  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
1559      if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1560  #endif  #endif
1561      break;      break;
1562    
1563      CASE_ITERATOR_LOCAL2B      CASE_ITERATOR_PRIVATE_DATA_2B
1564      if (PRIV_DATA(cc))      if (PRIVATE_DATA(cc))
1565        localsize += 2;        private_data_length += 2;
1566      cc += 2 + IMM2_SIZE;      cc += 2 + IMM2_SIZE;
1567  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
1568      if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1569  #endif  #endif
1570      break;      break;
1571    
1572      CASE_ITERATOR_TYPE_LOCAL1      CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1573      if (PRIV_DATA(cc))      if (PRIVATE_DATA(cc))
1574        localsize++;        private_data_length++;
1575      cc += 1;      cc += 1;
1576      break;      break;
1577    
1578      CASE_ITERATOR_TYPE_LOCAL2A      CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1579      if (PRIV_DATA(cc))      if (PRIVATE_DATA(cc))
1580        localsize += 2;        private_data_length += 2;
1581      cc += 1;      cc += 1;
1582      break;      break;
1583    
1584      CASE_ITERATOR_TYPE_LOCAL2B      CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1585      if (PRIV_DATA(cc))      if (PRIVATE_DATA(cc))
1586        localsize += 2;        private_data_length += 2;
1587      cc += 1 + IMM2_SIZE;      cc += 1 + IMM2_SIZE;
1588      break;      break;
1589    
# Line 1320  while (cc < ccend) Line 1595  while (cc < ccend)
1595  #else  #else
1596      size = 1 + 32 / (int)sizeof(pcre_uchar);      size = 1 + 32 / (int)sizeof(pcre_uchar);
1597  #endif  #endif
1598      if (PRIV_DATA(cc))      if (PRIVATE_DATA(cc))
1599        localsize += get_class_iterator_size(cc + size);        private_data_length += get_class_iterator_size(cc + size);
1600      cc += size;      cc += size;
1601      break;      break;
1602    
# Line 1332  while (cc < ccend) Line 1607  while (cc < ccend)
1607      }      }
1608    }    }
1609  SLJIT_ASSERT(cc == ccend);  SLJIT_ASSERT(cc == ccend);
1610  return localsize;  return private_data_length;
1611  }  }
1612    
1613  static void copy_locals(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend,  static void copy_private_data(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend,
1614    BOOL save, int stackptr, int stacktop)    BOOL save, int stackptr, int stacktop, BOOL needs_control_head)
1615  {  {
1616  DEFINE_COMPILER;  DEFINE_COMPILER;
1617  int srcw[2];  int srcw[2];
# Line 1357  stacktop = STACK(stacktop - 1); Line 1632  stacktop = STACK(stacktop - 1);
1632    
1633  if (!save)  if (!save)
1634    {    {
1635    stackptr += sizeof(sljit_w);    stackptr += (needs_control_head ? 2 : 1) * sizeof(sljit_sw);
1636    if (stackptr < stacktop)    if (stackptr < stacktop)
1637      {      {
1638      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1639      stackptr += sizeof(sljit_w);      stackptr += sizeof(sljit_sw);
1640      tmp1empty = FALSE;      tmp1empty = FALSE;
1641      }      }
1642    if (stackptr < stacktop)    if (stackptr < stacktop)
1643      {      {
1644      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1645      stackptr += sizeof(sljit_w);      stackptr += sizeof(sljit_sw);
1646      tmp2empty = FALSE;      tmp2empty = FALSE;
1647      }      }
1648    /* The tmp1next must be TRUE in either way. */    /* The tmp1next must be TRUE in either way. */
1649    }    }
1650    
1651  while (status != end)  do
1652    {    {
1653    count = 0;    count = 0;
1654    switch(status)    switch(status)
1655      {      {
1656      case start:      case start:
1657      SLJIT_ASSERT(save && common->recursive_head != 0);      SLJIT_ASSERT(save && common->recursive_head_ptr != 0);
1658      count = 1;      count = 1;
1659      srcw[0] = common->recursive_head;      srcw[0] = common->recursive_head_ptr;
1660        if (needs_control_head)
1661          {
1662          SLJIT_ASSERT(common->control_head_ptr != 0);
1663          count = 2;
1664          srcw[1] = common->control_head_ptr;
1665          }
1666      status = loop;      status = loop;
1667      break;      break;
1668    
# Line 1394  while (status != end) Line 1675  while (status != end)
1675    
1676      switch(*cc)      switch(*cc)
1677        {        {
1678          case OP_KET:
1679          if (PRIVATE_DATA(cc) != 0)
1680            {
1681            count = 1;
1682            srcw[0] = PRIVATE_DATA(cc);
1683            }
1684          cc += 1 + LINK_SIZE;
1685          break;
1686    
1687        case OP_ASSERT:        case OP_ASSERT:
1688        case OP_ASSERT_NOT:        case OP_ASSERT_NOT:
1689        case OP_ASSERTBACK:        case OP_ASSERTBACK:
# Line 1405  while (status != end) Line 1695  while (status != end)
1695        case OP_SBRAPOS:        case OP_SBRAPOS:
1696        case OP_SCOND:        case OP_SCOND:
1697        count = 1;        count = 1;
1698        srcw[0] = PRIV_DATA(cc);        srcw[0] = PRIVATE_DATA(cc);
1699        SLJIT_ASSERT(srcw[0] != 0);        SLJIT_ASSERT(srcw[0] != 0);
1700        cc += 1 + LINK_SIZE;        cc += 1 + LINK_SIZE;
1701        break;        break;
1702    
1703        case OP_CBRA:        case OP_CBRA:
1704        case OP_SCBRA:        case OP_SCBRA:
1705        count = 1;        if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1706        srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));          {
1707            count = 1;
1708            srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1709            }
1710        cc += 1 + LINK_SIZE + IMM2_SIZE;        cc += 1 + LINK_SIZE + IMM2_SIZE;
1711        break;        break;
1712    
1713        case OP_CBRAPOS:        case OP_CBRAPOS:
1714        case OP_SCBRAPOS:        case OP_SCBRAPOS:
1715        count = 2;        count = 2;
1716        srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));        srcw[0] = PRIVATE_DATA(cc);
1717        srcw[1] = PRIV_DATA(cc);        srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1718        SLJIT_ASSERT(srcw[0] != 0);        SLJIT_ASSERT(srcw[0] != 0 && srcw[1] != 0);
1719        cc += 1 + LINK_SIZE + IMM2_SIZE;        cc += 1 + LINK_SIZE + IMM2_SIZE;
1720        break;        break;
1721    
# Line 1432  while (status != end) Line 1725  while (status != end)
1725        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1726          {          {
1727          count = 1;          count = 1;
1728          srcw[0] = PRIV_DATA(cc);          srcw[0] = PRIVATE_DATA(cc);
1729          SLJIT_ASSERT(srcw[0] != 0);          SLJIT_ASSERT(srcw[0] != 0);
1730          }          }
1731        cc += 1 + LINK_SIZE;        cc += 1 + LINK_SIZE;
1732        break;        break;
1733    
1734        CASE_ITERATOR_LOCAL1        CASE_ITERATOR_PRIVATE_DATA_1
1735        if (PRIV_DATA(cc))        if (PRIVATE_DATA(cc))
1736          {          {
1737          count = 1;          count = 1;
1738          srcw[0] = PRIV_DATA(cc);          srcw[0] = PRIVATE_DATA(cc);
1739          }          }
1740        cc += 2;        cc += 2;
1741  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
# Line 1450  while (status != end) Line 1743  while (status != end)
1743  #endif  #endif
1744        break;        break;
1745    
1746        CASE_ITERATOR_LOCAL2A        CASE_ITERATOR_PRIVATE_DATA_2A
1747        if (PRIV_DATA(cc))        if (PRIVATE_DATA(cc))
1748          {          {
1749          count = 2;          count = 2;
1750          srcw[0] = PRIV_DATA(cc);          srcw[0] = PRIVATE_DATA(cc);
1751          srcw[1] = PRIV_DATA(cc) + sizeof(sljit_w);          srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1752          }          }
1753        cc += 2;        cc += 2;
1754  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
# Line 1463  while (status != end) Line 1756  while (status != end)
1756  #endif  #endif
1757        break;        break;
1758    
1759        CASE_ITERATOR_LOCAL2B        CASE_ITERATOR_PRIVATE_DATA_2B
1760        if (PRIV_DATA(cc))        if (PRIVATE_DATA(cc))
1761          {          {
1762          count = 2;          count = 2;
1763          srcw[0] = PRIV_DATA(cc);          srcw[0] = PRIVATE_DATA(cc);
1764          srcw[1] = PRIV_DATA(cc) + sizeof(sljit_w);          srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1765          }          }
1766        cc += 2 + IMM2_SIZE;        cc += 2 + IMM2_SIZE;
1767  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
# Line 1476  while (status != end) Line 1769  while (status != end)
1769  #endif  #endif
1770        break;        break;
1771    
1772        CASE_ITERATOR_TYPE_LOCAL1        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1773        if (PRIV_DATA(cc))        if (PRIVATE_DATA(cc))
1774          {          {
1775          count = 1;          count = 1;
1776          srcw[0] = PRIV_DATA(cc);          srcw[0] = PRIVATE_DATA(cc);
1777          }          }
1778        cc += 1;        cc += 1;
1779        break;        break;
1780    
1781        CASE_ITERATOR_TYPE_LOCAL2A        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1782        if (PRIV_DATA(cc))        if (PRIVATE_DATA(cc))
1783          {          {
1784          count = 2;          count = 2;
1785          srcw[0] = PRIV_DATA(cc);          srcw[0] = PRIVATE_DATA(cc);
1786          srcw[1] = srcw[0] + sizeof(sljit_w);          srcw[1] = srcw[0] + sizeof(sljit_sw);
1787          }          }
1788        cc += 1;        cc += 1;
1789        break;        break;
1790    
1791        CASE_ITERATOR_TYPE_LOCAL2B        CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1792        if (PRIV_DATA(cc))        if (PRIVATE_DATA(cc))
1793          {          {
1794          count = 2;          count = 2;
1795          srcw[0] = PRIV_DATA(cc);          srcw[0] = PRIVATE_DATA(cc);
1796          srcw[1] = srcw[0] + sizeof(sljit_w);          srcw[1] = srcw[0] + sizeof(sljit_sw);
1797          }          }
1798        cc += 1 + IMM2_SIZE;        cc += 1 + IMM2_SIZE;
1799        break;        break;
# Line 1513  while (status != end) Line 1806  while (status != end)
1806  #else  #else
1807        size = 1 + 32 / (int)sizeof(pcre_uchar);        size = 1 + 32 / (int)sizeof(pcre_uchar);
1808  #endif  #endif
1809        if (PRIV_DATA(cc))        if (PRIVATE_DATA(cc))
1810          switch(get_class_iterator_size(cc + size))          switch(get_class_iterator_size(cc + size))
1811            {            {
1812            case 1:            case 1:
1813            count = 1;            count = 1;
1814            srcw[0] = PRIV_DATA(cc);            srcw[0] = PRIVATE_DATA(cc);
1815            break;            break;
1816    
1817            case 2:            case 2:
1818            count = 2;            count = 2;
1819            srcw[0] = PRIV_DATA(cc);            srcw[0] = PRIVATE_DATA(cc);
1820            srcw[1] = srcw[0] + sizeof(sljit_w);            srcw[1] = srcw[0] + sizeof(sljit_sw);
1821            break;            break;
1822    
1823            default:            default:
# Line 1556  while (status != end) Line 1849  while (status != end)
1849          if (!tmp1empty)          if (!tmp1empty)
1850            {            {
1851            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1852            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1853            }            }
1854          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);
1855          tmp1empty = FALSE;          tmp1empty = FALSE;
# Line 1567  while (status != end) Line 1860  while (status != end)
1860          if (!tmp2empty)          if (!tmp2empty)
1861            {            {
1862            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1863            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1864            }            }
