/[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 929 by zherczeg, Fri Feb 24 11:07:47 2012 UTC revision 1422 by zherczeg, Mon Dec 30 19:05:36 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 82  The code generator follows the recursive Line 94  The code generator follows the recursive
94  expressions. The basic blocks of regular expressions are condition checkers  expressions. The basic blocks of regular expressions are condition checkers
95  whose execute different commands depending on the result of the condition check.  whose execute different commands depending on the result of the condition check.
96  The relationship between the operators can be horizontal (concatenation) and  The relationship between the operators can be horizontal (concatenation) and
97  vertical (sub-expression) (See struct fallback_common for more details).  vertical (sub-expression) (See struct backtrack_common for more details).
98    
99    'ab' - 'a' and 'b' regexps are concatenated    'ab' - 'a' and 'b' regexps are concatenated
100    'a+' - 'a' is the sub-expression of the '+' operator    'a+' - 'a' is the sub-expression of the '+' operator
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 hot 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 'fallback' 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 hot path.  branches on the matching path.
107    
108   Greedy star operator (*) :   Greedy star operator (*) :
109     Hot path: match happens.     Matching path: match happens.
110     Fallback path: match failed.     Backtrack path: match failed.
111   Non-greedy star operator (*?) :   Non-greedy star operator (*?) :
112     Hot path: no need to perform a match.     Matching path: no need to perform a match.
113     Fallback 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
116  with two alternatives. Let A, B, C, D are arbirary regular expressions, and  with two alternatives. Let A, B, C, D are arbirary regular expressions, and
# 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 hot path   A matching path
124   '(' hot path (pushing arguments to the stack)   '(' matching path (pushing arguments to the stack)
125   B hot path   B matching path
126   ')' hot path (pushing arguments to the stack)   ')' matching path (pushing arguments to the stack)
127   D hot path   D matching path
128   return with successful match   return with successful match
129    
130   D fallback path   D backtrack path
131   ')' fallback path (If we arrived from "C" jump to the fallback of "C")   ')' backtrack path (If we arrived from "C" jump to the backtrack of "C")
132   B fallback path   B backtrack path
133   C expected path   C expected path
134   jump to D hot path   jump to D matching path
135   C fallback path   C backtrack path
136   A fallback path   A backtrack path
137    
138   Notice, that the order of fallback code paths are the opposite of the fast   Notice, that the order of backtrack code paths are the opposite of the fast
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 fallback code path. The fallback code 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 hot 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 fallback 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 fallback 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 *leave;    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_hotpath, and contains  code generator. It is allocated by compile_matchingpath, and contains
214  the aguments for compile_fallbackpath. 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 fallback_common {  typedef struct backtrack_common {
217    /* Concatenation stack. */    /* Concatenation stack. */
218    struct fallback_common *prev;    struct backtrack_common *prev;
219    jump_list *nextfallbacks;    jump_list *nextbacktracks;
220    /* Internal stack (for component operators). */    /* Internal stack (for component operators). */
221    struct fallback_common *top;    struct backtrack_common *top;
222    jump_list *topfallbacks;    jump_list *topbacktracks;
223    /* Opcode pointer. */    /* Opcode pointer. */
224    pcre_uchar *cc;    pcre_uchar *cc;
225  } fallback_common;  } backtrack_common;
226    
227  typedef struct assert_fallback {  typedef struct assert_backtrack {
228    fallback_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 *hotpath;    struct sljit_label *matchingpath;
236  } assert_fallback;  } assert_backtrack;
237    
238  typedef struct bracket_fallback {  typedef struct bracket_backtrack {
239    fallback_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 *althotpath;    struct sljit_label *alternative_matchingpath;
242    /* For rmin and rmax iterators. */    /* For rmin and rmax iterators. */
243    struct sljit_label *recursivehotpath;    struct sljit_label *recursive_matchingpath;
244    /* For greedy ? operator. */    /* For greedy ? operator. */
245    struct sljit_label *zerohotpath;    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_fallback *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_fallback;  } bracket_backtrack;
257    
258  typedef struct bracketpos_fallback {  typedef struct bracketpos_backtrack {
259    fallback_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. */
265    int stacksize;    int stacksize;
266  } bracketpos_fallback;  } bracketpos_backtrack;
267    
268  typedef struct braminzero_fallback {  typedef struct braminzero_backtrack {
269    fallback_common common;    backtrack_common common;
270    struct sljit_label *hotpath;    struct sljit_label *matchingpath;
271  } braminzero_fallback;  } braminzero_backtrack;
272    
273  typedef struct iterator_fallback {  typedef struct iterator_backtrack {
274    fallback_common common;    backtrack_common common;
275    /* Next iteration. */    /* Next iteration. */
276    struct sljit_label *hotpath;    struct sljit_label *matchingpath;
277  } iterator_fallback;  } iterator_backtrack;
278    
279  typedef struct recurse_entry {  typedef struct recurse_entry {
280    struct recurse_entry *next;    struct recurse_entry *next;
# 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_fallback {  typedef struct recurse_backtrack {
290    fallback_common common;    backtrack_common common;
291  } recurse_fallback;    BOOL inlined_pattern;
292    } recurse_backtrack;
293    
294    #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    /* Local stack area size and variable pointers. */    sljit_si *private_data_ptrs;
318    int localsize;    /* Tells whether the capturing bracket is optimized. */
319    int *localptrs;    pcre_uint8 *optimized_cbracket;
320    int cbraptr;    /* Tells whether the starting offset is a target of then. */
321    /* OVector starting point. Must be divisible by 2. */    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 290  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    /* Other  */    /* 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. */
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. */
365    int nltype;    int nltype;
366    int newline;    int newline;
367    int bsr_nltype;    int bsr_nltype;
368      /* Dollar endonly. */
369    int endonly;    int endonly;
370    BOOL has_set_som;    /* Tables. */
371    sljit_w ctypes;    sljit_sw ctypes;
372    sljit_uw name_table;    /* Named capturing brackets. */
373    sljit_w name_count;    pcre_uchar *name_table;
374    sljit_w name_entry_size;    sljit_sw name_count;
375      sljit_sw name_entry_size;
376    
377    /* Labels and jump lists. */    /* Labels and jump lists. */
378    struct sljit_label *partialmatchlabel;    struct sljit_label *partialmatchlabel;
379    struct sljit_label *acceptlabel;    struct sljit_label *quit_label;
380      struct sljit_label *forced_quit_label;
381      struct sljit_label *accept_label;
382    stub_list *stubs;    stub_list *stubs;
383    recurse_entry *entries;    recurse_entry *entries;
384    recurse_entry *currententry;    recurse_entry *currententry;
385    jump_list *partialmatch;    jump_list *partialmatch;
386      jump_list *quit;
387      jump_list *positive_assert_quit;
388      jump_list *forced_quit;
389    jump_list *accept;    jump_list *accept;
390    jump_list *calllimit;    jump_list *calllimit;
391    jump_list *stackalloc;    jump_list *stackalloc;
# Line 322  typedef struct compiler_common { Line 396  typedef struct compiler_common {
396    jump_list *vspace;    jump_list *vspace;
397    jump_list *casefulcmp;    jump_list *casefulcmp;
398    jump_list *caselesscmp;    jump_list *caselesscmp;
399      jump_list *reset_match;
400    BOOL jscript_compat;    BOOL jscript_compat;
401  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
402    BOOL utf;    BOOL utf;
403  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
404    BOOL use_ucp;    BOOL use_ucp;
405  #endif  #endif
   jump_list *utfreadchar;  
406  #ifdef COMPILE_PCRE8  #ifdef COMPILE_PCRE8
407      jump_list *utfreadchar;
408      jump_list *utfreadchar16;
409    jump_list *utfreadtype8;    jump_list *utfreadtype8;
410  #endif  #endif
411  #endif /* SUPPORT_UTF */  #endif /* SUPPORT_UTF */
# Line 346  typedef struct compare_context { Line 422  typedef struct compare_context {
422  #if defined SLJIT_UNALIGNED && SLJIT_UNALIGNED  #if defined SLJIT_UNALIGNED && SLJIT_UNALIGNED
423    int ucharptr;    int ucharptr;
424    union {    union {
425      sljit_i asint;      sljit_si asint;
426      sljit_uh asushort;      sljit_uh asushort;
427  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
428      sljit_ub asbyte;      sljit_ub asbyte;
429      sljit_ub asuchars[4];      sljit_ub asuchars[4];
430  #else  #elif defined COMPILE_PCRE16
 #ifdef COMPILE_PCRE16  
431      sljit_uh asuchars[2];      sljit_uh asuchars[2];
432  #endif  #elif defined COMPILE_PCRE32
433        sljit_ui asuchars[1];
434  #endif  #endif
435    } c;    } c;
436    union {    union {
437      sljit_i asint;      sljit_si asint;
438      sljit_uh asushort;      sljit_uh asushort;
439  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
440      sljit_ub asbyte;      sljit_ub asbyte;
441      sljit_ub asuchars[4];      sljit_ub asuchars[4];
442  #else  #elif defined COMPILE_PCRE16
 #ifdef COMPILE_PCRE16  
443      sljit_uh asuchars[2];      sljit_uh asuchars[2];
444  #endif  #elif defined COMPILE_PCRE32
445        sljit_ui asuchars[1];
446  #endif  #endif
447    } oc;    } oc;
448  #endif  #endif
449  } compare_context;  } compare_context;
450    
 enum {  
   frame_end = 0,  
   frame_setstrbegin = -1,  
   frame_setmark = -2  
 };  
   
451  /* Undefine sljit macros. */  /* Undefine sljit macros. */
452  #undef CMP  #undef CMP
453    