1865          OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);          OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);
1866          tmp2empty = FALSE;          tmp2empty = FALSE;
# Line 1584  while (status != end) Line 1877  while (status != end)
1877          if (!tmp1empty)          if (!tmp1empty)
1878            {            {
1879            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1880            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1881            }            }
1882          tmp1next = FALSE;          tmp1next = FALSE;
1883          }          }
# Line 1596  while (status != end) Line 1889  while (status != end)
1889          if (!tmp2empty)          if (!tmp2empty)
1890            {            {
1891            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1892            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1893            }            }
1894          tmp1next = TRUE;          tmp1next = TRUE;
1895          }          }
1896        }        }
1897      }      }
1898    }    }
1899    while (status != end);
1900    
1901  if (save)  if (save)
1902    {    {
# Line 1611  if (save) Line 1905  if (save)
1905      if (!tmp1empty)      if (!tmp1empty)
1906        {        {
1907        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1908        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1909        }        }
1910      if (!tmp2empty)      if (!tmp2empty)
1911        {        {
1912        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1913        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1914        }        }
1915      }      }
1916    else    else
# Line 1624  if (save) Line 1918  if (save)
1918      if (!tmp2empty)      if (!tmp2empty)
1919        {        {
1920        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1921        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1922        }        }
1923      if (!tmp1empty)      if (!tmp1empty)
1924        {        {
1925        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1926        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1927        }        }
1928      }      }
1929    }    }
1930  SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));  SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));
1931  }  }
1932    
1933  #undef CASE_ITERATOR_LOCAL1  static SLJIT_INLINE pcre_uchar *set_then_offsets(compiler_common *common, pcre_uchar *cc, pcre_uint8 *current_offset)
1934  #undef CASE_ITERATOR_LOCAL2A  {
1935  #undef CASE_ITERATOR_LOCAL2B  pcre_uchar *end = bracketend(cc);
1936  #undef CASE_ITERATOR_TYPE_LOCAL1  BOOL has_alternatives = cc[GET(cc, 1)] == OP_ALT;
 #undef CASE_ITERATOR_TYPE_LOCAL2A  
 #undef CASE_ITERATOR_TYPE_LOCAL2B  
1937    
1938  static SLJIT_INLINE BOOL ispowerof2(unsigned int value)  /* Assert captures then. */
1939    if (*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT)
1940      current_offset = NULL;
1941    /* Conditional block does not. */
1942    if (*cc == OP_COND || *cc == OP_SCOND)
1943      has_alternatives = FALSE;
1944    
1945    cc = next_opcode(common, cc);
1946    if (has_alternatives)
1947      current_offset = common->then_offsets + (cc - common->start);
1948    
1949    while (cc < end)
1950      {
1951      if ((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND))
1952        cc = set_then_offsets(common, cc, current_offset);
1953      else
1954        {
1955        if (*cc == OP_ALT && has_alternatives)
1956          current_offset = common->then_offsets + (cc + 1 + LINK_SIZE - common->start);
1957        if (*cc >= OP_THEN && *cc <= OP_THEN_ARG && current_offset != NULL)
1958          *current_offset = 1;
1959        cc = next_opcode(common, cc);
1960        }
1961      }
1962    
1963    return end;
1964    }
1965    
1966    #undef CASE_ITERATOR_PRIVATE_DATA_1
1967    #undef CASE_ITERATOR_PRIVATE_DATA_2A
1968    #undef CASE_ITERATOR_PRIVATE_DATA_2B
1969    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1970    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1971    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1972    
1973    static SLJIT_INLINE BOOL is_powerof2(unsigned int value)
1974  {  {
1975  return (value & (value - 1)) == 0;  return (value & (value - 1)) == 0;
1976  }  }
# Line 1653  static SLJIT_INLINE void set_jumps(jump_ Line 1980  static SLJIT_INLINE void set_jumps(jump_
1980  while (list)  while (list)
1981    {    {
1982    /* sljit_set_label is clever enough to do nothing    /* sljit_set_label is clever enough to do nothing
1983    if either the jump or the label is NULL */    if either the jump or the label is NULL. */
1984    sljit_set_label(list->jump, label);    SET_LABEL(list->jump, label);
1985    list = list->next;    list = list->next;
1986    }    }
1987  }  }
# Line 1670  if (list_item) Line 1997  if (list_item)
1997    }    }
1998  }  }
1999    
2000  static void add_stub(compiler_common *common, enum stub_types type, int data, struct sljit_jump *start)  static void add_stub(compiler_common *common, struct sljit_jump *start)
2001  {  {
2002  DEFINE_COMPILER;  DEFINE_COMPILER;
2003  stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));  stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));
2004    
2005  if (list_item)  if (list_item)
2006    {    {
   list_item->type = type;  
   list_item->data = data;  
2007    list_item->start = start;    list_item->start = start;
2008    list_item->quit = LABEL();    list_item->quit = LABEL();
2009    list_item->next = common->stubs;    list_item->next = common->stubs;
# Line 1694  stub_list* list_item = common->stubs; Line 2019  stub_list* list_item = common->stubs;
2019  while (list_item)  while (list_item)
2020    {    {
2021    JUMPHERE(list_item->start);    JUMPHERE(list_item->start);
2022    switch(list_item->type)    add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));
     {  
     case stack_alloc:  
     add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));  
     break;  
     }  
2023    JUMPTO(SLJIT_JUMP, list_item->quit);    JUMPTO(SLJIT_JUMP, list_item->quit);
2024    list_item = list_item->next;    list_item = list_item->next;
2025    }    }
2026  common->stubs = NULL;  common->stubs = NULL;
2027  }  }
2028    
2029  static SLJIT_INLINE void decrease_call_count(compiler_common *common)  static SLJIT_INLINE void count_match(compiler_common *common)
2030  {  {
2031  DEFINE_COMPILER;  DEFINE_COMPILER;
2032    
2033  OP2(SLJIT_SUB | SLJIT_SET_E, CALL_COUNT, 0, CALL_COUNT, 0, SLJIT_IMM, 1);  OP2(SLJIT_SUB | SLJIT_SET_E, COUNT_MATCH, 0, COUNT_MATCH, 0, SLJIT_IMM, 1);
2034  add_jump(compiler, &common->calllimit, JUMP(SLJIT_C_ZERO));  add_jump(compiler, &common->calllimit, JUMP(SLJIT_C_ZERO));
2035  }  }
2036    
# Line 1719  static SLJIT_INLINE void allocate_stack( Line 2039  static SLJIT_INLINE void allocate_stack(
2039  /* May destroy all locals and registers except TMP2. */  /* May destroy all locals and registers except TMP2. */
2040  DEFINE_COMPILER;  DEFINE_COMPILER;
2041    
2042  OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_w));  OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_sw));
2043  #ifdef DESTROY_REGISTERS  #ifdef DESTROY_REGISTERS
2044  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);
2045  OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);  OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);
# Line 1727  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0); Line 2047  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);
2047  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);
2048  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);
2049  #endif  #endif
2050  add_stub(common, stack_alloc, 0, CMP(SLJIT_C_GREATER, STACK_TOP, 0, STACK_LIMIT, 0));  add_stub(common, CMP(SLJIT_C_GREATER, STACK_TOP, 0, STACK_LIMIT, 0));
2051  }  }
2052    
2053  static SLJIT_INLINE void free_stack(compiler_common *common, int size)  static SLJIT_INLINE void free_stack(compiler_common *common, int size)
2054  {  {
2055  DEFINE_COMPILER;  DEFINE_COMPILER;
2056  OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_w));  OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_sw));
2057  }  }
2058    
2059  static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)  static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)
# Line 1741  static SLJIT_INLINE void reset_ovector(c Line 2061  static SLJIT_INLINE void reset_ovector(c
2061  DEFINE_COMPILER;  DEFINE_COMPILER;
2062  struct sljit_label *loop;  struct sljit_label *loop;
2063  int i;  int i;
2064    
2065  /* At this point we can freely use all temporary registers. */  /* At this point we can freely use all temporary registers. */
2066    SLJIT_ASSERT(length > 1);
2067  /* TMP1 returns with begin - 1. */  /* TMP1 returns with begin - 1. */
2068  OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG1, 0, SLJIT_MEM1(SLJIT_SAVED_REG1), SLJIT_OFFSETOF(jit_arguments, begin), SLJIT_IMM, IN_UCHARS(1));  OP2(SLJIT_SUB, SLJIT_SCRATCH_REG1, 0, SLJIT_MEM1(SLJIT_SAVED_REG1), SLJIT_OFFSETOF(jit_arguments, begin), SLJIT_IMM, IN_UCHARS(1));
2069    if (length < 8)
2070      {
2071      for (i = 1; i < length; i++)
2072        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), SLJIT_SCRATCH_REG1, 0);
2073      }
2074    else
2075      {
2076      GET_LOCAL_BASE(SLJIT_SCRATCH_REG2, 0, OVECTOR_START);
2077      OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, length - 1);
2078      loop = LABEL();
2079      OP1(SLJIT_MOVU, SLJIT_MEM1(SLJIT_SCRATCH_REG2), sizeof(sljit_sw), SLJIT_SCRATCH_REG1, 0);
2080      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 1);
2081      JUMPTO(SLJIT_C_NOT_ZERO, loop);
2082      }
2083    }
2084    
2085    static SLJIT_INLINE void do_reset_match(compiler_common *common, int length)
2086    {
2087    DEFINE_COMPILER;
2088    struct sljit_label *loop;
2089    int i;
2090    
2091    SLJIT_ASSERT(length > 1);
2092    /* OVECTOR(1) contains the "string begin - 1" constant. */
2093    if (length > 2)
2094      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));
2095  if (length < 8)  if (length < 8)
2096    {    {
2097    for (i = 0; i < length; i++)    for (i = 2; i < length; i++)
2098      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), SLJIT_TEMPORARY_REG1, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), TMP1, 0);
2099    }    }
2100  else  else
2101    {    {
2102    GET_LOCAL_BASE(SLJIT_TEMPORARY_REG2, 0, OVECTOR_START - sizeof(sljit_w));    GET_LOCAL_BASE(TMP2, 0, OVECTOR_START + sizeof(sljit_sw));
2103    OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, length);    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_IMM, length - 2);
2104    loop = LABEL();    loop = LABEL();
2105    OP1(SLJIT_MOVU, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), sizeof(sljit_w), SLJIT_TEMPORARY_REG1, 0);    OP1(SLJIT_MOVU, SLJIT_MEM1(TMP2), sizeof(sljit_sw), TMP1, 0);
2106    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_TEMPORARY_REG3, 0, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, 1);    OP2(SLJIT_SUB | SLJIT_SET_E, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, 1);
2107    JUMPTO(SLJIT_C_NOT_ZERO, loop);    JUMPTO(SLJIT_C_NOT_ZERO, loop);
2108    }    }
2109    
2110    OP1(SLJIT_MOV, STACK_TOP, 0, ARGUMENTS, 0);
2111    if (common->mark_ptr != 0)
2112      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr, SLJIT_IMM, 0);
2113    if (common->control_head_ptr != 0)
2114      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->control_head_ptr, SLJIT_IMM, 0);
2115    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(jit_arguments, stack));
2116    OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_ptr);
2117    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(struct sljit_stack, base));
2118    }
2119    
2120    static sljit_sw SLJIT_CALL do_search_mark(sljit_sw *current, const pcre_uchar *skip_arg)
2121    {
2122    while (current != NULL)
2123      {
2124      switch (current[-2])
2125        {
2126        case type_then_trap:
2127        break;
2128    
2129        case type_mark:
2130        if (STRCMP_UC_UC(skip_arg, (pcre_uchar *)current[-3]) == 0)
2131          return current[-4];
2132        break;
2133    
2134        default:
2135        SLJIT_ASSERT_STOP();
2136        break;
2137        }
2138      current = (sljit_sw*)current[-1];
2139      }
2140    return -1;
2141  }  }
2142    
2143  static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)  static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)
2144  {  {
2145  DEFINE_COMPILER;  DEFINE_COMPILER;
2146  struct sljit_label *loop;  struct sljit_label *loop;
2147  struct sljit_jump *earlyexit;  struct sljit_jump *early_quit;
2148    
2149  /* At this point we can freely use all registers. */  /* At this point we can freely use all registers. */
2150  OP1(SLJIT_MOV, SLJIT_SAVED_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));  OP1(SLJIT_MOV, SLJIT_SAVED_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));
2151  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1), STR_PTR, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1), STR_PTR, 0);
2152    
2153  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, ARGUMENTS, 0);  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG1, 0, ARGUMENTS, 0);
2154  if (common->mark_ptr != 0)  if (common->mark_ptr != 0)
2155    OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);
2156  OP1(SLJIT_MOV_SI, SLJIT_TEMPORARY_REG2, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, offsetcount));  OP1(SLJIT_MOV_SI, SLJIT_SCRATCH_REG2, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, offset_count));
2157  if (common->mark_ptr != 0)  if (common->mark_ptr != 0)
2158    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, mark_ptr), SLJIT_TEMPORARY_REG3, 0);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, mark_ptr), SLJIT_SCRATCH_REG3, 0);
2159  OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));  OP2(SLJIT_SUB, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));
2160  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, begin));  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG1, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, begin));
2161  GET_LOCAL_BASE(SLJIT_SAVED_REG1, 0, OVECTOR_START);  GET_LOCAL_BASE(SLJIT_SAVED_REG1, 0, OVECTOR_START);
2162  /* Unlikely, but possible */  /* Unlikely, but possible */
2163  earlyexit = CMP(SLJIT_C_EQUAL, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, 0);  early_quit = CMP(SLJIT_C_EQUAL, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, 0);
2164  loop = LABEL();  loop = LABEL();
2165  OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, SLJIT_MEM1(SLJIT_SAVED_REG1), 0, SLJIT_TEMPORARY_REG1, 0);  OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, SLJIT_MEM1(SLJIT_SAVED_REG1), 0, SLJIT_SCRATCH_REG1, 0);
2166  OP2(SLJIT_ADD, SLJIT_SAVED_REG1, 0, SLJIT_SAVED_REG1, 0, SLJIT_IMM, sizeof(sljit_w));  OP2(SLJIT_ADD, SLJIT_SAVED_REG1, 0, SLJIT_SAVED_REG1, 0, SLJIT_IMM, sizeof(sljit_sw));
2167  /* Copy the integer value to the output buffer */  /* Copy the integer value to the output buffer */
2168  #ifdef COMPILE_PCRE16  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2169  OP2(SLJIT_ASHR, SLJIT_SAVED_REG2, 0, SLJIT_SAVED_REG2, 0, SLJIT_IMM, 1);  OP2(SLJIT_ASHR, SLJIT_SAVED_REG2, 0, SLJIT_SAVED_REG2, 0, SLJIT_IMM, UCHAR_SHIFT);
2170  #endif  #endif
2171  OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_TEMPORARY_REG3), sizeof(int), SLJIT_SAVED_REG2, 0);  OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG3), sizeof(int), SLJIT_SAVED_REG2, 0);
2172  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_TEMPORARY_REG2, 0, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, 1);  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_SCRATCH_REG2, 0, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, 1);
2173  JUMPTO(SLJIT_C_NOT_ZERO, loop);  JUMPTO(SLJIT_C_NOT_ZERO, loop);
2174  JUMPHERE(earlyexit);  JUMPHERE(early_quit);
2175    
2176  /* Calculate the return value, which is the maximum ovector value. */  /* Calculate the return value, which is the maximum ovector value. */
2177  if (topbracket > 1)  if (topbracket > 1)
2178    {    {
2179    GET_LOCAL_BASE(SLJIT_TEMPORARY_REG1, 0, OVECTOR_START + topbracket * 2 * sizeof(sljit_w));    GET_LOCAL_BASE(SLJIT_SCRATCH_REG1, 0, OVECTOR_START + topbracket * 2 * sizeof(sljit_sw));
2180    OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, topbracket + 1);    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, topbracket + 1);
2181    
2182    /* OVECTOR(0) is never equal to SLJIT_SAVED_REG3. */    /* OVECTOR(0) is never equal to SLJIT_SAVED_REG3. */
2183    loop = LABEL();    loop = LABEL();
2184    OP1(SLJIT_MOVU, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), -(2 * (sljit_w)sizeof(sljit_w)));    OP1(SLJIT_MOVU, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), -(2 * (sljit_sw)sizeof(sljit_sw)));
2185    OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG2, 0, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, 1);    OP2(SLJIT_SUB, SLJIT_SCRATCH_REG2, 0, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, 1);
2186    CMPTO(SLJIT_C_EQUAL, SLJIT_TEMPORARY_REG3, 0, SLJIT_SAVED_REG3, 0, loop);    CMPTO(SLJIT_C_EQUAL, SLJIT_SCRATCH_REG3, 0, SLJIT_SAVED_REG3, 0, loop);
2187    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_TEMPORARY_REG2, 0);    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_SCRATCH_REG2, 0);
2188    }    }
2189  else  else
2190    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, 1);    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, 1);
# Line 1813  else Line 2193  else
2193  static SLJIT_INLINE void return_with_partial_match(compiler_common *common, struct sljit_label *quit)  static SLJIT_INLINE void return_with_partial_match(compiler_common *common, struct sljit_label *quit)
2194  {  {
2195  DEFINE_COMPILER;  DEFINE_COMPILER;
2196    struct sljit_jump *jump;
2197    
2198  SLJIT_COMPILE_ASSERT(STR_END == SLJIT_SAVED_REG2, str_end_must_be_saved_reg2);  SLJIT_COMPILE_ASSERT(STR_END == SLJIT_SAVED_REG2, str_end_must_be_saved_reg2);
2199  SLJIT_ASSERT(common->start_used_ptr != 0 && (common->mode == JIT_PARTIAL_SOFT_COMPILE ? common->hit_start != 0 : common->hit_start == 0));  SLJIT_ASSERT(common->start_used_ptr != 0 && common->start_ptr != 0
2200      && (common->mode == JIT_PARTIAL_SOFT_COMPILE ? common->hit_start != 0 : common->hit_start == 0));
2201    
2202  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG2, 0, ARGUMENTS, 0);  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, ARGUMENTS, 0);
2203  OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, PCRE_ERROR_PARTIAL);  OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, PCRE_ERROR_PARTIAL);
2204  OP1(SLJIT_MOV_SI, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), SLJIT_OFFSETOF(jit_arguments, offsetcount));  OP1(SLJIT_MOV_SI, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, real_offset_count));
2205  CMPTO(SLJIT_C_LESS, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, 2, quit);  CMPTO(SLJIT_C_SIG_LESS, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 2, quit);
2206    
2207  /* Store match begin and end. */  /* Store match begin and end. */
2208  OP1(SLJIT_MOV, SLJIT_SAVED_REG1, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), SLJIT_OFFSETOF(jit_arguments, begin));  OP1(SLJIT_MOV, SLJIT_SAVED_REG1, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, begin));
2209  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG2, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), SLJIT_OFFSETOF(jit_arguments, offsets));  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, offsets));
2210  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mode == JIT_PARTIAL_HARD_COMPILE ? common->start_used_ptr : common->hit_start);  
2211    jump = CMP(SLJIT_C_SIG_LESS, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 3);
2212    OP2(SLJIT_SUB, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mode == JIT_PARTIAL_HARD_COMPILE ? common->start_ptr : (common->hit_start + (int)sizeof(sljit_sw)), SLJIT_SAVED_REG1, 0);
2213    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2214    OP2(SLJIT_ASHR, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, UCHAR_SHIFT);
2215    #endif
2216    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG2), 2 * sizeof(int), SLJIT_SCRATCH_REG3, 0);
2217    JUMPHERE(jump);
2218    
2219    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mode == JIT_PARTIAL_HARD_COMPILE ? common->start_used_ptr : common->hit_start);
2220  OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, STR_END, 0, SLJIT_SAVED_REG1, 0);  OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, STR_END, 0, SLJIT_SAVED_REG1, 0);
2221  #ifdef COMPILE_PCRE16  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2222  OP2(SLJIT_ASHR, SLJIT_SAVED_REG2, 0, SLJIT_SAVED_REG2, 0, SLJIT_IMM, 1);  OP2(SLJIT_ASHR, SLJIT_SAVED_REG2, 0, SLJIT_SAVED_REG2, 0, SLJIT_IMM, UCHAR_SHIFT);
2223  #endif  #endif
2224  OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), sizeof(int), SLJIT_SAVED_REG2, 0);  OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG2), sizeof(int), SLJIT_SAVED_REG2, 0);
2225    
2226  OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG3, 0, SLJIT_TEMPORARY_REG3, 0, SLJIT_SAVED_REG1, 0);  OP2(SLJIT_SUB, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_SAVED_REG1, 0);
2227  #ifdef COMPILE_PCRE16  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2228  OP2(SLJIT_ASHR, SLJIT_TEMPORARY_REG3, 0, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, 1);  OP2(SLJIT_ASHR, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, UCHAR_SHIFT);
2229  #endif  #endif
2230  OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), 0, SLJIT_TEMPORARY_REG3, 0);  OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG2), 0, SLJIT_SCRATCH_REG3, 0);
2231    
2232  JUMPTO(SLJIT_JUMP, quit);  JUMPTO(SLJIT_JUMP, quit);
2233  }  }
# Line 1949  if (c <= 127 && bit == 0x20) Line 2340  if (c <= 127 && bit == 0x20)
2340    return (0 << 8) | 0x20;    return (0 << 8) | 0x20;
2341    
2342  /* Since c != oc, they must have at least 1 bit difference. */  /* Since c != oc, they must have at least 1 bit difference. */
2343  if (!ispowerof2(bit))  if (!is_powerof2(bit))
2344    return 0;    return 0;
2345    
2346  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
2347    
2348  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
2349  if (common->utf && c > 127)  if (common->utf && c > 127)
# Line 1968  if (common->utf && c > 127) Line 2359  if (common->utf && c > 127)
2359  #endif /* SUPPORT_UTF */  #endif /* SUPPORT_UTF */
2360  return (0 << 8) | bit;  return (0 << 8) | bit;
2361    
2362  #else /* COMPILE_PCRE8 */  #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2363    
 #ifdef COMPILE_PCRE16  
2364  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
2365  if (common->utf && c > 65535)  if (common->utf && c > 65535)
2366    {    {
# Line 1981  if (common->utf && c > 65535) Line 2371  if (common->utf && c > 65535)
2371    }    }
2372  #endif /* SUPPORT_UTF */  #endif /* SUPPORT_UTF */
2373  return (bit < 256) ? ((0 << 8) | bit) : ((1 << 8) | (bit >> 8));  return (bit < 256) ? ((0 << 8) | bit) : ((1 << 8) | (bit >> 8));
 #endif /* COMPILE_PCRE16 */  
2374    
2375  #endif /* COMPILE_PCRE8 */  #endif /* COMPILE_PCRE[8|16|32] */
2376  }  }
2377    
2378  static void check_partial(compiler_common *common, BOOL force)  static void check_partial(compiler_common *common, BOOL force)
2379  {  {
2380  /* Checks whether a partial matching is occured. Does not modify registers. */  /* Checks whether a partial matching is occurred. Does not modify registers. */