454  /* Used for accessing the elements of the stack. */  /* Used for accessing the elements of the stack. */
455  #define STACK(i)      ((-(i) - 1) * (int)sizeof(sljit_w))  #define STACK(i)      ((-(i) - 1) * (int)sizeof(sljit_sw))
456    
457  #define TMP1          SLJIT_TEMPORARY_REG1  #define TMP1          SLJIT_SCRATCH_REG1
458  #define TMP2          SLJIT_TEMPORARY_REG3  #define TMP2          SLJIT_SCRATCH_REG3
459  #define TMP3          SLJIT_TEMPORARY_EREG2  #define TMP3          SLJIT_TEMPORARY_EREG2
460  #define STR_PTR       SLJIT_SAVED_REG1  #define STR_PTR       SLJIT_SAVED_REG1
461  #define STR_END       SLJIT_SAVED_REG2  #define STR_END       SLJIT_SAVED_REG2
462  #define STACK_TOP     SLJIT_TEMPORARY_REG2  #define STACK_TOP     SLJIT_SCRATCH_REG2
463  #define STACK_LIMIT   SLJIT_SAVED_REG3  #define STACK_LIMIT   SLJIT_SAVED_REG3
464  #define ARGUMENTS     SLJIT_SAVED_EREG1  #define ARGUMENTS     SLJIT_SAVED_EREG1
465  #define CALL_COUNT    SLJIT_SAVED_EREG2  #define COUNT_MATCH   SLJIT_SAVED_EREG2
466  #define RETURN_ADDR   SLJIT_TEMPORARY_EREG1  #define RETURN_ADDR   SLJIT_TEMPORARY_EREG1
467    
468  /* Locals layout. */  /* Local space layout. */
469  /* These two locals can be used by the current opcode. */  /* These two locals can be used by the current opcode. */
470  #define LOCALS0          (0 * sizeof(sljit_w))  #define LOCALS0          (0 * sizeof(sljit_sw))
471  #define LOCALS1          (1 * sizeof(sljit_w))  #define LOCALS1          (1 * sizeof(sljit_sw))
472  /* Two local variables for possessive quantifiers (char1 cannot use them). */  /* Two local variables for possessive quantifiers (char1 cannot use them). */
473  #define POSSESSIVE0      (2 * sizeof(sljit_w))  #define POSSESSIVE0      (2 * sizeof(sljit_sw))
474  #define POSSESSIVE1      (3 * sizeof(sljit_w))  #define POSSESSIVE1      (3 * sizeof(sljit_sw))
475  /* Max limit of recursions. */  /* Max limit of recursions. */
476  #define CALL_LIMIT       (4 * sizeof(sljit_w))  #define LIMIT_MATCH      (4 * sizeof(sljit_sw))
477  /* The output vector is stored on the stack, and contains pointers  /* The output vector is stored on the stack, and contains pointers
478  to characters. The vector data is divided into two groups: the first  to characters. The vector data is divided into two groups: the first
479  group contains the start / end character pointers, and the second is  group contains the start / end character pointers, and the second is
480  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. */
481  #define OVECTOR_START    (common->ovector_start)  #define OVECTOR_START    (common->ovector_start)
482  #define OVECTOR(i)       (OVECTOR_START + (i) * sizeof(sljit_w))  #define OVECTOR(i)       (OVECTOR_START + (i) * (sljit_sw)sizeof(sljit_sw))
483  #define OVECTOR_PRIV(i)  (common->cbraptr + (i) * sizeof(sljit_w))  #define OVECTOR_PRIV(i)  (common->cbra_ptr + (i) * (sljit_sw)sizeof(sljit_sw))
484  #define PRIV_DATA(cc)    (common->localptrs[(cc) - common->start])  #define PRIVATE_DATA(cc) (common->private_data_ptrs[(cc) - common->start])
485    
486  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
487  #define MOV_UCHAR  SLJIT_MOV_UB  #define MOV_UCHAR  SLJIT_MOV_UB
488  #define MOVU_UCHAR SLJIT_MOVU_UB  #define MOVU_UCHAR SLJIT_MOVU_UB
489  #else  #elif defined COMPILE_PCRE16
 #ifdef COMPILE_PCRE16  
490  #define MOV_UCHAR  SLJIT_MOV_UH  #define MOV_UCHAR  SLJIT_MOV_UH
491  #define MOVU_UCHAR SLJIT_MOVU_UH  #define MOVU_UCHAR SLJIT_MOVU_UH
492    #elif defined COMPILE_PCRE32
493    #define MOV_UCHAR  SLJIT_MOV_UI
494    #define MOVU_UCHAR SLJIT_MOVU_UI
495  #else  #else
496  #error Unsupported compiling mode  #error Unsupported compiling mode
497  #endif  #endif
 #endif  
498    
499  /* Shortcuts. */  /* Shortcuts. */
500  #define DEFINE_COMPILER \  #define DEFINE_COMPILER \
# Line 440  the start pointers when the end of the c Line 511  the start pointers when the end of the c
511    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))
512  #define JUMPHERE(jump) \  #define JUMPHERE(jump) \
513    sljit_set_label((jump), sljit_emit_label(compiler))    sljit_set_label((jump), sljit_emit_label(compiler))
514    #define SET_LABEL(jump, label) \
515      sljit_set_label((jump), (label))
516  #define CMP(type, src1, src1w, src2, src2w) \  #define CMP(type, src1, src1w, src2, src2w) \
517    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))
518  #define CMPTO(type, src1, src1w, src2, src2w, label) \  #define CMPTO(type, src1, src1w, src2, src2w, label) \
519    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))
520  #define COND_VALUE(op, dst, dstw, type) \  #define OP_FLAGS(op, dst, dstw, src, srcw, type) \
521    sljit_emit_cond_value(compiler, (op), (dst), (dstw), (type))    sljit_emit_op_flags(compiler, (op), (dst), (dstw), (src), (srcw), (type))
522    #define GET_LOCAL_BASE(dst, dstw, offset) \
523      sljit_get_local_base(compiler, (dst), (dstw), (offset))
524    
525  static pcre_uchar* bracketend(pcre_uchar* cc)  static pcre_uchar* bracketend(pcre_uchar* cc)
526  {  {
# Line 456  cc += 1 + LINK_SIZE; Line 531  cc += 1 + LINK_SIZE;
531  return cc;  return cc;
532  }  }
533    
534    static int ones_in_half_byte[16] = {
535      /* 0 */ 0, 1, 1, 2, /* 4 */ 1, 2, 2, 3,
536      /* 8 */ 1, 2, 2, 3, /* 12 */ 2, 3, 3, 4
537    };
538    
539  /* Functions whose might need modification for all new supported opcodes:  /* Functions whose might need modification for all new supported opcodes:
540   next_opcode   next_opcode
541   get_localspace   check_opcode_types
542   set_localptrs   set_private_data_ptrs
543   get_framesize   get_framesize
544   init_frame   init_frame
545   get_localsize   get_private_data_copy_length
546   copy_locals   copy_private_data
547   compile_hotpath   compile_matchingpath
548   compile_fallbackpath   compile_backtrackingpath
549  */  */
550    
551  static pcre_uchar *next_opcode(compiler_common *common, pcre_uchar *cc)  static pcre_uchar *next_opcode(compiler_common *common, pcre_uchar *cc)
# Line 486  switch(*cc) Line 566  switch(*cc)
566    case OP_WORDCHAR:    case OP_WORDCHAR:
567    case OP_ANY:    case OP_ANY:
568    case OP_ALLANY:    case OP_ALLANY:
569      case OP_NOTPROP:
570      case OP_PROP:
571    case OP_ANYNL:    case OP_ANYNL:
572    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
573    case OP_HSPACE:    case OP_HSPACE:
# Line 498  switch(*cc) Line 580  switch(*cc)
580    case OP_CIRCM:    case OP_CIRCM:
581    case OP_DOLL:    case OP_DOLL:
582    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:  
583    case OP_CRSTAR:    case OP_CRSTAR:
584    case OP_CRMINSTAR:    case OP_CRMINSTAR:
585    case OP_CRPLUS:    case OP_CRPLUS:
586    case OP_CRMINPLUS:    case OP_CRMINPLUS:
587    case OP_CRQUERY:    case OP_CRQUERY:
588    case OP_CRMINQUERY:    case OP_CRMINQUERY:
589      case OP_CRRANGE:
590      case OP_CRMINRANGE:
591      case OP_CRPOSSTAR:
592      case OP_CRPOSPLUS:
593      case OP_CRPOSQUERY:
594      case OP_CRPOSRANGE:
595      case OP_CLASS:
596      case OP_NCLASS:
597      case OP_REF:
598      case OP_REFI:
599      case OP_DNREF:
600      case OP_DNREFI:
601      case OP_RECURSE:
602      case OP_CALLOUT:
603      case OP_ALT:
604      case OP_KET:
605      case OP_KETRMAX:
606      case OP_KETRMIN:
607      case OP_KETRPOS:
608      case OP_REVERSE:
609      case OP_ASSERT:
610      case OP_ASSERT_NOT:
611      case OP_ASSERTBACK:
612      case OP_ASSERTBACK_NOT:
613      case OP_ONCE:
614      case OP_ONCE_NC:
615      case OP_BRA:
616      case OP_BRAPOS:
617      case OP_CBRA:
618      case OP_CBRAPOS:
619      case OP_COND:
620      case OP_SBRA:
621      case OP_SBRAPOS:
622      case OP_SCBRA:
623      case OP_SCBRAPOS:
624      case OP_SCOND:
625      case OP_CREF:
626      case OP_DNCREF:
627      case OP_RREF:
628      case OP_DNRREF:
629    case OP_DEF:    case OP_DEF:
630    case OP_BRAZERO:    case OP_BRAZERO:
631    case OP_BRAMINZERO:    case OP_BRAMINZERO:
632    case OP_BRAPOSZERO:    case OP_BRAPOSZERO:
633      case OP_PRUNE:
634      case OP_SKIP:
635      case OP_THEN:
636      case OP_COMMIT:
637    case OP_FAIL:    case OP_FAIL:
638    case OP_ACCEPT:    case OP_ACCEPT:
639    case OP_ASSERT_ACCEPT:    case OP_ASSERT_ACCEPT:
640      case OP_CLOSE:
641    case OP_SKIPZERO:    case OP_SKIPZERO:
642    return cc + 1;    return cc + PRIV(OP_lengths)[*cc];
   
   case OP_ANYBYTE:  
 #ifdef SUPPORT_UTF  
   if (common->utf) return NULL;  
 #endif  
   return cc + 1;  
643    
644    case OP_CHAR:    case OP_CHAR:
645    case OP_CHARI:    case OP_CHARI:
# Line 539  switch(*cc) Line 651  switch(*cc)
651    case OP_MINPLUS:    case OP_MINPLUS:
652    case OP_QUERY:    case OP_QUERY:
653    case OP_MINQUERY:    case OP_MINQUERY:
654      case OP_UPTO:
655      case OP_MINUPTO:
656      case OP_EXACT:
657    case OP_POSSTAR:    case OP_POSSTAR:
658    case OP_POSPLUS:    case OP_POSPLUS:
659    case OP_POSQUERY:    case OP_POSQUERY:
660      case OP_POSUPTO:
661    case OP_STARI:    case OP_STARI:
662    case OP_MINSTARI:    case OP_MINSTARI:
663    case OP_PLUSI:    case OP_PLUSI:
664    case OP_MINPLUSI:    case OP_MINPLUSI:
665    case OP_QUERYI:    case OP_QUERYI:
666    case OP_MINQUERYI:    case OP_MINQUERYI:
667      case OP_UPTOI:
668      case OP_MINUPTOI:
669      case OP_EXACTI:
670    case OP_POSSTARI:    case OP_POSSTARI:
671    case OP_POSPLUSI:    case OP_POSPLUSI:
672    case OP_POSQUERYI:    case OP_POSQUERYI:
673      case OP_POSUPTOI:
674    case OP_NOTSTAR:    case OP_NOTSTAR:
675    case OP_NOTMINSTAR:    case OP_NOTMINSTAR:
676    case OP_NOTPLUS:    case OP_NOTPLUS:
677    case OP_NOTMINPLUS:    case OP_NOTMINPLUS:
678    case OP_NOTQUERY:    case OP_NOTQUERY:
679    case OP_NOTMINQUERY:    case OP_NOTMINQUERY:
680      case OP_NOTUPTO:
681      case OP_NOTMINUPTO:
682      case OP_NOTEXACT:
683    case OP_NOTPOSSTAR:    case OP_NOTPOSSTAR:
684    case OP_NOTPOSPLUS:    case OP_NOTPOSPLUS:
685    case OP_NOTPOSQUERY:    case OP_NOTPOSQUERY:
686      case OP_NOTPOSUPTO:
687    case OP_NOTSTARI:    case OP_NOTSTARI:
688    case OP_NOTMINSTARI:    case OP_NOTMINSTARI:
689    case OP_NOTPLUSI:    case OP_NOTPLUSI:
690    case OP_NOTMINPLUSI:    case OP_NOTMINPLUSI:
691    case OP_NOTQUERYI:    case OP_NOTQUERYI:
692    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:  
693    case OP_NOTUPTOI:    case OP_NOTUPTOI:
694    case OP_NOTMINUPTOI:    case OP_NOTMINUPTOI:
695    case OP_NOTEXACTI:    case OP_NOTEXACTI:
696      case OP_NOTPOSSTARI:
697      case OP_NOTPOSPLUSI:
698      case OP_NOTPOSQUERYI:
699    case OP_NOTPOSUPTOI:    case OP_NOTPOSUPTOI:
700    cc += 2 + IMM2_SIZE;    cc += PRIV(OP_lengths)[*cc];