2381  DEFINE_COMPILER;  DEFINE_COMPILER;
2382  struct sljit_jump *jump = NULL;  struct sljit_jump *jump = NULL;
2383    
# Line 2003  else if (common->mode == JIT_PARTIAL_SOF Line 2392  else if (common->mode == JIT_PARTIAL_SOF
2392    jump = CMP(SLJIT_C_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, SLJIT_IMM, -1);    jump = CMP(SLJIT_C_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, SLJIT_IMM, -1);
2393    
2394  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2395    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, -1);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2396  else  else
2397    {    {
2398    if (common->partialmatchlabel != NULL)    if (common->partialmatchlabel != NULL)
# Line 2016  if (jump != NULL) Line 2405  if (jump != NULL)
2405    JUMPHERE(jump);    JUMPHERE(jump);
2406  }  }
2407    
2408  static struct sljit_jump *check_str_end(compiler_common *common)  static void check_str_end(compiler_common *common, jump_list **end_reached)
2409  {  {
2410  /* Does not affect registers. Usually used in a tight spot. */  /* Does not affect registers. Usually used in a tight spot. */
2411  DEFINE_COMPILER;  DEFINE_COMPILER;
2412  struct sljit_jump *jump;  struct sljit_jump *jump;
 struct sljit_jump *nohit;  
 struct sljit_jump *return_value;  
2413    
2414  if (common->mode == JIT_COMPILE)  if (common->mode == JIT_COMPILE)
2415    return CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);    {
2416      add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2417      return;
2418      }
2419    
2420  jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);  jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);
2421  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2422    {    {
2423    nohit = CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);    add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));
2424    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, -1);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2425    JUMPHERE(nohit);    add_jump(compiler, end_reached, JUMP(SLJIT_JUMP));
   return_value = JUMP(SLJIT_JUMP);  
2426    }    }
2427  else  else
2428    {    {
2429    return_value = CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);    add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));
2430    if (common->partialmatchlabel != NULL)    if (common->partialmatchlabel != NULL)
2431      JUMPTO(SLJIT_JUMP, common->partialmatchlabel);      JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2432    else    else
2433      add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));      add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2434    }    }
2435  JUMPHERE(jump);  JUMPHERE(jump);
 return return_value;  
2436  }  }
2437    
2438  static void detect_partial_match(compiler_common *common, jump_list **backtracks)  static void detect_partial_match(compiler_common *common, jump_list **backtracks)
# Line 2063  jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR Line 2451  jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR
2451  add_jump(compiler, backtracks, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));  add_jump(compiler, backtracks, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));
2452  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2453    {    {
2454    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, -1);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2455    add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));    add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));
2456    }    }
2457  else  else
2458    {    {
2459    if (common->partialmatchlabel != NULL)    if (common->partialmatchlabel != NULL)
2460      JUMPTO(SLJIT_JUMP, common->partialmatchlabel);      JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2461    else    else
2462      add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));      add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2463      }
2464    JUMPHERE(jump);
2465    }
2466    
2467    static void peek_char(compiler_common *common)
2468    {
2469    /* Reads the character into TMP1, keeps STR_PTR.
2470    Does not check STR_END. TMP2 Destroyed. */
2471    DEFINE_COMPILER;
2472    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2473    struct sljit_jump *jump;
2474    #endif
2475    
2476    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2477    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2478    if (common->utf)
2479      {
2480      jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2481      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2482      add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2483      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2484      JUMPHERE(jump);
2485      }
2486    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2487    
2488    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2489    if (common->utf)
2490      {
2491      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2492      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2493      /* TMP2 contains the high surrogate. */
2494      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2495      OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2496      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2497      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2498      OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2499      JUMPHERE(jump);
2500      }
2501    #endif
2502    }
2503    
2504    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2505    
2506    static BOOL is_char7_bitset(const pcre_uint8* bitset, BOOL nclass)
2507    {
2508    /* Tells whether the character codes below 128 are enough
2509    to determine a match. */
2510    const pcre_uint8 value = nclass ? 0xff : 0;
2511    const pcre_uint8* end = bitset + 32;
2512    
2513    bitset += 16;
2514    do
2515      {
2516      if (*bitset++ != value)
2517        return FALSE;
2518      }
2519    while (bitset < end);
2520    return TRUE;
2521    }
2522    
2523    static void read_char7_type(compiler_common *common, BOOL full_read)
2524    {
2525    /* Reads the precise character type of a character into TMP1, if the character
2526    is less than 128. Otherwise it returns with zero. Does not check STR_END. The
2527    full_read argument tells whether characters above max are accepted or not. */
2528    DEFINE_COMPILER;
2529    struct sljit_jump *jump;
2530    
2531    SLJIT_ASSERT(common->utf);
2532    
2533    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2534    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2535    
2536    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2537    
2538    if (full_read)
2539      {
2540      jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2541      OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2542      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2543      JUMPHERE(jump);
2544    }    }
 JUMPHERE(jump);  
2545  }  }
2546    
2547  static void read_char(compiler_common *common)  #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2548    
2549    static void read_char_max(compiler_common *common, pcre_uint32 max, BOOL full_read)
2550  {  {
2551  /* Reads the character into TMP1, updates STR_PTR.  /* Reads the precise value of a character into TMP1, if the character is
2552  Does not check STR_END. TMP2 Destroyed. */  less than or equal to max. Otherwise it returns with a value greater than max.
2553    Does not check STR_END. The full_read argument tells whether characters above
2554    max are accepted or not. */
2555  DEFINE_COMPILER;  DEFINE_COMPILER;
2556  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2557  struct sljit_jump *jump;  struct sljit_jump *jump;
2558  #endif  #endif
2559    
2560    SLJIT_UNUSED_ARG(full_read);
2561    SLJIT_UNUSED_ARG(max);
2562    
2563  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2564  #ifdef SUPPORT_UTF  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2565    
2566    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2567  if (common->utf)  if (common->utf)
2568    {    {
2569  #ifdef COMPILE_PCRE8    if (max < 128 && !full_read)
2570        return;
2571    
2572    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2573  #else    if (max >= 0x800)
2574  #ifdef COMPILE_PCRE16      add_jump(compiler, (max < 0x10000) ? &common->utfreadchar16 : &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2575    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);    else if (max < 128)
2576  #endif      {
2577  #endif /* COMPILE_PCRE8 */      OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2578    add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2579        }
2580      else
2581        {
2582        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2583        if (!full_read)
2584          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2585        else
2586          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2587        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2588        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2589        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2590        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2591        if (full_read)
2592          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2593        }
2594    JUMPHERE(jump);    JUMPHERE(jump);
2595    }    }
2596  #endif  #endif
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  
 }  
   
 static void peek_char(compiler_common *common)  
 {  
 /* Reads the character into TMP1, keeps STR_PTR.  
 Does not check STR_END. TMP2 Destroyed. */  
 DEFINE_COMPILER;  
 #ifdef SUPPORT_UTF  
 struct sljit_jump *jump;  
 #endif  
2597    
2598  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  #if defined SUPPORT_UTF && defined COMPILE_PCRE16
 #ifdef SUPPORT_UTF  
2599  if (common->utf)  if (common->utf)
2600    {    {
2601  #ifdef COMPILE_PCRE8    if (max >= 0x10000)
2602    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);      {
2603  #else      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2604  #ifdef COMPILE_PCRE16      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2605    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);      /* TMP2 contains the high surrogate. */
2606  #endif      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2607  #endif /* COMPILE_PCRE8 */      OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2608    add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2609    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2610        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2611        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2612        JUMPHERE(jump);
2613        return;
2614        }
2615    
2616      if (max < 0xd800 && !full_read)
2617        return;
2618    
2619      /* Skip low surrogate if necessary. */
2620      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2621      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2622      if (full_read)
2623        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2624      if (max >= 0xd800)
2625        OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0x10000);
2626    JUMPHERE(jump);    JUMPHERE(jump);
2627    }    }
2628  #endif  #endif
2629  }  }
2630    
2631  static void read_char8_type(compiler_common *common)  static SLJIT_INLINE void read_char(compiler_common *common)
2632    {
2633    read_char_max(common, READ_CHAR_ANY, TRUE);
2634    }
2635    
2636    static void read_char8_type(compiler_common *common, BOOL full_read)
2637  {  {
2638  /* Reads the character type into TMP1, updates STR_PTR. Does not check STR_END. */  /* Reads the character type into TMP1, updates STR_PTR. Does not check STR_END.
2639    The full_read argument tells whether characters above max are accepted or not. */
2640  DEFINE_COMPILER;  DEFINE_COMPILER;
2641  #if defined SUPPORT_UTF || defined COMPILE_PCRE16  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2642  struct sljit_jump *jump;  struct sljit_jump *jump;
2643  #endif  #endif
2644    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2645    struct sljit_jump *jump2;
2646    #endif
2647    
2648  #ifdef SUPPORT_UTF  SLJIT_UNUSED_ARG(full_read);
2649    
2650    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2651    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2652    
2653    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2654  if (common->utf)  if (common->utf)
2655    {    {
   OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);  
   OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  
 #ifdef COMPILE_PCRE8  
2656    /* This can be an extra read in some situations, but hopefully    /* This can be an extra read in some situations, but hopefully
2657    it is needed in most cases. */    it is needed in most cases. */
2658    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2659    jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);    jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2660    add_jump(compiler, &common->utfreadtype8, JUMP(SLJIT_FAST_CALL));    if (!full_read)
2661    JUMPHERE(jump);      {
2662  #else      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2663  #ifdef COMPILE_PCRE16      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2664    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);      OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2665    jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2666    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2667        OP2(SLJIT_OR, TMP2, 0, TMP2, 0, TMP1, 0);
2668        OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2669        jump2 = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);
2670        OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2671        JUMPHERE(jump2);
2672        }
2673      else
2674        add_jump(compiler, &common->utfreadtype8, JUMP(SLJIT_FAST_CALL));
2675    JUMPHERE(jump);    JUMPHERE(jump);
   /* Skip low surrogate if necessary. */  
   OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0xfc00);  
   OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0xd800);  
   COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_EQUAL);  
   OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 1);  
   OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);  
 #endif  
 #endif /* COMPILE_PCRE8 */  
2676    return;    return;
2677    }    }
2678  #endif  #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2679  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);  
2680  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  #if !defined COMPILE_PCRE8
 #ifdef COMPILE_PCRE16  
2681  /* The ctypes array contains only 256 values. */  /* The ctypes array contains only 256 values. */
2682  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2683  jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);  jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);
2684  #endif  #endif
2685  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2686  #ifdef COMPILE_PCRE16  #if !defined COMPILE_PCRE8
2687  JUMPHERE(jump);  JUMPHERE(jump);
2688  #endif  #endif
2689    
2690    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2691    if (common->utf && full_read)
2692      {
2693      /* Skip low surrogate if necessary. */
2694      OP2(SLJIT_SUB, TMP2, 0, TMP2, 0, SLJIT_IMM, 0xd800);
2695      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2696      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2697      JUMPHERE(jump);
2698      }
2699    #endif /* SUPPORT_UTF && COMPILE_PCRE16 */
2700  }  }
2701    
2702  static void skip_char_back(compiler_common *common)  static void skip_char_back(compiler_common *common)
2703  {  {
2704  /* Goes one character back. Affects STR_PTR and TMP1. Does not check begin. */  /* Goes one character back. Affects STR_PTR and TMP1. Does not check begin. */
2705  DEFINE_COMPILER;  DEFINE_COMPILER;
2706  #if defined SUPPORT_UTF && defined COMPILE_PCRE8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2707    #if defined COMPILE_PCRE8
2708  struct sljit_label *label;  struct sljit_label *label;
2709    
2710  if (common->utf)  if (common->utf)
# Line 2196  if (common->utf) Line 2716  if (common->utf)
2716    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, 0x80, label);    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, 0x80, label);
2717    return;    return;
2718    }    }
2719  #endif  #elif defined COMPILE_PCRE16
 #if defined SUPPORT_UTF && defined COMPILE_PCRE16  
2720  if (common->utf)  if (common->utf)
2721    {    {
2722    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
# Line 2205  if (common->utf) Line 2724  if (common->utf)
2724    /* Skip low surrogate if necessary. */    /* Skip low surrogate if necessary. */
2725    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
2726    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xdc00);    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xdc00);
2727    COND_VALUE(SLJIT_MOV, TMP1, 0, SLJIT_C_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
2728    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
2729    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2730    return;    return;
2731    }    }
2732  #endif  #endif /* COMPILE_PCRE[8|16] */
2733    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2734  OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2735  }  }
2736    
2737  static void check_newlinechar(compiler_common *common, int nltype, jump_list **backtracks, BOOL jumpiftrue)  static void check_newlinechar(compiler_common *common, int nltype, jump_list **backtracks, BOOL jumpifmatch)
2738  {  {
2739  /* Character comes in TMP1. Checks if it is a newline. TMP2 may be destroyed. */  /* Character comes in TMP1. Checks if it is a newline. TMP2 may be destroyed. */
2740  DEFINE_COMPILER;  DEFINE_COMPILER;
2741    struct sljit_jump *jump;
2742    
2743  if (nltype == NLTYPE_ANY)  if (nltype == NLTYPE_ANY)
2744    {    {
2745    add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));
2746    add_jump(compiler, backtracks, JUMP(jumpiftrue ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));    add_jump(compiler, backtracks, JUMP(jumpifmatch ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));
2747    }    }
2748  else if (nltype == NLTYPE_ANYCRLF)  else if (nltype == NLTYPE_ANYCRLF)
2749    {    {
2750    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_CR);    if (jumpifmatch)
2751    COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_EQUAL);      {
2752    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_NL);      add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR));
2753    COND_VALUE(SLJIT_OR | SLJIT_SET_E, TMP2, 0, SLJIT_C_EQUAL);      add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL));
2754    add_jump(compiler, backtracks, JUMP(jumpiftrue ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));      }
2755      else
2756        {
2757        jump = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR);
2758        add_jump(compiler, backtracks, CMP(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL));
2759        JUMPHERE(jump);
2760        }
2761    }    }
2762  else  else
2763    {    {
2764    SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline < 256);    SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline < 256);
2765    add_jump(compiler, backtracks, CMP(jumpiftrue ? SLJIT_C_EQUAL : SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, common->newline));    add_jump(compiler, backtracks, CMP(jumpifmatch ? SLJIT_C_EQUAL : SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, common->newline));
2766    }    }
2767  }  }
2768    
2769  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
2770    
2771  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
2772  static void do_utfreadchar(compiler_common *common)  static void do_utfreadchar(compiler_common *common)
2773  {  {
2774  /* Fast decoding a UTF-8 character. TMP1 contains the first byte  /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2775  of the character (>= 0xc0). Return char value in TMP1, length - 1 in TMP2. */  of the character (>= 0xc0). Return char value in TMP1, length in TMP2. */
2776  DEFINE_COMPILER;  DEFINE_COMPILER;
2777  struct sljit_jump *jump;  struct sljit_jump *jump;
2778    
2779  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2780    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2781    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2782    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2783    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2784    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2785    
2786  /* Searching for the first zero. */  /* Searching for the first zero. */
2787  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x20);  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
2788  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_C_NOT_ZERO);
2789  /* Two byte sequence. */  /* Two byte sequence. */
 OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  
2790  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2791  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x1f);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(2));
2792    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2793    
2794    JUMPHERE(jump);
2795    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2796    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
2797  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2798  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2799  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
 OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(1));  
 sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  
 JUMPHERE(jump);  
2800    
2801  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x10);  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2802  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_C_NOT_ZERO);
2803  /* Three byte sequence. */  /* Three byte sequence. */
 OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  
 OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x0f);  
 OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 12);  
 OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  
 OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
 OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));  
2804  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2805  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(3));
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
 OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(2));  
2806  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
 JUMPHERE(jump);  
2807    
2808  /* Four byte sequence. */  /* Four byte sequence. */
2809  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  JUMPHERE(jump);
2810  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x07);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2811  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 18);  OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2812    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2813    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2814  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
 OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 12);  
2815  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2816  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(4));
2817    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2818    }
2819    
2820    static void do_utfreadchar16(compiler_common *common)
2821    {
2822    /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2823    of the character (>= 0xc0). Return value in TMP1. */
2824    DEFINE_COMPILER;
2825    struct sljit_jump *jump;
2826    
2827    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2828    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2829    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2830    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2831  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
 OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  
2832  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2833  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(3));  
2834  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));  /* Searching for the first zero. */
2835    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
2836    jump = JUMP(SLJIT_C_NOT_ZERO);
2837    /* Two byte sequence. */
2838    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2839    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2840    
2841    JUMPHERE(jump);
2842    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x400);
2843    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_NOT_ZERO);
2844    /* This code runs only in 8 bit mode. No need to shift the value. */