701  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
702    if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);    if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
703  #endif  #endif
704    return cc;    return cc;
705    
706    case OP_NOTPROP:    /* Special cases. */
707    case OP_PROP:    case OP_TYPESTAR:
708    return cc + 1 + 2;    case OP_TYPEMINSTAR:
709      case OP_TYPEPLUS:
710      case OP_TYPEMINPLUS:
711      case OP_TYPEQUERY:
712      case OP_TYPEMINQUERY:
713    case OP_TYPEUPTO:    case OP_TYPEUPTO:
714    case OP_TYPEMINUPTO:    case OP_TYPEMINUPTO:
715    case OP_TYPEEXACT:    case OP_TYPEEXACT:
716      case OP_TYPEPOSSTAR:
717      case OP_TYPEPOSPLUS:
718      case OP_TYPEPOSQUERY:
719    case OP_TYPEPOSUPTO:    case OP_TYPEPOSUPTO:
720    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;  
721    
722    case OP_CLASS:    case OP_ANYBYTE:
723    case OP_NCLASS:  #ifdef SUPPORT_UTF
724    return cc + 1 + 32 / sizeof(pcre_uchar);    if (common->utf) return NULL;
725    #endif
726      return cc + 1;
727    
728  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
729    case OP_XCLASS:    case OP_XCLASS:
730    return cc + GET(cc, 1);    return cc + GET(cc, 1);
731  #endif  #endif
732    
   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;  
   
733    case OP_MARK:    case OP_MARK:
734      case OP_PRUNE_ARG:
735      case OP_SKIP_ARG:
736      case OP_THEN_ARG:
737    return cc + 1 + 2 + cc[1];    return cc + 1 + 2 + cc[1];
738    
739    default:    default:
740      /* All opcodes are supported now! */
741      SLJIT_ASSERT_STOP();
742    return NULL;    return NULL;
743    }    }
744  }  }
745    
746  static int get_localspace(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend)  static BOOL check_opcode_types(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend)
747  {  {
748  int localspace = 0;  int count;
749  pcre_uchar *alternative;  pcre_uchar *slot;
750    
751  /* 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. */
752  while (cc < ccend)  while (cc < ccend)
753    {    {
# Line 677  while (cc < ccend) Line 758  while (cc < ccend)
758      cc += 1;      cc += 1;
759      break;      break;
760    
761      case OP_ASSERT:      case OP_REF:
762      case OP_ASSERT_NOT:      case OP_REFI:
763      case OP_ASSERTBACK:      common->optimized_cbracket[GET2(cc, 1)] = 0;
764      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);  
     cc += 1 + LINK_SIZE;  
765      break;      break;
766    
767      case OP_CBRAPOS:      case OP_CBRAPOS:
768      case OP_SCBRAPOS:      case OP_SCBRAPOS:
769      localspace += sizeof(sljit_w);      common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] = 0;
770      cc += 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
771      break;      break;
772    
773      case OP_COND:      case OP_COND:
774      /* Might be a hidden SCOND. */      case OP_SCOND:
775      alternative = cc + GET(cc, 1);      /* Only AUTO_CALLOUT can insert this opcode. We do
776      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)         not intend to support this case. */
777        localspace += sizeof(sljit_w);      if (cc[1 + LINK_SIZE] == OP_CALLOUT)
778          return FALSE;
779      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
780      break;      break;
781    
782        case OP_CREF:
783        common->optimized_cbracket[GET2(cc, 1)] = 0;
784        cc += 1 + IMM2_SIZE;
785        break;
786    
787        case OP_DNREF:
788        case OP_DNREFI:
789        case OP_DNCREF:
790        count = GET2(cc, 1 + IMM2_SIZE);
791        slot = common->name_table + GET2(cc, 1) * common->name_entry_size;
792        while (count-- > 0)
793          {
794          common->optimized_cbracket[GET2(slot, 0)] = 0;
795          slot += common->name_entry_size;
796          }
797        cc += 1 + 2 * IMM2_SIZE;
798        break;
799    
800      case OP_RECURSE:      case OP_RECURSE:
801      /* Set its value only once. */      /* Set its value only once. */
802      if (common->recursive_head == 0)      if (common->recursive_head_ptr == 0)
803        {        {
804        common->recursive_head = common->ovector_start;        common->recursive_head_ptr = common->ovector_start;
805        common->ovector_start += sizeof(sljit_w);        common->ovector_start += sizeof(sljit_sw);
806        }        }
807      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
808      break;      break;
809    
810        case OP_CALLOUT:
811        if (common->capture_last_ptr == 0)
812          {
813          common->capture_last_ptr = common->ovector_start;
814          common->ovector_start += sizeof(sljit_sw);
815          }
816        cc += 2 + 2 * LINK_SIZE;
817        break;
818    
819        case OP_THEN_ARG:
820        common->has_then = TRUE;
821        common->control_head_ptr = 1;
822        /* Fall through. */
823    
824        case OP_PRUNE_ARG:
825        common->needs_start_ptr = TRUE;
826        /* Fall through. */
827    
828      case OP_MARK:      case OP_MARK:
829      if (common->mark_ptr == 0)      if (common->mark_ptr == 0)
830        {        {
831        common->mark_ptr = common->ovector_start;        common->mark_ptr = common->ovector_start;
832        common->ovector_start += sizeof(sljit_w);        common->ovector_start += sizeof(sljit_sw);
833        }        }
834      cc += 1 + 2 + cc[1];      cc += 1 + 2 + cc[1];
835      break;      break;
836    
837        case OP_THEN:
838        common->has_then = TRUE;
839        common->control_head_ptr = 1;
840        /* Fall through. */
841    
842        case OP_PRUNE:
843        case OP_SKIP:
844        common->needs_start_ptr = TRUE;
845        cc += 1;
846        break;
847    
848        case OP_SKIP_ARG:
849        common->control_head_ptr = 1;
850        common->has_skip_arg = TRUE;
851        cc += 1 + 2 + cc[1];
852        break;
853    
854      default:      default:
855      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
856      if (cc == NULL)      if (cc == NULL)
857        return -1;        return FALSE;
858      break;      break;
859      }      }
860    }    }
861  return localspace;  return TRUE;
862  }  }
863    
864  static void set_localptrs(compiler_common *common, int localptr, pcre_uchar *ccend)  static int get_class_iterator_size(pcre_uchar *cc)
865  {  {
866  pcre_uchar *cc = common->start;  switch(*cc)
 pcre_uchar *alternative;  
 while (cc < ccend)  
867    {    {
868    switch(*cc)    case OP_CRSTAR:
869      {    case OP_CRPLUS:
870      case OP_ASSERT:    return 2;
     case OP_ASSERT_NOT:  
     case OP_ASSERTBACK:  
     case OP_ASSERTBACK_NOT:  
     case OP_ONCE:  
     case OP_ONCE_NC:  
     case OP_BRAPOS:  
     case OP_SBRA:  
     case OP_SBRAPOS:  
     case OP_SCOND:  
     common->localptrs[cc - common->start] = localptr;  
     localptr += sizeof(sljit_w);  
     cc += 1 + LINK_SIZE;  
     break;  
871    
872      case OP_CBRAPOS:    case OP_CRMINSTAR:
873      case OP_SCBRAPOS:    case OP_CRMINPLUS:
874      common->localptrs[cc - common->start] = localptr;    case OP_CRQUERY:
875      localptr += sizeof(sljit_w);    case OP_CRMINQUERY:
876      cc += 1 + LINK_SIZE + IMM2_SIZE;    return 1;
     break;  
877    
878      case OP_COND:    case OP_CRRANGE:
879      /* Might be a hidden SCOND. */    case OP_CRMINRANGE:
880      alternative = cc + GET(cc, 1);    if (GET2(cc, 1) == GET2(cc, 1 + IMM2_SIZE))
881      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)      return 0;
882        {    return 2;
       common->localptrs[cc - common->start] = localptr;  
       localptr += sizeof(sljit_w);  
       }  
     cc += 1 + LINK_SIZE;  
     break;  
883    
884      default:    default:
885      cc = next_opcode(common, cc);    return 0;
     SLJIT_ASSERT(cc != NULL);  
     break;  
     }  
886    }    }
887  }  }
888    
889  /* Returns with -1 if no need for frame. */  static BOOL detect_repeat(compiler_common *common, pcre_uchar *begin)
 static int get_framesize(compiler_common *common, pcre_uchar *cc, BOOL recursive)  
890  {  {
891  pcre_uchar *ccend = bracketend(cc);  pcre_uchar *end = bracketend(begin);
892  int length = 0;  pcre_uchar *next;
893  BOOL possessive = FALSE;  pcre_uchar *next_end;
894  BOOL setsom_found = recursive;  pcre_uchar *max_end;
895  BOOL setmark_found = recursive;  pcre_uchar type;
896    sljit_sw length = end - begin;
897    int min, max, i;
898    
899  if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))  /* Detect fixed iterations first. */
900    if (end[-(1 + LINK_SIZE)] != OP_KET)
901      return FALSE;
902    
903    /* Already detected repeat. */
904    if (common->private_data_ptrs[end - common->start - LINK_SIZE] != 0)
905      return TRUE;
906    
907    next = end;
908    min = 1;
909    while (1)
910    {    {
911    length = 3;    if (*next != *begin)
912    possessive = TRUE;      break;
913      next_end = bracketend(next);
914      if (next_end - next != length || memcmp(begin, next, IN_UCHARS(length)) != 0)
915        break;
916      next = next_end;
917      min++;
918      }
919    
920    if (min == 2)
921      return FALSE;
922    
923    max = 0;
924    max_end = next;
925    if (*next == OP_BRAZERO || *next == OP_BRAMINZERO)
926      {
927      type = *next;
928      while (1)
929        {
930        if (next[0] != type || next[1] != OP_BRA || next[2 + LINK_SIZE] != *begin)
931          break;
932        next_end = bracketend(next + 2 + LINK_SIZE);
933        if (next_end - next != (length + 2 + LINK_SIZE) || memcmp(begin, next + 2 + LINK_SIZE, IN_UCHARS(length)) != 0)
934          break;
935        next = next_end;
936        max++;
937        }
938    
939      if (next[0] == type && next[1] == *begin && max >= 1)
940        {
941        next_end = bracketend(next + 1);
942        if (next_end - next == (length + 1) && memcmp(begin, next + 1, IN_UCHARS(length)) == 0)
943          {
944          for (i = 0; i < max; i++, next_end += 1 + LINK_SIZE)
945            if (*next_end != OP_KET)
946              break;
947    
948          if (i == max)
949            {
950            common->private_data_ptrs[max_end - common->start - LINK_SIZE] = next_end - max_end;
951            common->private_data_ptrs[max_end - common->start - LINK_SIZE + 1] = (type == OP_BRAZERO) ? OP_UPTO : OP_MINUPTO;
952            /* +2 the original and the last. */
953            common->private_data_ptrs[max_end - common->start - LINK_SIZE + 2] = max + 2;
954            if (min == 1)
955              return TRUE;
956            min--;
957            max_end -= (1 + LINK_SIZE) + GET(max_end, -LINK_SIZE);
958            }
959          }
960        }
961      }
962    
963    if (min >= 3)
964      {
965      common->private_data_ptrs[end - common->start - LINK_SIZE] = max_end - end;
966      common->private_data_ptrs[end - common->start - LINK_SIZE + 1] = OP_EXACT;
967      common->private_data_ptrs[end - common->start - LINK_SIZE + 2] = min;
968      return TRUE;
969      }
970    
971    return FALSE;
972    }
973    
974    #define CASE_ITERATOR_PRIVATE_DATA_1 \
975        case OP_MINSTAR: \
976        case OP_MINPLUS: \
977        case OP_QUERY: \
978        case OP_MINQUERY: \
979        case OP_MINSTARI: \
980        case OP_MINPLUSI: \
981        case OP_QUERYI: \
982        case OP_MINQUERYI: \
983        case OP_NOTMINSTAR: \
984        case OP_NOTMINPLUS: \
985        case OP_NOTQUERY: \
986        case OP_NOTMINQUERY: \
987        case OP_NOTMINSTARI: \
988        case OP_NOTMINPLUSI: \
989        case OP_NOTQUERYI: \
990        case OP_NOTMINQUERYI:
991    
992    #define CASE_ITERATOR_PRIVATE_DATA_2A \
993        case OP_STAR: \
994        case OP_PLUS: \
995        case OP_STARI: \
996        case OP_PLUSI: \
997        case OP_NOTSTAR: \
998        case OP_NOTPLUS: \
999        case OP_NOTSTARI: \
1000        case OP_NOTPLUSI:
1001    
1002    #define CASE_ITERATOR_PRIVATE_DATA_2B \
1003        case OP_UPTO: \
1004        case OP_MINUPTO: \
1005        case OP_UPTOI: \
1006        case OP_MINUPTOI: \
1007        case OP_NOTUPTO: \
1008        case OP_NOTMINUPTO: \
1009        case OP_NOTUPTOI: \
1010        case OP_NOTMINUPTOI:
1011    
1012    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_1 \
1013        case OP_TYPEMINSTAR: \
1014        case OP_TYPEMINPLUS: \
1015        case OP_TYPEQUERY: \
1016        case OP_TYPEMINQUERY:
1017    
1018    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2A \
1019        case OP_TYPESTAR: \
1020        case OP_TYPEPLUS:
1021    
1022    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2B \
1023        case OP_TYPEUPTO: \
1024        case OP_TYPEMINUPTO:
1025    
1026    static void set_private_data_ptrs(compiler_common *common, int *private_data_start, pcre_uchar *ccend)
1027    {
1028    pcre_uchar *cc = common->start;
1029    pcre_uchar *alternative;
1030    pcre_uchar *end = NULL;
1031    int private_data_ptr = *private_data_start;
1032    int space, size, bracketlen;
1033    
1034    while (cc < ccend)
1035      {
1036      space = 0;
1037      size = 0;
1038      bracketlen = 0;
1039      if (private_data_ptr > SLJIT_MAX_LOCAL_SIZE)
1040        return;
1041    
1042      if (*cc == OP_ONCE || *cc == OP_ONCE_NC || *cc == OP_BRA || *cc == OP_CBRA || *cc == OP_COND)
1043        if (detect_repeat(common, cc))
1044          {
1045          /* These brackets are converted to repeats, so no global
1046          based single character repeat is allowed. */
1047          if (cc >= end)
1048            end = bracketend(cc);
1049          }
1050    
1051      switch(*cc)
1052        {
1053        case OP_KET:
1054        if (common->private_data_ptrs[cc + 1 - common->start] != 0)
1055          {
1056          common->private_data_ptrs[cc - common->start] = private_data_ptr;
1057          private_data_ptr += sizeof(sljit_sw);
1058          cc += common->private_data_ptrs[cc + 1 - common->start];
1059          }
1060        cc += 1 + LINK_SIZE;
1061        break;
1062    
1063        case OP_ASSERT:
1064        case OP_ASSERT_NOT:
1065        case OP_ASSERTBACK:
1066        case OP_ASSERTBACK_NOT:
1067        case OP_ONCE:
1068        case OP_ONCE_NC:
1069        case OP_BRAPOS:
1070        case OP_SBRA:
1071        case OP_SBRAPOS:
1072        case OP_SCOND:
1073        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1074        private_data_ptr += sizeof(sljit_sw);
1075        bracketlen = 1 + LINK_SIZE;
1076        break;
1077    
1078        case OP_CBRAPOS:
1079        case OP_SCBRAPOS:
1080        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1081        private_data_ptr += sizeof(sljit_sw);
1082        bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
1083        break;
1084    
1085        case OP_COND:
1086        /* Might be a hidden SCOND. */
1087        alternative = cc + GET(cc, 1);
1088        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1089          {
1090          common->private_data_ptrs[cc - common->start] = private_data_ptr;
1091          private_data_ptr += sizeof(sljit_sw);
1092          }
1093        bracketlen = 1 + LINK_SIZE;
1094        break;
1095    
1096        case OP_BRA:
1097        bracketlen = 1 + LINK_SIZE;
1098        break;
1099    
1100        case OP_CBRA:
1101        case OP_SCBRA:
1102        bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
1103        break;
1104    
1105        CASE_ITERATOR_PRIVATE_DATA_1
1106        space = 1;
1107        size = -2;
1108        break;
1109    
1110        CASE_ITERATOR_PRIVATE_DATA_2A
1111        space = 2;
1112        size = -2;
1113        break;
1114    
1115        CASE_ITERATOR_PRIVATE_DATA_2B
1116        space = 2;
1117        size = -(2 + IMM2_SIZE);
1118        break;
1119    
1120        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1121        space = 1;
1122        size = 1;
1123        break;
1124    
1125        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1126        if (cc[1] != OP_ANYNL && cc[1] != OP_EXTUNI)
1127          space = 2;
1128        size = 1;
1129        break;
1130    
1131        CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1132        if (cc[1 + IMM2_SIZE] != OP_ANYNL && cc[1 + IMM2_SIZE] != OP_EXTUNI)
1133          space = 2;
1134        size = 1 + IMM2_SIZE;
1135        break;
1136    
1137        case OP_CLASS:
1138        case OP_NCLASS:
1139        size += 1 + 32 / sizeof(pcre_uchar);
1140        space = get_class_iterator_size(cc + size);
1141        break;
1142    
1143    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1144        case OP_XCLASS:
1145        size = GET(cc, 1);
1146        space = get_class_iterator_size(cc + size);
1147        break;
1148    #endif
1149    
1150        default:
1151        cc = next_opcode(common, cc);
1152        SLJIT_ASSERT(cc != NULL);
1153        break;
1154        }
1155    
1156      /* Character iterators, which are not inside a repeated bracket,
1157         gets a private slot instead of allocating it on the stack. */
1158      if (space > 0 && cc >= end)
1159        {
1160        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1161        private_data_ptr += sizeof(sljit_sw) * space;
1162        }
1163    
1164      if (size != 0)
1165        {
1166        if (size < 0)
1167          {
1168          cc += -size;
1169    #ifdef SUPPORT_UTF
1170          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1171    #endif
1172          }
1173        else
1174          cc += size;
1175        }
1176    
1177      if (bracketlen > 0)
1178        {
1179        if (cc >= end)
1180          {
1181          end = bracketend(cc);
1182          if (end[-1 - LINK_SIZE] == OP_KET)
1183            end = NULL;
1184          }
1185        cc += bracketlen;
1186        }
1187      }
1188    *private_data_start = private_data_ptr;
1189    }
1190    
1191    /* Returns with a frame_types (always < 0) if no need for frame. */
1192    static int get_framesize(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, BOOL recursive, BOOL* needs_control_head)
1193    {
1194    int length = 0;
1195    int possessive = 0;
1196    BOOL stack_restore = FALSE;
1197    BOOL setsom_found = recursive;
1198    BOOL setmark_found = recursive;
1199    /* The last capture is a local variable even for recursions. */
1200    BOOL capture_last_found = FALSE;
1201    
1202    #if defined DEBUG_FORCE_CONTROL_HEAD && DEBUG_FORCE_CONTROL_HEAD
1203    SLJIT_ASSERT(common->control_head_ptr != 0);
1204    *needs_control_head = TRUE;
1205    #else
1206    *needs_control_head = FALSE;
1207    #endif
1208    
1209    if (ccend == NULL)
1210      {
1211      ccend = bracketend(cc) - (1 + LINK_SIZE);
1212      if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))
1213        {
1214        possessive = length = (common->capture_last_ptr != 0) ? 5 : 3;
1215        /* This is correct regardless of common->capture_last_ptr. */
1216        capture_last_found = TRUE;
1217        }
1218      cc = next_opcode(common, cc);
1219    }    }
1220    
 cc = next_opcode(common, cc);  
1221  SLJIT_ASSERT(cc != NULL);  SLJIT_ASSERT(cc != NULL);
1222  while (cc < ccend)  while (cc < ccend)
1223    switch(*cc)    switch(*cc)
1224      {      {
1225      case OP_SET_SOM:      case OP_SET_SOM:
1226      SLJIT_ASSERT(common->has_set_som);      SLJIT_ASSERT(common->has_set_som);
1227        stack_restore = TRUE;
1228      if (!setsom_found)      if (!setsom_found)
1229        {        {
1230        length += 2;        length += 2;
# Line 814  while (cc < ccend) Line 1234  while (cc < ccend)
1234      break;      break;
1235    
1236      case OP_MARK:      case OP_MARK:
1237        case OP_PRUNE_ARG:
1238        case OP_THEN_ARG:
1239      SLJIT_ASSERT(common->mark_ptr != 0);      SLJIT_ASSERT(common->mark_ptr != 0);
1240        stack_restore = TRUE;
1241      if (!setmark_found)      if (!setmark_found)
1242        {        {
1243        length += 2;        length += 2;
1244        setmark_found = TRUE;        setmark_found = TRUE;
1245        }        }
1246        if (common->control_head_ptr != 0)
1247          *needs_control_head = TRUE;
1248      cc += 1 + 2 + cc[1];      cc += 1 + 2 + cc[1];
1249      break;      break;
1250    
1251      case OP_RECURSE:      case OP_RECURSE:
1252        stack_restore = TRUE;
1253      if (common->has_set_som && !setsom_found)      if (common->has_set_som && !setsom_found)
1254        {        {
1255        length += 2;        length += 2;
# Line 834  while (cc < ccend) Line 1260  while (cc < ccend)
1260        length += 2;        length += 2;
1261        setmark_found = TRUE;        setmark_found = TRUE;
1262        }        }
1263        if (common->capture_last_ptr != 0 && !capture_last_found)
1264          {
1265          length += 2;
1266          capture_last_found = TRUE;
1267          }
1268      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1269      break;      break;
1270    
# Line 841  while (cc < ccend) Line 1272  while (cc < ccend)
1272      case OP_CBRAPOS:      case OP_CBRAPOS:
1273      case OP_SCBRA:      case OP_SCBRA:
1274      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1275        stack_restore = TRUE;
1276        if (common->capture_last_ptr != 0 && !capture_last_found)
1277          {
1278          length += 2;
1279          capture_last_found = TRUE;