2845    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2846    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2847    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
2848    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2849  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2850  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2851  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(3));  /* Three byte sequence. */
2852    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2853  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2854  }  }
2855    
# Line 2316  jump = JUMP(SLJIT_C_NOT_ZERO); Line 2869  jump = JUMP(SLJIT_C_NOT_ZERO);
2869  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2870  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2871  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);
2872    /* The upper 5 bits are known at this point. */
2873    compare = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0x3);
2874  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2875  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2876  OP2(SLJIT_OR, TMP2, 0, TMP2, 0, TMP1, 0);  OP2(SLJIT_OR, TMP2, 0, TMP2, 0, TMP1, 0);
 compare = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);  
2877  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2878  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2879    
2880  JUMPHERE(compare);  JUMPHERE(compare);
2881  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2882  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
 JUMPHERE(jump);  
2883    
2884  /* We only have types for characters less than 256. */  /* We only have types for characters less than 256. */
 OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), (sljit_w)PRIV(utf8_table4) - 0xc0);  
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);  
 OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);  
 sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  
 }  
   
 #else /* COMPILE_PCRE8 */  
   
 #ifdef COMPILE_PCRE16  
 static void do_utfreadchar(compiler_common *common)  
 {  
 /* Fast decoding a UTF-16 character. TMP1 contains the first 16 bit char  
 of the character (>= 0xd800). Return char value in TMP1, length - 1 in TMP2. */  
 DEFINE_COMPILER;  
 struct sljit_jump *jump;  
   
 sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);  
 jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xdc00);  
 /* Do nothing, only return. */  
 sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  
   
2885  JUMPHERE(jump);  JUMPHERE(jump);
2886  /* Combine two 16 bit characters. */  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2887  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2888  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
 OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);  
 OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 10);  
 OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3ff);  
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
 OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(1));  
 OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x10000);  
2889  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2890  }  }
 #endif /* COMPILE_PCRE16 */  
2891    
2892  #endif /* COMPILE_PCRE8 */  #endif /* COMPILE_PCRE8 */
2893    
# Line 2384  SLJIT_ASSERT(UCD_BLOCK_SIZE == 128 && si Line 2909  SLJIT_ASSERT(UCD_BLOCK_SIZE == 128 && si
2909    
2910  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2911  OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);  OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
2912  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_w)PRIV(ucd_stage1));  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(ucd_stage1));
2913  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);
2914  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
2915  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, TMP2, 0);
2916  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, (sljit_w)PRIV(ucd_stage2));  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_stage2));
2917  OP1(SLJIT_MOV_UH, TMP2, 0, SLJIT_MEM2(TMP2, TMP1), 1);  OP1(SLJIT_MOV_UH, TMP2, 0, SLJIT_MEM2(TMP2, TMP1), 1);
2918  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, (sljit_w)PRIV(ucd_records) + SLJIT_OFFSETOF(ucd_record, chartype));  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_records) + SLJIT_OFFSETOF(ucd_record, chartype));
2919  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM2(TMP1, TMP2), 3);  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM2(TMP1, TMP2), 3);
2920  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2921  }  }
# Line 2404  struct sljit_label *newlinelabel = NULL; Line 2929  struct sljit_label *newlinelabel = NULL;
2929  struct sljit_jump *start;  struct sljit_jump *start;
2930  struct sljit_jump *end = NULL;  struct sljit_jump *end = NULL;
2931  struct sljit_jump *nl = NULL;  struct sljit_jump *nl = NULL;
2932  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2933  struct sljit_jump *singlechar;  struct sljit_jump *singlechar;
2934  #endif  #endif
2935  jump_list *newline = NULL;  jump_list *newline = NULL;
# Line 2420  if (firstline) Line 2945  if (firstline)
2945    /* Search for the end of the first line. */    /* Search for the end of the first line. */
2946    SLJIT_ASSERT(common->first_line_end != 0);    SLJIT_ASSERT(common->first_line_end != 0);
2947    OP1(SLJIT_MOV, TMP3, 0, STR_PTR, 0);    OP1(SLJIT_MOV, TMP3, 0, STR_PTR, 0);
   OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_END, 0);  
2948    
2949    if (common->nltype == NLTYPE_FIXED && common->newline > 255)    if (common->nltype == NLTYPE_FIXED && common->newline > 255)
2950      {      {
# Line 2431  if (firstline) Line 2955  if (firstline)
2955      OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));      OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2956      CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, mainloop);      CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, mainloop);
2957      CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, mainloop);      CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, mainloop);
2958        JUMPHERE(end);
2959      OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));      OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2960      }      }
2961    else    else
# Line 2439  if (firstline) Line 2964  if (firstline)
2964      mainloop = LABEL();      mainloop = LABEL();
2965      /* Continual stores does not cause data dependency. */      /* Continual stores does not cause data dependency. */
2966      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);
2967      read_char(common);      read_char_max(common, common->nlmax, TRUE);
2968      check_newlinechar(common, common->nltype, &newline, TRUE);      check_newlinechar(common, common->nltype, &newline, TRUE);
2969      CMPTO(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0, mainloop);      CMPTO(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0, mainloop);
2970        JUMPHERE(end);
2971      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);
2972      set_jumps(newline, LABEL());      set_jumps(newline, LABEL());
2973      }      }
2974    
   JUMPHERE(end);  
2975    OP1(SLJIT_MOV, STR_PTR, 0, TMP3, 0);    OP1(SLJIT_MOV, STR_PTR, 0, TMP3, 0);
2976    }    }
2977    
# Line 2459  if (newlinecheck) Line 2984  if (newlinecheck)
2984    end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);    end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
2985    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2986    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, common->newline & 0xff);    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, common->newline & 0xff);
2987    COND_VALUE(SLJIT_MOV, TMP1, 0, SLJIT_C_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
2988  #ifdef COMPILE_PCRE16  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2989    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, UCHAR_SHIFT);
2990  #endif  #endif
2991    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2992    nl = JUMP(SLJIT_JUMP);    nl = JUMP(SLJIT_JUMP);
# Line 2482  if (newlinecheck) Line 3007  if (newlinecheck)
3007    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, newlinelabel);    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, newlinelabel);
3008    
3009  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3010  #if defined SUPPORT_UTF && defined COMPILE_PCRE8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3011    #if defined COMPILE_PCRE8
3012  if (common->utf)  if (common->utf)
3013    {    {
3014    singlechar = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);    singlechar = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
3015    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_w)PRIV(utf8_table4) - 0xc0);    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
3016    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3017    JUMPHERE(singlechar);    JUMPHERE(singlechar);
3018    }    }
3019  #endif  #elif defined COMPILE_PCRE16
 #if defined SUPPORT_UTF && defined COMPILE_PCRE16  
3020  if (common->utf)  if (common->utf)
3021    {    {
3022    singlechar = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);    singlechar = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);
3023    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
3024    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);
3025    COND_VALUE(SLJIT_MOV, TMP1, 0, SLJIT_C_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
3026    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
3027    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3028    JUMPHERE(singlechar);    JUMPHERE(singlechar);
3029    }    }
3030  #endif  #endif /* COMPILE_PCRE[8|16] */
3031    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
3032  JUMPHERE(start);  JUMPHERE(start);
3033    
3034  if (newlinecheck)  if (newlinecheck)
# Line 2514  if (newlinecheck) Line 3040  if (newlinecheck)
3040  return mainloop;  return mainloop;
3041  }  }
3042    
3043  static SLJIT_INLINE BOOL fast_forward_first_two_chars(compiler_common *common, BOOL firstline)  static int scan_prefix(compiler_common *common, pcre_uchar *cc, pcre_uint32 *chars, int max_chars)
3044  {  {
3045  DEFINE_COMPILER;  /* Recursive function, which scans prefix literals. */
3046  struct sljit_label *start;  int len, repeat, len_save, consumed = 0;
3047  struct sljit_jump *quit;  pcre_uint32 caseless, chr, mask;
3048  struct sljit_jump *found;  pcre_uchar *alternative, *cc_save;
3049  pcre_int32 chars[4];  BOOL last, any;
 pcre_uchar *cc = common->start + 1 + IMM2_SIZE;  
 int location = 0;  
 pcre_int32 len, c, bit, caseless;  
 BOOL must_end;  
   
 #ifdef COMPILE_PCRE8  
 union {  
     sljit_uh ascombined;  
     sljit_ub asuchars[2];  
 } pair;  
 #else  
 union {  
     sljit_ui ascombined;  
     sljit_uh asuchars[2];  
 } pair;  
 #endif  
   
 if (*(common->start + GET(common->start, 1)) == OP_ALT)  
   return FALSE;  
3050    
3051    repeat = 1;
3052  while (TRUE)  while (TRUE)
3053    {    {
3054      last = TRUE;
3055      any = FALSE;
3056    caseless = 0;    caseless = 0;
3057    must_end = TRUE;    switch (*cc)
   switch(*cc)  
3058      {      {
3059        case OP_CHARI:
3060        caseless = 1;
3061      case OP_CHAR:      case OP_CHAR:
3062      must_end = FALSE;      last = FALSE;
3063      cc++;      cc++;
3064      break;      break;
3065    
3066      case OP_CHARI:      case OP_SOD:
3067        case OP_SOM:
3068        case OP_SET_SOM:
3069        case OP_NOT_WORD_BOUNDARY:
3070        case OP_WORD_BOUNDARY:
3071        case OP_EODN:
3072        case OP_EOD:
3073        case OP_CIRC:
3074        case OP_CIRCM:
3075        case OP_DOLL:
3076        case OP_DOLLM:
3077        /* Zero width assertions. */
3078        cc++;
3079        continue;
3080    
3081        case OP_PLUS:
3082        case OP_MINPLUS:
3083        case OP_POSPLUS:
3084        cc++;
3085        break;
3086    
3087        case OP_EXACTI:
3088        caseless = 1;
3089        case OP_EXACT:
3090        repeat = GET2(cc, 1);
3091        last = FALSE;
3092        cc += 1 + IMM2_SIZE;
3093        break;
3094    
3095        case OP_PLUSI:
3096        case OP_MINPLUSI:
3097        case OP_POSPLUSI:
3098      caseless = 1;      caseless = 1;
     must_end = FALSE;  
3099      cc++;      cc++;
3100      break;      break;
3101    
3102      case OP_SOD:      case OP_KET:
3103      case OP_SOM:      cc += 1 + LINK_SIZE;
3104      case OP_SET_SOM:      continue;
3105      case OP_NOT_WORD_BOUNDARY:  
3106      case OP_WORD_BOUNDARY:      case OP_ALT:
3107      case OP_EODN:      cc += GET(cc, 1);
3108      case OP_EOD:      continue;
3109      case OP_CIRC:  
3110      case OP_CIRCM:      case OP_ONCE:
3111      case OP_DOLL:      case OP_ONCE_NC:
3112      case OP_DOLLM:      case OP_BRA:
3113      /* Zero width assertions. */      case OP_BRAPOS:
3114      cc++;      case OP_CBRA:
3115        case OP_CBRAPOS:
3116        alternative = cc + GET(cc, 1);
3117        while (*alternative == OP_ALT)
3118          {