1280          }
1281      length += 3;      length += 3;
1282      cc += 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1283      break;      break;
1284    
1285      default:      default:
1286        stack_restore = TRUE;
1287        /* Fall through. */
1288    
1289        case OP_NOT_WORD_BOUNDARY:
1290        case OP_WORD_BOUNDARY:
1291        case OP_NOT_DIGIT:
1292        case OP_DIGIT:
1293        case OP_NOT_WHITESPACE:
1294        case OP_WHITESPACE:
1295        case OP_NOT_WORDCHAR:
1296        case OP_WORDCHAR:
1297        case OP_ANY:
1298        case OP_ALLANY:
1299        case OP_ANYBYTE:
1300        case OP_NOTPROP:
1301        case OP_PROP:
1302        case OP_ANYNL:
1303        case OP_NOT_HSPACE:
1304        case OP_HSPACE:
1305        case OP_NOT_VSPACE:
1306        case OP_VSPACE:
1307        case OP_EXTUNI:
1308        case OP_EODN:
1309        case OP_EOD:
1310        case OP_CIRC:
1311        case OP_CIRCM:
1312        case OP_DOLL:
1313        case OP_DOLLM:
1314        case OP_CHAR:
1315        case OP_CHARI:
1316        case OP_NOT:
1317        case OP_NOTI:
1318    
1319        case OP_EXACT:
1320        case OP_POSSTAR:
1321        case OP_POSPLUS:
1322        case OP_POSQUERY:
1323        case OP_POSUPTO:
1324    
1325        case OP_EXACTI:
1326        case OP_POSSTARI:
1327        case OP_POSPLUSI:
1328        case OP_POSQUERYI:
1329        case OP_POSUPTOI:
1330    
1331        case OP_NOTEXACT:
1332        case OP_NOTPOSSTAR:
1333        case OP_NOTPOSPLUS:
1334        case OP_NOTPOSQUERY:
1335        case OP_NOTPOSUPTO:
1336    
1337        case OP_NOTEXACTI:
1338        case OP_NOTPOSSTARI:
1339        case OP_NOTPOSPLUSI:
1340        case OP_NOTPOSQUERYI:
1341        case OP_NOTPOSUPTOI:
1342    
1343        case OP_TYPEEXACT:
1344        case OP_TYPEPOSSTAR:
1345        case OP_TYPEPOSPLUS:
1346        case OP_TYPEPOSQUERY:
1347        case OP_TYPEPOSUPTO:
1348    
1349        case OP_CLASS:
1350        case OP_NCLASS:
1351        case OP_XCLASS:
1352    
1353      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
1354      SLJIT_ASSERT(cc != NULL);      SLJIT_ASSERT(cc != NULL);
1355      break;      break;
1356      }      }
1357    
1358  /* Possessive quantifiers can use a special case. */  /* Possessive quantifiers can use a special case. */
1359  if (SLJIT_UNLIKELY(possessive) && length == 3)  if (SLJIT_UNLIKELY(possessive == length))
1360    return -1;    return stack_restore ? no_frame : no_stack;
1361    
1362  if (length > 0)  if (length > 0)
1363    return length + 1;    return length + 1;
1364  return -1;  return stack_restore ? no_frame : no_stack;
1365  }  }
1366    
1367  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)
1368  {  {
1369  DEFINE_COMPILER;  DEFINE_COMPILER;
 pcre_uchar *ccend = bracketend(cc);  
1370  BOOL setsom_found = recursive;  BOOL setsom_found = recursive;
1371  BOOL setmark_found = recursive;  BOOL setmark_found = recursive;
1372    /* The last capture is a local variable even for recursions. */
1373    BOOL capture_last_found = FALSE;
1374  int offset;  int offset;
1375    
1376  /* >= 1 + shortest item size (2) */  /* >= 1 + shortest item size (2) */
# Line 873  SLJIT_UNUSED_ARG(stacktop); Line 1378  SLJIT_UNUSED_ARG(stacktop);
1378  SLJIT_ASSERT(stackpos >= stacktop + 2);  SLJIT_ASSERT(stackpos >= stacktop + 2);
1379    
1380  stackpos = STACK(stackpos);  stackpos = STACK(stackpos);
1381  if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))  if (ccend == NULL)
1382    cc = next_opcode(common, cc);    {
1383      ccend = bracketend(cc) - (1 + LINK_SIZE);
1384      if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))
1385        cc = next_opcode(common, cc);
1386      }
1387    
1388  SLJIT_ASSERT(cc != NULL);  SLJIT_ASSERT(cc != NULL);
1389  while (cc < ccend)  while (cc < ccend)
1390    switch(*cc)    switch(*cc)
# Line 884  while (cc < ccend) Line 1394  while (cc < ccend)
1394      if (!setsom_found)      if (!setsom_found)
1395        {        {
1396        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));
1397        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setstrbegin);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1398        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1399        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1400        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1401        setsom_found = TRUE;        setsom_found = TRUE;
1402        }        }
1403      cc += 1;      cc += 1;
1404      break;      break;
1405    
1406      case OP_MARK:      case OP_MARK:
1407        case OP_PRUNE_ARG:
1408        case OP_THEN_ARG:
1409      SLJIT_ASSERT(common->mark_ptr != 0);      SLJIT_ASSERT(common->mark_ptr != 0);
1410      if (!setmark_found)      if (!setmark_found)
1411        {        {
1412        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);
1413        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);
1414        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1415        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1416        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1417        setmark_found = TRUE;        setmark_found = TRUE;
1418        }        }
1419      cc += 1 + 2 + cc[1];      cc += 1 + 2 + cc[1];
# Line 911  while (cc < ccend) Line 1423  while (cc < ccend)
1423      if (common->has_set_som && !setsom_found)      if (common->has_set_som && !setsom_found)
1424        {        {
1425        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));
1426        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setstrbegin);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1427        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1428        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1429        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1430        setsom_found = TRUE;        setsom_found = TRUE;
1431        }        }
1432      if (common->mark_ptr != 0 && !setmark_found)      if (common->mark_ptr != 0 && !setmark_found)
1433        {        {
1434        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);
1435        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);
1436        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1437        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1438        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1439        setmark_found = TRUE;        setmark_found = TRUE;
1440        }        }
1441        if (common->capture_last_ptr != 0 && !capture_last_found)
1442          {
1443          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->capture_last_ptr);
1444          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1445          stackpos += (int)sizeof(sljit_sw);
1446          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1447          stackpos += (int)sizeof(sljit_sw);
1448          capture_last_found = TRUE;
1449          }
1450      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1451      break;      break;
1452    
# Line 933  while (cc < ccend) Line 1454  while (cc < ccend)
1454      case OP_CBRAPOS:      case OP_CBRAPOS:
1455      case OP_SCBRA:      case OP_SCBRA:
1456      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1457        if (common->capture_last_ptr != 0 && !capture_last_found)
1458          {
1459          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->capture_last_ptr);
1460          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1461          stackpos += (int)sizeof(sljit_sw);
1462          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1463          stackpos += (int)sizeof(sljit_sw);
1464          capture_last_found = TRUE;
1465          }
1466      offset = (GET2(cc, 1 + LINK_SIZE)) << 1;      offset = (GET2(cc, 1 + LINK_SIZE)) << 1;
1467      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));
1468      stackpos += (int)sizeof(sljit_w);      stackpos += (int)sizeof(sljit_sw);
1469      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset));      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset));
1470      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));
1471      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1472      stackpos += (int)sizeof(sljit_w);      stackpos += (int)sizeof(sljit_sw);
1473      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);
1474      stackpos += (int)sizeof(sljit_w);      stackpos += (int)sizeof(sljit_sw);
1475    
1476      cc += 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1477      break;      break;
# Line 952  while (cc < ccend) Line 1482  while (cc < ccend)
1482      break;      break;
1483      }      }
1484    
1485  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_end);  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, 0);
1486  SLJIT_ASSERT(stackpos == STACK(stacktop));  SLJIT_ASSERT(stackpos == STACK(stacktop));
1487  }  }
1488    
1489  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)
1490  {  {
1491  int localsize = 2;  int private_data_length = needs_control_head ? 3 : 2;
1492    int size;
1493  pcre_uchar *alternative;  pcre_uchar *alternative;
1494  /* Calculate the sum of the local variables. */  /* Calculate the sum of the private machine words. */
1495  while (cc < ccend)  while (cc < ccend)
1496    {    {
1497      size = 0;
1498    switch(*cc)    switch(*cc)
1499      {      {
1500        case OP_KET:
1501        if (PRIVATE_DATA(cc) != 0)
1502          private_data_length++;
1503        cc += 1 + LINK_SIZE;
1504        break;
1505    
1506      case OP_ASSERT:      case OP_ASSERT:
1507      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1508      case OP_ASSERTBACK:      case OP_ASSERTBACK:
# Line 975  while (cc < ccend) Line 1513  while (cc < ccend)
1513      case OP_SBRA:      case OP_SBRA:
1514      case OP_SBRAPOS:      case OP_SBRAPOS:
1515      case OP_SCOND:      case OP_SCOND:
1516      localsize++;      private_data_length++;
1517      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1518      break;      break;
1519    
1520      case OP_CBRA:      case OP_CBRA:
1521      case OP_SCBRA:      case OP_SCBRA:
1522      localsize++;      if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1523          private_data_length++;
1524      cc += 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1525      break;      break;
1526    
1527      case OP_CBRAPOS:      case OP_CBRAPOS:
1528      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1529      localsize += 2;      private_data_length += 2;
1530      cc += 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1531      break;      break;
1532    
# Line 995  while (cc < ccend) Line 1534  while (cc < ccend)
1534      /* Might be a hidden SCOND. */      /* Might be a hidden SCOND. */
1535      alternative = cc + GET(cc, 1);      alternative = cc + GET(cc, 1);
1536      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1537        localsize++;        private_data_length++;
1538      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1539      break;      break;
1540    
1541      default:      CASE_ITERATOR_PRIVATE_DATA_1
1542      cc = next_opcode(common, cc);      if (PRIVATE_DATA(cc))
1543      SLJIT_ASSERT(cc != NULL);        private_data_length++;
1544        cc += 2;
1545    #ifdef SUPPORT_UTF
1546        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1547    #endif
1548      break;      break;
     }  
   }  
 SLJIT_ASSERT(cc == ccend);  
 return localsize;  
 }  
1549    
1550  static void copy_locals(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend,      CASE_ITERATOR_PRIVATE_DATA_2A
1551    BOOL save, int stackptr, int stacktop)      if (PRIVATE_DATA(cc))
1552  {        private_data_length += 2;
1553  DEFINE_COMPILER;      cc += 2;
1554  int srcw[2];  #ifdef SUPPORT_UTF
1555  int count;      if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1556  BOOL tmp1next = TRUE;  #endif
1557  BOOL tmp1empty = TRUE;      break;
1558  BOOL tmp2empty = TRUE;  
1559  pcre_uchar *alternative;      CASE_ITERATOR_PRIVATE_DATA_2B
1560  enum {      if (PRIVATE_DATA(cc))
1561    start,        private_data_length += 2;
1562    loop,      cc += 2 + IMM2_SIZE;
1563    end  #ifdef SUPPORT_UTF
1564  } status;      if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1565    #endif
1566        break;
1567    
1568        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1569        if (PRIVATE_DATA(cc))
1570          private_data_length++;
1571        cc += 1;
1572        break;
1573    
1574        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1575        if (PRIVATE_DATA(cc))
1576          private_data_length += 2;
1577        cc += 1;
1578        break;
1579    
1580        CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1581        if (PRIVATE_DATA(cc))
1582          private_data_length += 2;
1583        cc += 1 + IMM2_SIZE;
1584        break;
1585    
1586        case OP_CLASS:
1587        case OP_NCLASS:
1588    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1589        case OP_XCLASS:
1590        size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
1591    #else
1592        size = 1 + 32 / (int)sizeof(pcre_uchar);
1593    #endif
1594        if (PRIVATE_DATA(cc))
1595          private_data_length += get_class_iterator_size(cc + size);
1596        cc += size;
1597        break;
1598    
1599        default:
1600        cc = next_opcode(common, cc);
1601        SLJIT_ASSERT(cc != NULL);
1602        break;
1603        }
1604      }
1605    SLJIT_ASSERT(cc == ccend);
1606    return private_data_length;
1607    }
1608    
1609    static void copy_private_data(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend,
1610      BOOL save, int stackptr, int stacktop, BOOL needs_control_head)
1611    {
1612    DEFINE_COMPILER;
1613    int srcw[2];
1614    int count, size;
1615    BOOL tmp1next = TRUE;
1616    BOOL tmp1empty = TRUE;
1617    BOOL tmp2empty = TRUE;
1618    pcre_uchar *alternative;
1619    enum {
1620      start,
1621      loop,
1622      end
1623    } status;
1624    
1625  status = save ? start : loop;  status = save ? start : loop;
1626  stackptr = STACK(stackptr - 2);  stackptr = STACK(stackptr - 2);
# Line 1031  stacktop = STACK(stacktop - 1); Line 1628  stacktop = STACK(stacktop - 1);
1628    
1629  if (!save)  if (!save)
1630    {    {
1631    stackptr += sizeof(sljit_w);    stackptr += (needs_control_head ? 2 : 1) * sizeof(sljit_sw);
1632    if (stackptr < stacktop)    if (stackptr < stacktop)
1633      {      {
1634      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1635      stackptr += sizeof(sljit_w);      stackptr += sizeof(sljit_sw);
1636      tmp1empty = FALSE;      tmp1empty = FALSE;
1637      }      }
1638    if (stackptr < stacktop)    if (stackptr < stacktop)
1639      {      {
1640      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1641      stackptr += sizeof(sljit_w);      stackptr += sizeof(sljit_sw);
1642      tmp2empty = FALSE;      tmp2empty = FALSE;
1643      }      }
1644    /* The tmp1next must be TRUE in either way. */    /* The tmp1next must be TRUE in either way. */
1645    }    }
1646    
1647  while (status != end)  do
1648    {    {
1649    count = 0;    count = 0;
1650    switch(status)    switch(status)
1651      {      {
1652      case start:      case start:
1653      SLJIT_ASSERT(save && common->recursive_head != 0);      SLJIT_ASSERT(save && common->recursive_head_ptr != 0);
1654      count = 1;      count = 1;
1655      srcw[0] = common->recursive_head;      srcw[0] = common->recursive_head_ptr;
1656        if (needs_control_head)
1657          {
1658          SLJIT_ASSERT(common->control_head_ptr != 0);
1659          count = 2;
1660          srcw[1] = common->control_head_ptr;
1661          }
1662      status = loop;      status = loop;
1663      break;      break;
1664    
# Line 1068  while (status != end) Line 1671  while (status != end)
1671    
1672      switch(*cc)      switch(*cc)
1673        {        {
1674          case OP_KET:
1675          if (PRIVATE_DATA(cc) != 0)
1676            {
1677            count = 1;
1678            srcw[0] = PRIVATE_DATA(cc);
1679            }
1680          cc += 1 + LINK_SIZE;
1681          break;
1682    
1683        case OP_ASSERT:        case OP_ASSERT:
1684        case OP_ASSERT_NOT:        case OP_ASSERT_NOT:
1685        case OP_ASSERTBACK:        case OP_ASSERTBACK:
# Line 1079  while (status != end) Line 1691  while (status != end)
1691        case OP_SBRAPOS:        case OP_SBRAPOS:
1692        case OP_SCOND:        case OP_SCOND:
1693        count = 1;        count = 1;
1694        srcw[0] = PRIV_DATA(cc);        srcw[0] = PRIVATE_DATA(cc);
1695        SLJIT_ASSERT(srcw[0] != 0);        SLJIT_ASSERT(srcw[0] != 0);
1696        cc += 1 + LINK_SIZE;        cc += 1 + LINK_SIZE;
1697        break;        break;
1698    
1699        case OP_CBRA:        case OP_CBRA:
1700        case OP_SCBRA:        case OP_SCBRA:
1701        count = 1;        if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1702        srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));          {
1703            count = 1;
1704            srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1705            }
1706        cc += 1 + LINK_SIZE + IMM2_SIZE;        cc += 1 + LINK_SIZE + IMM2_SIZE;
1707        break;        break;
1708    
1709        case OP_CBRAPOS:        case OP_CBRAPOS:
1710        case OP_SCBRAPOS:        case OP_SCBRAPOS:
1711        count = 2;        count = 2;
1712          srcw[0] = PRIVATE_DATA(cc);
1713        srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));        srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1714        srcw[0] = PRIV_DATA(cc);        SLJIT_ASSERT(srcw[0] != 0 && srcw[1] != 0);
       SLJIT_ASSERT(srcw[0] != 0);  
1715        cc += 1 + LINK_SIZE + IMM2_SIZE;        cc += 1 + LINK_SIZE + IMM2_SIZE;
1716        break;        break;
1717    
# Line 1106  while (status != end) Line 1721  while (status != end)
1721        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1722          {          {
1723          count = 1;          count = 1;
1724          srcw[0] = PRIV_DATA(cc);          srcw[0] = PRIVATE_DATA(cc);
1725          SLJIT_ASSERT(srcw[0] != 0);          SLJIT_ASSERT(srcw[0] != 0);
1726          }          }
1727        cc += 1 + LINK_SIZE;        cc += 1 + LINK_SIZE;
1728        break;        break;
1729    
1730          CASE_ITERATOR_PRIVATE_DATA_1
1731          if (PRIVATE_DATA(cc))
1732            {
1733            count = 1;
1734            srcw[0] = PRIVATE_DATA(cc);
1735            }
1736          cc += 2;
1737    #ifdef SUPPORT_UTF
1738          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1739    #endif
1740          break;
1741    
1742          CASE_ITERATOR_PRIVATE_DATA_2A
1743          if (PRIVATE_DATA(cc))
1744            {
1745            count = 2;
1746            srcw[0] = PRIVATE_DATA(cc);
1747            srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1748            }
1749          cc += 2;
1750    #ifdef SUPPORT_UTF
1751          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1752    #endif
1753          break;
1754    
1755          CASE_ITERATOR_PRIVATE_DATA_2B
1756          if (PRIVATE_DATA(cc))
1757            {
1758            count = 2;
1759            srcw[0] = PRIVATE_DATA(cc);
1760            srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1761            }
1762          cc += 2 + IMM2_SIZE;
1763    #ifdef SUPPORT_UTF
1764          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1765    #endif
1766          break;
1767    
1768          CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1769          if (PRIVATE_DATA(cc))
1770            {
1771            count = 1;
1772            srcw[0] = PRIVATE_DATA(cc);
1773            }
1774          cc += 1;
1775          break;
1776    
1777          CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1778          if (PRIVATE_DATA(cc))
1779            {
1780            count = 2;
1781            srcw[0] = PRIVATE_DATA(cc);
1782            srcw[1] = srcw[0] + sizeof(sljit_sw);
1783            }
1784          cc += 1;
1785          break;
1786    
1787          CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1788          if (PRIVATE_DATA(cc))
1789            {
1790            count = 2;
1791            srcw[0] = PRIVATE_DATA(cc);
1792            srcw[1] = srcw[0] + sizeof(sljit_sw);
1793            }
1794          cc += 1 + IMM2_SIZE;
1795          break;
1796    
1797          case OP_CLASS:
1798          case OP_NCLASS:
1799    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1800          case OP_XCLASS:
1801          size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
1802    #else
1803          size = 1 + 32 / (int)sizeof(pcre_uchar);
1804    #endif
1805          if (PRIVATE_DATA(cc))
1806            switch(get_class_iterator_size(cc + size))
1807              {
1808              case 1:
1809              count = 1;
1810              srcw[0] = PRIVATE_DATA(cc);
1811              break;
1812    
1813              case 2:
1814              count = 2;
1815              srcw[0] = PRIVATE_DATA(cc);
1816              srcw[1] = srcw[0] + sizeof(sljit_sw);
1817              break;
1818    
1819              default:
1820              SLJIT_ASSERT_STOP();
1821              break;
1822              }
1823          cc += size;
1824          break;
1825    
1826        default:        default:
1827        cc = next_opcode(common, cc);        cc = next_opcode(common, cc);
1828        SLJIT_ASSERT(cc != NULL);        SLJIT_ASSERT(cc != NULL);
# Line 1134  while (status != end) Line 1845  while (status != end)
1845          if (!tmp1empty)          if (!tmp1empty)
1846            {            {
1847            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1848            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1849            }            }
1850          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);
1851          tmp1empty = FALSE;          tmp1empty = FALSE;
# Line 1145  while (status != end) Line 1856  while (status != end)
1856          if (!tmp2empty)          if (!tmp2empty)
1857            {            {
1858            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1859            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1860            }            }
1861          OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);          OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);
1862          tmp2empty = FALSE;          tmp2empty = FALSE;
# Line 1162  while (status != end) Line 1873  while (status != end)
1873          if (!tmp1empty)          if (!tmp1empty)
1874            {            {
1875            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1876            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1877            }            }
1878          tmp1next = FALSE;          tmp1next = FALSE;
1879          }          }
# Line 1174  while (status != end) Line 1885  while (status != end)
1885          if (!tmp2empty)          if (!tmp2empty)
1886            {            {
1887            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1888            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1889            }            }
1890          tmp1next = TRUE;          tmp1next = TRUE;
1891          }          }
1892        }        }
1893      }      }
1894    }    }
1895    while (status != end);
1896    
1897  if (save)  if (save)
1898    {    {
# Line 1189  if (save) Line 1901  if (save)
1901      if (!tmp1empty)      if (!tmp1empty)
1902        {        {
1903        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1904        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1905        }        }
1906      if (!tmp2empty)      if (!tmp2empty)
1907        {        {
1908        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1909        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1910        }        }
1911      }      }
1912    else    else
# Line 1202  if (save) Line 1914  if (save)
1914      if (!tmp2empty)      if (!tmp2empty)
1915        {        {
1916        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1917        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1918        }        }
1919      if (!tmp1empty)      if (!tmp1empty)
1920        {        {
1921        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1922        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1923        }        }
1924      }      }
1925    }    }
1926  SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));  SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));
1927  }  }
1928    
1929  static SLJIT_INLINE BOOL ispowerof2(unsigned int value)  static SLJIT_INLINE pcre_uchar *set_then_offsets(compiler_common *common, pcre_uchar *cc, pcre_uint8 *current_offset)
1930    {
1931    pcre_uchar *end = bracketend(cc);
1932    BOOL has_alternatives = cc[GET(cc, 1)] == OP_ALT;
1933    
1934    /* Assert captures then. */
1935    if (*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT)
1936      current_offset = NULL;
1937    /* Conditional block does not. */
1938    if (*cc == OP_COND || *cc == OP_SCOND)
1939      has_alternatives = FALSE;
1940    
1941    cc = next_opcode(common, cc);
1942    if (has_alternatives)
1943      current_offset = common->then_offsets + (cc - common->start);
1944    
1945    while (cc < end)
1946      {
1947      if ((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND))
1948        cc = set_then_offsets(common, cc, current_offset);
1949      else
1950        {
1951        if (*cc == OP_ALT && has_alternatives)
1952          current_offset = common->then_offsets + (cc + 1 + LINK_SIZE - common->start);
1953        if (*cc >= OP_THEN && *cc <= OP_THEN_ARG && current_offset != NULL)
1954          *current_offset = 1;
1955        cc = next_opcode(common, cc);
1956        }
1957      }
1958    
1959    return end;
1960    }
1961    
1962    #undef CASE_ITERATOR_PRIVATE_DATA_1
1963    #undef CASE_ITERATOR_PRIVATE_DATA_2A
1964    #undef CASE_ITERATOR_PRIVATE_DATA_2B
1965    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1966    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1967    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1968    
1969    static SLJIT_INLINE BOOL is_powerof2(unsigned int value)
1970  {  {
1971  return (value & (value - 1)) == 0;  return (value & (value - 1)) == 0;
1972  }  }
# Line 1224  static SLJIT_INLINE void set_jumps(jump_ Line 1976  static SLJIT_INLINE void set_jumps(jump_
1976  while (list)  while (list)
1977    {    {
1978    /* sljit_set_label is clever enough to do nothing    /* sljit_set_label is clever enough to do nothing
1979    if either the jump or the label is NULL */    if either the jump or the label is NULL. */
1980    sljit_set_label(list->jump, label);    SET_LABEL(list->jump, label);
1981    list = list->next;    list = list->next;
1982    }    }
1983  }  }
# Line 1241  if (list_item) Line 1993  if (list_item)
1993    }    }
1994  }  }
1995    
1996  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)
1997  {  {
1998  DEFINE_COMPILER;  DEFINE_COMPILER;
1999  stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));  stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));
2000    
2001  if (list_item)  if (list_item)
2002    {    {
   list_item->type = type;  
   list_item->data = data;  
2003    list_item->start = start;    list_item->start = start;
2004    list_item->leave = LABEL();    list_item->quit = LABEL();
2005    list_item->next = common->stubs;    list_item->next = common->stubs;
2006    common->stubs = list_item;    common->stubs = list_item;
2007    }    }
# Line 1265  stub_list* list_item = common->stubs; Line 2015  stub_list* list_item = common->stubs;
2015  while (list_item)  while (list_item)
2016    {    {
2017    JUMPHERE(list_item->start);    JUMPHERE(list_item->start);
2018    switch(list_item->type)    add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));
2019      {    JUMPTO(SLJIT_JUMP, list_item->quit);
     case stack_alloc:  
     add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));  
     break;  
     }  
   JUMPTO(SLJIT_JUMP, list_item->leave);  
2020    list_item = list_item->next;    list_item = list_item->next;
2021    }    }
2022  common->stubs = NULL;  common->stubs = NULL;
2023  }  }
2024    
2025  static SLJIT_INLINE void decrease_call_count(compiler_common *common)  static SLJIT_INLINE void count_match(compiler_common *common)
2026  {  {
2027  DEFINE_COMPILER;  DEFINE_COMPILER;
2028    
2029  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);
2030  add_jump(compiler, &common->calllimit, JUMP(SLJIT_C_ZERO));  add_jump(compiler, &common->calllimit, JUMP(SLJIT_C_ZERO));
2031  }  }
2032    
# Line 1290  static SLJIT_INLINE void allocate_stack( Line 2035  static SLJIT_INLINE void allocate_stack(
2035  /* May destroy all locals and registers except TMP2. */  /* May destroy all locals and registers except TMP2. */
2036  DEFINE_COMPILER;  DEFINE_COMPILER;
2037    
2038  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));
2039  #ifdef DESTROY_REGISTERS  #ifdef DESTROY_REGISTERS
2040  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);
2041  OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);  OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);
# Line 1298  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0); Line 2043  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);
2043  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);
2044  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);
2045  #endif  #endif
2046  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));
2047  }  }
2048    
2049  static SLJIT_INLINE void free_stack(compiler_common *common, int size)  static SLJIT_INLINE void free_stack(compiler_common *common, int size)
2050  {  {
2051  DEFINE_COMPILER;  DEFINE_COMPILER;
2052  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));
2053  }  }
2054    
2055  static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)  static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)
# Line 1312  static SLJIT_INLINE void reset_ovector(c Line 2057  static SLJIT_INLINE void reset_ovector(c
2057  DEFINE_COMPILER;  DEFINE_COMPILER;
2058  struct sljit_label *loop;  struct sljit_label *loop;
2059  int i;  int i;
2060    
2061  /* At this point we can freely use all temporary registers. */  /* At this point we can freely use all temporary registers. */
2062    SLJIT_ASSERT(length > 1);
2063  /* TMP1 returns with begin - 1. */  /* TMP1 returns with begin - 1. */
2064  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));
2065    if (length < 8)
2066      {
2067      for (i = 1; i < length; i++)
2068        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), SLJIT_SCRATCH_REG1, 0);
2069      }
2070    else
2071      {
2072      GET_LOCAL_BASE(SLJIT_SCRATCH_REG2, 0, OVECTOR_START);
2073      OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, length - 1);
2074      loop = LABEL();
2075      OP1(SLJIT_MOVU, SLJIT_MEM1(SLJIT_SCRATCH_REG2), sizeof(sljit_sw), SLJIT_SCRATCH_REG1, 0);
2076      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 1);
2077      JUMPTO(SLJIT_C_NOT_ZERO, loop);
2078      }
2079    }
2080    
2081    static SLJIT_INLINE void do_reset_match(compiler_common *common, int length)
2082    {
2083    DEFINE_COMPILER;
2084    struct sljit_label *loop;
2085    int i;
2086    
2087    SLJIT_ASSERT(length > 1);
2088    /* OVECTOR(1) contains the "string begin - 1" constant. */
2089    if (length > 2)
2090      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));
2091  if (length < 8)  if (length < 8)
2092    {    {
2093    for (i = 0; i < length; i++)    for (i = 2; i < length; i++)
2094      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);
2095    }    }
2096  else  else
2097    {    {