3119          max_chars = scan_prefix(common, alternative + 1 + LINK_SIZE, chars, max_chars);
3120          if (max_chars == 0)
3121            return consumed;
3122          alternative += GET(alternative, 1);
3123          }
3124    
3125        if (*cc == OP_CBRA || *cc == OP_CBRAPOS)
3126          cc += IMM2_SIZE;
3127        cc += 1 + LINK_SIZE;
3128      continue;      continue;
3129    
3130      case OP_PLUS:      case OP_CLASS:
3131      case OP_MINPLUS:      case OP_NCLASS:
3132      case OP_POSPLUS:      any = TRUE;
3133      cc++;      cc += 1 + 32 / sizeof(pcre_uchar);
3134      break;      break;
3135    
3136      case OP_EXACT:  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3137      cc += 1 + IMM2_SIZE;      case OP_XCLASS:
3138        any = TRUE;
3139        cc += GET(cc, 1);
3140      break;      break;
3141    #endif
3142    
3143      case OP_PLUSI:      case OP_NOT_DIGIT:
3144      case OP_MINPLUSI:      case OP_DIGIT:
3145      case OP_POSPLUSI:      case OP_NOT_WHITESPACE:
3146      caseless = 1;      case OP_WHITESPACE:
3147        case OP_NOT_WORDCHAR:
3148        case OP_WORDCHAR:
3149        case OP_ANY:
3150        case OP_ALLANY:
3151        any = TRUE;
3152      cc++;      cc++;
3153      break;      break;
3154    
3155      case OP_EXACTI:  #ifdef SUPPORT_UCP
3156      caseless = 1;      case OP_NOTPROP:
3157      cc += 1 + IMM2_SIZE;      case OP_PROP:
3158        any = TRUE;
3159        cc += 1 + 2;
3160      break;      break;
3161    #endif
3162    
3163        case OP_TYPEEXACT:
3164        repeat = GET2(cc, 1);
3165        cc += 1 + IMM2_SIZE;
3166        continue;
3167    
3168      default:      default:
3169      return FALSE;      return consumed;
3170        }
3171    
3172      if (any)
3173        {
3174    #ifdef SUPPORT_UTF
3175        if (common->utf) return consumed;
3176    #endif
3177    #if defined COMPILE_PCRE8
3178        mask = 0xff;
3179    #elif defined COMPILE_PCRE16
3180        mask = 0xffff;
3181    #elif defined COMPILE_PCRE32
3182        mask = 0xffffffff;
3183    #else
3184        SLJIT_ASSERT_STOP();
3185    #endif
3186    
3187        do
3188          {
3189          chars[0] = mask;
3190          chars[1] = mask;
3191    
3192          if (--max_chars == 0)
3193            return consumed;
3194          consumed++;
3195          chars += 2;
3196          }
3197        while (--repeat > 0);
3198    
3199        repeat = 1;
3200        continue;
3201      }      }
3202    
3203    len = 1;    len = 1;
3204  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
3205    if (common->utf && HAS_EXTRALEN(cc[0])) len += GET_EXTRALEN(cc[0]);    if (common->utf && HAS_EXTRALEN(*cc)) len += GET_EXTRALEN(*cc);
3206  #endif  #endif
3207    
3208    if (caseless && char_has_othercase(common, cc))    if (caseless != 0 && char_has_othercase(common, cc))
3209      {      {
3210      caseless = char_get_othercase_bit(common, cc);      caseless = char_get_othercase_bit(common, cc);
3211      if (caseless == 0)      if (caseless == 0)
3212        return FALSE;        return consumed;
3213  #ifdef COMPILE_PCRE8  #ifdef COMPILE_PCRE8
3214      caseless = ((caseless & 0xff) << 8) | (len - (caseless >> 8));      caseless = ((caseless & 0xff) << 8) | (len - (caseless >> 8));
3215  #else  #else
# Line 2621  while (TRUE) Line 3222  while (TRUE)
3222    else    else
3223      caseless = 0;      caseless = 0;
3224    
3225    while (len > 0 && location < 2 * 2)    len_save = len;
3226      {    cc_save = cc;
3227      c = *cc;    while (TRUE)
3228      bit = 0;      {
3229      if (len == (caseless & 0xff))      do
3230        {        {
3231        bit = caseless >> 8;        chr = *cc;
3232        c |= bit;  #ifdef COMPILE_PCRE32
3233          if (SLJIT_UNLIKELY(chr == NOTACHAR))
3234            return consumed;
3235    #endif
3236          mask = 0;
3237          if (len == (caseless & 0xff))
3238            {
3239            mask = caseless >> 8;
3240            chr |= mask;
3241            }
3242    
3243          if (chars[0] == NOTACHAR)
3244            {
3245            chars[0] = chr;
3246            chars[1] = mask;
3247            }
3248          else
3249            {
3250            mask |= chars[0] ^ chr;
3251            chr |= mask;
3252            chars[0] = chr;
3253            chars[1] |= mask;
3254            }
3255    
3256          len--;
3257          if (--max_chars == 0)
3258            return consumed;
3259          consumed++;
3260          chars += 2;
3261          cc++;
3262        }        }
3263        while (len > 0);
3264    
3265        if (--repeat == 0)
3266          break;
3267    
3268      chars[location] = c;      len = len_save;
3269      chars[location + 1] = bit;      cc = cc_save;
3270        }
3271    
3272      len--;    repeat = 1;
3273      location += 2;    if (last)
3274      cc++;      return consumed;
3275      }
3276    }
3277    
3278    #define MAX_N_CHARS 16
3279    
3280    static SLJIT_INLINE BOOL fast_forward_first_n_chars(compiler_common *common, BOOL firstline)
3281    {
3282    DEFINE_COMPILER;
3283    struct sljit_label *start;
3284    struct sljit_jump *quit;
3285    pcre_uint32 chars[MAX_N_CHARS * 2];
3286    pcre_uint8 ones[MAX_N_CHARS];
3287    pcre_uint32 mask;
3288    int i, max;
3289    int offsets[3];
3290    
3291    for (i = 0; i < MAX_N_CHARS; i++)
3292      {
3293      chars[i << 1] = NOTACHAR;
3294      chars[(i << 1) + 1] = 0;
3295      }
3296    
3297    max = scan_prefix(common, common->start, chars, MAX_N_CHARS);
3298    
3299    if (max <= 1)
3300      return FALSE;
3301    
3302    for (i = 0; i < max; i++)
3303      {
3304      mask = chars[(i << 1) + 1];
3305      ones[i] = ones_in_half_byte[mask & 0xf];
3306      mask >>= 4;
3307      while (mask != 0)
3308        {
3309        ones[i] += ones_in_half_byte[mask & 0xf];
3310        mask >>= 4;
3311      }      }
3312      }
3313    
3314    if (location == 2 * 2)  offsets[0] = -1;
3315    /* Scan forward. */
3316    for (i = 0; i < max; i++)
3317      if (ones[i] <= 2) {
3318        offsets[0] = i;
3319      break;      break;
   else if (must_end)  
     return FALSE;  
3320    }    }
3321    
3322    if (offsets[0] == -1)
3323      return FALSE;
3324    
3325    /* Scan backward. */
3326    offsets[1] = -1;
3327    for (i = max - 1; i > offsets[0]; i--)
3328      if (ones[i] <= 2) {
3329        offsets[1] = i;
3330        break;
3331      }
3332    
3333    offsets[2] = -1;
3334    if (offsets[1] >= 0)
3335      {
3336      /* Scan from middle. */
3337      for (i = (offsets[0] + offsets[1]) / 2 + 1; i < offsets[1]; i++)
3338        if (ones[i] <= 2)
3339          {
3340          offsets[2] = i;
3341          break;
3342          }
3343    
3344      if (offsets[2] == -1)
3345        {
3346        for (i = (offsets[0] + offsets[1]) / 2; i > offsets[0]; i--)
3347          if (ones[i] <= 2)
3348            {
3349            offsets[2] = i;
3350            break;
3351            }
3352        }
3353      }
3354    
3355    SLJIT_ASSERT(offsets[1] == -1 || (offsets[0] < offsets[1]));
3356    SLJIT_ASSERT(offsets[2] == -1 || (offsets[0] < offsets[2] && offsets[1] > offsets[2]));
3357    
3358    chars[0] = chars[offsets[0] << 1];
3359    chars[1] = chars[(offsets[0] << 1) + 1];
3360    if (offsets[2] >= 0)
3361      {
3362      chars[2] = chars[offsets[2] << 1];
3363      chars[3] = chars[(offsets[2] << 1) + 1];
3364      }
3365    if (offsets[1] >= 0)
3366      {
3367      chars[4] = chars[offsets[1] << 1];
3368      chars[5] = chars[(offsets[1] << 1) + 1];
3369      }
3370    
3371    max -= 1;
3372  if (firstline)  if (firstline)
3373    {    {
3374    SLJIT_ASSERT(common->first_line_end != 0);    SLJIT_ASSERT(common->first_line_end != 0);
3375    OP1(SLJIT_MOV, TMP3, 0, STR_END, 0);    OP1(SLJIT_MOV, TMP3, 0, STR_END, 0);
3376    OP2(SLJIT_SUB, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, SLJIT_IMM, 1);    OP2(SLJIT_SUB, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, SLJIT_IMM, IN_UCHARS(max));
3377    }    }
3378  else  else
3379    OP2(SLJIT_SUB, STR_END, 0, STR_END, 0, SLJIT_IMM, 1);    OP2(SLJIT_SUB, STR_END, 0, STR_END, 0, SLJIT_IMM, IN_UCHARS(max));
3380    
3381  start = LABEL();  start = LABEL();
3382  quit = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);  quit = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
 #if defined SLJIT_UNALIGNED && SLJIT_UNALIGNED  
 #ifdef COMPILE_PCRE8  
 OP1(SLJIT_MOV_UH, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  
 #else /* COMPILE_PCRE8 */  
 OP1(SLJIT_MOV_UI, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  
 #endif  
   
 #else /* SLJIT_UNALIGNED */  
3383    
3384  #if defined SLJIT_BIG_ENDIAN && SLJIT_BIG_ENDIAN  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(offsets[0]));
3385  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);  if (offsets[1] >= 0)
3386  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(offsets[1]));
3387  #else /* SLJIT_BIG_ENDIAN */  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
 OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  
 OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  
 #endif /* SLJIT_BIG_ENDIAN */  
   
 #ifdef COMPILE_PCRE8  
 OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 8);  
 #else /* COMPILE_PCRE8 */  
 OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 16);  
 #endif  
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
3388    
3389  #endif  if (chars[1] != 0)
3390      OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, chars[1]);
3391    CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, chars[0], start);
3392    if (offsets[2] >= 0)
3393      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(offsets[2] - 1));
3394    
3395  if (chars[1] != 0 || chars[3] != 0)  if (offsets[1] >= 0)
3396    {    {
3397    pair.asuchars[0] = chars[1];    if (chars[5] != 0)
3398    pair.asuchars[1] = chars[3];      OP2(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_IMM, chars[5]);
3399    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, pair.ascombined);    CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, chars[4], start);
3400    }    }
3401    
3402  pair.asuchars[0] = chars[0];  if (offsets[2] >= 0)
3403  pair.asuchars[1] = chars[2];    {
3404  found = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, pair.ascombined);    if (chars[3] != 0)
3405        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, chars[3]);
3406      CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, chars[2], start);
3407      }
3408    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3409    
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  
 JUMPTO(SLJIT_JUMP, start);  
 JUMPHERE(found);  
3410  JUMPHERE(quit);  JUMPHERE(quit);
3411    
3412  if (firstline)  if (firstline)
3413    OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);    OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);
3414  else  else
3415    OP2(SLJIT_ADD, STR_END, 0, STR_END, 0, SLJIT_IMM, 1);    OP2(SLJIT_ADD, STR_END, 0, STR_END, 0, SLJIT_IMM, IN_UCHARS(max));
3416  return TRUE;  return TRUE;
3417  }  }
3418    
3419    #undef MAX_N_CHARS
3420    
3421  static SLJIT_INLINE void fast_forward_first_char(compiler_common *common, pcre_uchar first_char, BOOL caseless, BOOL firstline)  static SLJIT_INLINE void fast_forward_first_char(compiler_common *common, pcre_uchar first_char, BOOL caseless, BOOL firstline)
3422  {  {
3423  DEFINE_COMPILER;  DEFINE_COMPILER;
# Line 2738  if (first_char == oc) Line 3451  if (first_char == oc)
3451  else  else
3452    {    {
3453    bit = first_char ^ oc;    bit = first_char ^ oc;
3454    if (ispowerof2(bit))    if (is_powerof2(bit))
3455      {      {
3456      OP2(SLJIT_OR, TMP2, 0, TMP1, 0, SLJIT_IMM, bit);      OP2(SLJIT_OR, TMP2, 0, TMP1, 0, SLJIT_IMM, bit);
3457      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, first_char | bit);      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, first_char | bit);
# Line 2746  else Line 3459  else
3459    else    else
3460      {      {
3461      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, first_char);      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, first_char);
3462      COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_EQUAL);      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
3463      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, oc);      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, oc);
3464      COND_VALUE(SLJIT_OR | SLJIT_SET_E, TMP2, 0, SLJIT_C_EQUAL);      OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);
3465      found = JUMP(SLJIT_C_NOT_ZERO);      found = JUMP(SLJIT_C_NOT_ZERO);
3466      }      }
3467    }    }
# Line 2790  if (common->nltype == NLTYPE_FIXED && co Line 3503  if (common->nltype == NLTYPE_FIXED && co
3503    
3504    OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, IN_UCHARS(2));    OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, IN_UCHARS(2));
3505    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, STR_PTR, 0, TMP1, 0);
3506    COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_GREATER_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_GREATER_EQUAL);
3507  #ifdef COMPILE_PCRE16  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
3508    OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 1);    OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCHAR_SHIFT);
3509  #endif  #endif
3510    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
3511    
# Line 2819  firstchar = CMP(SLJIT_C_LESS_EQUAL, STR_ Line 3532  firstchar = CMP(SLJIT_C_LESS_EQUAL, STR_
3532  skip_char_back(common);  skip_char_back(common);
3533    
3534  loop = LABEL();  loop = LABEL();
3535  read_char(common);  read_char_max(common, common->nlmax, TRUE);
3536  lastchar = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);  lastchar = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3537  if (common->nltype == NLTYPE_ANY || common->nltype == NLTYPE_ANYCRLF)  if (common->nltype == NLTYPE_ANY || common->nltype == NLTYPE_ANYCRLF)
3538    foundcr = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR);    foundcr = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR);
# Line 2833  if (common->nltype == NLTYPE_ANY || comm Line 3546  if (common->nltype == NLTYPE_ANY || comm
3546    notfoundnl = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);    notfoundnl = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3547    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3548    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_NL);    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_NL);
3549    COND_VALUE(SLJIT_MOV, TMP1, 0, SLJIT_C_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
3550  #ifdef COMPILE_PCRE16  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
3551    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, UCHAR_SHIFT);
3552  #endif  #endif
3553    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3554    JUMPHERE(notfoundnl);    JUMPHERE(notfoundnl);
# Line 2848  if (firstline) Line 3561  if (firstline)
3561    OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);    OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);
3562  }  }
3563    
3564  static SLJIT_INLINE void fast_forward_start_bits(compiler_common *common, sljit_uw start_bits, BOOL firstline)  static BOOL check_class_ranges(compiler_common *common, const pcre_uint8 *bits, BOOL nclass, BOOL invert, jump_list **backtracks);
3565    
3566    static SLJIT_INLINE void fast_forward_start_bits(compiler_common *common, pcre_uint8 *start_bits, BOOL firstline)
3567  {  {
3568  DEFINE_COMPILER;  DEFINE_COMPILER;
3569  struct sljit_label *start;  struct sljit_label *start;
3570  struct sljit_jump *quit;  struct sljit_jump *quit;
3571  struct sljit_jump *found;  struct sljit_jump *found = NULL;
3572    jump_list *matches = NULL;
3573  #ifndef COMPILE_PCRE8  #ifndef COMPILE_PCRE8
3574  struct sljit_jump *jump;  struct sljit_jump *jump;
3575  #endif  #endif
# Line 2861  struct sljit_jump *jump; Line 3577  struct sljit_jump *jump;
3577  if (firstline)  if (firstline)
3578    {    {
3579    SLJIT_ASSERT(common->first_line_end != 0);    SLJIT_ASSERT(common->first_line_end != 0);
3580    OP1(SLJIT_MOV, TMP3, 0, STR_END, 0);    OP1(SLJIT_MOV, RETURN_ADDR, 0, STR_END, 0);
3581    OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end);    OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end);
3582    }    }
3583    
# Line 2872  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_P Line 3588  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_P
3588  if (common->utf)  if (common->utf)
3589    OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);    OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);
3590  #endif  #endif
3591    
3592    if (!check_class_ranges(common, start_bits, (start_bits[31] & 0x80) != 0, TRUE, &matches))
3593      {
3594  #ifndef COMPILE_PCRE8  #ifndef COMPILE_PCRE8
3595  jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 255);    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 255);
3596  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 255);    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 255);
3597  JUMPHERE(jump);    JUMPHERE(jump);
3598  #endif  #endif
3599  OP2(SLJIT_AND, TMP2, 0, TMP1, 0, SLJIT_IMM, 0x7);    OP2(SLJIT_AND, TMP2, 0, TMP1, 0, SLJIT_IMM, 0x7);
3600  OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 3);    OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 3);
3601  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), start_bits);    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)start_bits);
3602  OP2(SLJIT_SHL, TMP2, 0, SLJIT_IMM, 1, TMP2, 0);    OP2(SLJIT_SHL, TMP2, 0, SLJIT_IMM, 1, TMP2, 0);
3603  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);
3604  found = JUMP(SLJIT_C_NOT_ZERO);    found = JUMP(SLJIT_C_NOT_ZERO);
3605      }
3606    
3607  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
3608  if (common->utf)  if (common->utf)
3609    OP1(SLJIT_MOV, TMP1, 0, TMP3, 0);    OP1(SLJIT_MOV, TMP1, 0, TMP3, 0);
3610  #endif  #endif
3611  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3612  #if defined SUPPORT_UTF && defined COMPILE_PCRE8  #ifdef SUPPORT_UTF
3613    #if defined COMPILE_PCRE8
3614  if (common->utf)  if (common->utf)
3615    {    {
3616    CMPTO(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0, start);    CMPTO(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0, start);
3617    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_w)PRIV(utf8_table4) - 0xc0);    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
3618    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3619    }    }
3620  #endif  #elif defined COMPILE_PCRE16
 #if defined SUPPORT_UTF && defined COMPILE_PCRE16  
3621  if (common->utf)  if (common->utf)
3622    {    {
3623    CMPTO(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800, start);    CMPTO(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800, start);
3624    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
3625    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);
3626    COND_VALUE(SLJIT_MOV, TMP1, 0, SLJIT_C_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
3627    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
3628    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3629    }    }
3630  #endif  #endif /* COMPILE_PCRE[8|16] */
3631    #endif /* SUPPORT_UTF */
3632  JUMPTO(SLJIT_JUMP, start);  JUMPTO(SLJIT_JUMP, start);
3633  JUMPHERE(found);  if (found != NULL)
3634      JUMPHERE(found);
3635    if (matches != NULL)
3636      set_jumps(matches, LABEL());
3637  JUMPHERE(quit);  JUMPHERE(quit);
3638    
3639  if (firstline)  if (firstline)
3640    OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);    OP1(SLJIT_MOV, STR_END, 0, RETURN_ADDR, 0);
3641  }  }
3642    
3643  static SLJIT_INLINE struct sljit_jump *search_requested_char(compiler_common *common, pcre_uchar req_char, BOOL caseless, BOOL has_firstchar)  static SLJIT_INLINE struct sljit_jump *search_requested_char(compiler_common *common, pcre_uchar req_char, BOOL caseless, BOOL has_firstchar)
# Line 2925  struct sljit_jump *alreadyfound; Line 3649  struct sljit_jump *alreadyfound;
3649  struct sljit_jump *found;  struct sljit_jump *found;
3650  struct sljit_jump *foundoc = NULL;  struct sljit_jump *foundoc = NULL;
3651  struct sljit_jump *notfound;  struct sljit_jump *notfound;
3652  pcre_uchar oc, bit;  pcre_uint32 oc, bit;
3653    
3654  SLJIT_ASSERT(common->req_char_ptr != 0);  SLJIT_ASSERT(common->req_char_ptr != 0);
3655  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->req_char_ptr);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->req_char_ptr);
# Line 2956  if (req_char == oc) Line 3680  if (req_char == oc)
3680  else  else
3681    {    {
3682    bit = req_char ^ oc;    bit = req_char ^ oc;
3683    if (ispowerof2(bit))    if (is_powerof2(bit))
3684      {      {
3685      OP2(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_IMM, bit);      OP2(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_IMM, bit);
3686      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, req_char | bit);      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, req_char | bit);
# Line 2992  GET_LOCAL_BASE(TMP3, 0, 0); Line 3716  GET_LOCAL_BASE(TMP3, 0, 0);
3716  /* Drop frames until we reach STACK_TOP. */  /* Drop frames until we reach STACK_TOP. */
3717  mainloop = LABEL();  mainloop = LABEL();
3718  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(TMP1), 0);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(TMP1), 0);
3719  jump = CMP(SLJIT_C_SIG_LESS_EQUAL, TMP2, 0, SLJIT_IMM, frame_end);  OP2(SLJIT_SUB | SLJIT_SET_S, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0);
3720    jump = JUMP(SLJIT_C_SIG_LESS_EQUAL);
3721    
3722  OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, TMP3, 0);  OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, TMP3, 0);
3723  OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), 0, SLJIT_MEM1(TMP1), sizeof(sljit_w));  OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), 0, SLJIT_MEM1(TMP1), sizeof(sljit_sw));
3724  OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), sizeof(sljit_w), SLJIT_MEM1(TMP1), 2 * sizeof(sljit_w));  OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), sizeof(sljit_sw), SLJIT_MEM1(TMP1), 2 * sizeof(sljit_sw));
3725  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 3 * sizeof(sljit_w));  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 3 * sizeof(sljit_sw));
3726  JUMPTO(SLJIT_JUMP, mainloop);  JUMPTO(SLJIT_JUMP, mainloop);
3727    
3728  JUMPHERE(jump);  JUMPHERE(jump);
3729  jump = CMP(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, frame_end);  jump = JUMP(SLJIT_C_SIG_LESS);
3730  /* End of dropping frames. */  /* End of dropping frames. */
3731  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3732    
3733  JUMPHERE(jump);  JUMPHERE(jump);
3734  jump = CMP(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, frame_setstrbegin);  OP1(SLJIT_NEG, TMP2, 0, TMP2, 0);
3735  /* Set string begin. */  OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, TMP3, 0);
3736  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(TMP1), sizeof(sljit_w));  OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), 0, SLJIT_MEM1(TMP1), sizeof(sljit_sw));
3737  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 2 * sizeof(sljit_w));  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 2 * sizeof(sljit_sw));
 OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0), TMP2, 0);  
 JUMPTO(SLJIT_JUMP, mainloop);  
   
 JUMPHERE(jump);  
 if (common->mark_ptr != 0)  
   {  
   jump = CMP(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, frame_setmark);  
   OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(TMP1), sizeof(sljit_w));  
   OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 2 * sizeof(sljit_w));  
   OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr, TMP2, 0);  
   JUMPTO(SLJIT_JUMP, mainloop);  
   
   JUMPHERE(jump);  
   }  
   
 /* Unknown command. */  
 OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 2 * sizeof(sljit_w));  
3738  JUMPTO(SLJIT_JUMP, mainloop);  JUMPTO(SLJIT_JUMP, mainloop);
3739  }  }
3740    
# Line 3033  static void check_wordboundary(compiler_ Line 3742  static void check_wordboundary(compiler_
3742  {  {
3743  DEFINE_COMPILER;  DEFINE_COMPILER;
3744  struct sljit_jump *skipread;  struct sljit_jump *skipread;
3745    jump_list *skipread_list = NULL;
3746  #if !(defined COMPILE_PCRE8) || defined SUPPORT_UTF  #if !(defined COMPILE_PCRE8) || defined SUPPORT_UTF
3747  struct sljit_jump *jump;  struct sljit_jump *jump;
3748  #endif  #endif
# Line 3058  if (common->use_ucp) Line 3768  if (common->use_ucp)
3768    add_jump(compiler, &common->getucd, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->getucd, JUMP(SLJIT_FAST_CALL));
3769    OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Ll);    OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Ll);
3770    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_Lu - ucp_Ll);    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_Lu - ucp_Ll);
3771    COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_LESS_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_LESS_EQUAL);
3772    OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Nd - ucp_Ll);    OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Nd - ucp_Ll);
3773    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_No - ucp_Nd);    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_No - ucp_Nd);
3774    COND_VALUE(SLJIT_OR, TMP2, 0, SLJIT_C_LESS_EQUAL);    OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_C_LESS_EQUAL);
3775    JUMPHERE(jump);    JUMPHERE(jump);
3776    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP2, 0);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP2, 0);
3777    }    }
# Line 3090  else Line 3800  else
3800  JUMPHERE(skipread);  JUMPHERE(skipread);
3801    
3802  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 0);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 0);
3803  skipread = check_str_end(common);  check_str_end(common, &skipread_list);
3804  peek_char(common);  peek_char(common);
3805    
3806  /* Testing char type. This is a code duplication. */  /* Testing char type. This is a code duplication. */