2098    OP2(SLJIT_ADD, SLJIT_TEMPORARY_REG2, 0, SLJIT_LOCALS_REG, 0, SLJIT_IMM, OVECTOR_START - sizeof(sljit_w));    GET_LOCAL_BASE(TMP2, 0, OVECTOR_START + sizeof(sljit_sw));
2099    OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, length);    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_IMM, length - 2);
2100    loop = LABEL();    loop = LABEL();
2101    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);
2102    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);
2103    JUMPTO(SLJIT_C_NOT_ZERO, loop);    JUMPTO(SLJIT_C_NOT_ZERO, loop);
2104    }    }
2105    
2106    OP1(SLJIT_MOV, STACK_TOP, 0, ARGUMENTS, 0);
2107    if (common->mark_ptr != 0)
2108      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr, SLJIT_IMM, 0);
2109    if (common->control_head_ptr != 0)
2110      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->control_head_ptr, SLJIT_IMM, 0);
2111    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(jit_arguments, stack));
2112    OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_ptr);
2113    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(struct sljit_stack, base));
2114    }
2115    
2116    static sljit_sw SLJIT_CALL do_search_mark(sljit_sw *current, const pcre_uchar *skip_arg)
2117    {
2118    while (current != NULL)
2119      {
2120      switch (current[-2])
2121        {
2122        case type_then_trap:
2123        break;
2124    
2125        case type_mark:
2126        if (STRCMP_UC_UC(skip_arg, (pcre_uchar *)current[-3]) == 0)
2127          return current[-4];
2128        break;
2129    
2130        default:
2131        SLJIT_ASSERT_STOP();
2132        break;
2133        }
2134      current = (sljit_sw*)current[-1];
2135      }
2136    return -1;
2137  }  }
2138    
2139  static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)  static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)
2140  {  {
2141  DEFINE_COMPILER;  DEFINE_COMPILER;
2142  struct sljit_label *loop;  struct sljit_label *loop;
2143  struct sljit_jump *earlyexit;  struct sljit_jump *early_quit;
2144    
2145  /* At this point we can freely use all registers. */  /* At this point we can freely use all registers. */
2146  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));
2147  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);
2148    
2149  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, ARGUMENTS, 0);  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG1, 0, ARGUMENTS, 0);
2150  if (common->mark_ptr != 0)  if (common->mark_ptr != 0)
2151    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);
2152  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));
2153  if (common->mark_ptr != 0)  if (common->mark_ptr != 0)
2154    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);
2155  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));
2156  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));
2157  OP2(SLJIT_ADD, SLJIT_SAVED_REG1, 0, SLJIT_LOCALS_REG, 0, SLJIT_IMM, OVECTOR_START);  GET_LOCAL_BASE(SLJIT_SAVED_REG1, 0, OVECTOR_START);
2158  /* Unlikely, but possible */  /* Unlikely, but possible */
2159  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);
2160  loop = LABEL();  loop = LABEL();
2161  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);
2162  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));
2163  /* Copy the integer value to the output buffer */  /* Copy the integer value to the output buffer */
2164  #ifdef COMPILE_PCRE16  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2165  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);
2166  #endif  #endif
2167  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);
2168  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);
2169  JUMPTO(SLJIT_C_NOT_ZERO, loop);  JUMPTO(SLJIT_C_NOT_ZERO, loop);
2170  JUMPHERE(earlyexit);  JUMPHERE(early_quit);
2171    
2172  /* Calculate the return value, which is the maximum ovector value. */  /* Calculate the return value, which is the maximum ovector value. */
2173  if (topbracket > 1)  if (topbracket > 1)
2174    {    {
2175    OP2(SLJIT_ADD, SLJIT_TEMPORARY_REG1, 0, SLJIT_LOCALS_REG, 0, SLJIT_IMM, OVECTOR_START + topbracket * 2 * sizeof(sljit_w));    GET_LOCAL_BASE(SLJIT_SCRATCH_REG1, 0, OVECTOR_START + topbracket * 2 * sizeof(sljit_sw));
2176    OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, topbracket + 1);    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, topbracket + 1);
2177    
2178    /* OVECTOR(0) is never equal to SLJIT_SAVED_REG3. */    /* OVECTOR(0) is never equal to SLJIT_SAVED_REG3. */
2179    loop = LABEL();    loop = LABEL();
2180    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)));
2181    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);
2182    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);
2183    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_TEMPORARY_REG2, 0);    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_SCRATCH_REG2, 0);
2184    }    }
2185  else  else
2186    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, 1);    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, 1);
2187  }  }
2188    
2189  static SLJIT_INLINE void return_with_partial_match(compiler_common *common, struct sljit_label *leave)  static SLJIT_INLINE void return_with_partial_match(compiler_common *common, struct sljit_label *quit)
2190  {  {
2191  DEFINE_COMPILER;  DEFINE_COMPILER;
2192    struct sljit_jump *jump;
2193    
2194  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);
2195  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
2196      && (common->mode == JIT_PARTIAL_SOFT_COMPILE ? common->hit_start != 0 : common->hit_start == 0));
2197    
2198  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG2, 0, ARGUMENTS, 0);  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, ARGUMENTS, 0);
2199  OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, PCRE_ERROR_PARTIAL);  OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, PCRE_ERROR_PARTIAL);
2200  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));
2201  CMPTO(SLJIT_C_LESS, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, 2, leave);  CMPTO(SLJIT_C_SIG_LESS, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 2, quit);
2202    
2203  /* Store match begin and end. */  /* Store match begin and end. */
2204  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));
2205  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));
2206  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);  
2207    jump = CMP(SLJIT_C_SIG_LESS, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 3);
2208    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);
2209    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2210    OP2(SLJIT_ASHR, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, UCHAR_SHIFT);
2211    #endif
2212    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG2), 2 * sizeof(int), SLJIT_SCRATCH_REG3, 0);
2213    JUMPHERE(jump);
2214    
2215    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);
2216  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);
2217  #ifdef COMPILE_PCRE16  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2218  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);
2219  #endif  #endif
2220  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);
2221    
2222  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);
2223  #ifdef COMPILE_PCRE16  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2224  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);
2225  #endif  #endif
2226  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);
2227    
2228  JUMPTO(SLJIT_JUMP, leave);  JUMPTO(SLJIT_JUMP, quit);
2229  }  }
2230    
2231  static SLJIT_INLINE void check_start_used_ptr(compiler_common *common)  static SLJIT_INLINE void check_start_used_ptr(compiler_common *common)
# Line 1520  if (c <= 127 && bit == 0x20) Line 2336  if (c <= 127 && bit == 0x20)
2336    return (0 << 8) | 0x20;    return (0 << 8) | 0x20;
2337    
2338  /* Since c != oc, they must have at least 1 bit difference. */  /* Since c != oc, they must have at least 1 bit difference. */
2339  if (!ispowerof2(bit))  if (!is_powerof2(bit))
2340    return 0;    return 0;
2341    
2342  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
2343    
2344  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
2345  if (common->utf && c > 127)  if (common->utf && c > 127)
# Line 1539  if (common->utf && c > 127) Line 2355  if (common->utf && c > 127)
2355  #endif /* SUPPORT_UTF */  #endif /* SUPPORT_UTF */
2356  return (0 << 8) | bit;  return (0 << 8) | bit;
2357    
2358  #else /* COMPILE_PCRE8 */  #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2359    
 #ifdef COMPILE_PCRE16  
2360  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
2361  if (common->utf && c > 65535)  if (common->utf && c > 65535)
2362    {    {
# Line 1552  if (common->utf && c > 65535) Line 2367  if (common->utf && c > 65535)
2367    }    }
2368  #endif /* SUPPORT_UTF */  #endif /* SUPPORT_UTF */
2369  return (bit < 256) ? ((0 << 8) | bit) : ((1 << 8) | (bit >> 8));  return (bit < 256) ? ((0 << 8) | bit) : ((1 << 8) | (bit >> 8));
 #endif /* COMPILE_PCRE16 */  
2370    
2371  #endif /* COMPILE_PCRE8 */  #endif /* COMPILE_PCRE[8|16|32] */
2372  }  }
2373    
2374  static void check_partial(compiler_common *common, BOOL force)  static void check_partial(compiler_common *common, BOOL force)
2375  {  {
2376  /* Checks whether a partial matching is occured. Does not modify registers. */  /* Checks whether a partial matching is occurred. Does not modify registers. */
2377  DEFINE_COMPILER;  DEFINE_COMPILER;
2378  struct sljit_jump *jump = NULL;  struct sljit_jump *jump = NULL;
2379    
# Line 1574  else if (common->mode == JIT_PARTIAL_SOF Line 2388  else if (common->mode == JIT_PARTIAL_SOF
2388    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);
2389    
2390  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2391    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);
2392  else  else
2393    {    {
2394    if (common->partialmatchlabel != NULL)    if (common->partialmatchlabel != NULL)
# Line 1587  if (jump != NULL) Line 2401  if (jump != NULL)
2401    JUMPHERE(jump);    JUMPHERE(jump);
2402  }  }
2403    
2404  static struct sljit_jump *check_str_end(compiler_common *common)  static void check_str_end(compiler_common *common, jump_list **end_reached)
2405  {  {
2406  /* Does not affect registers. Usually used in a tight spot. */  /* Does not affect registers. Usually used in a tight spot. */
2407  DEFINE_COMPILER;  DEFINE_COMPILER;
2408  struct sljit_jump *jump;  struct sljit_jump *jump;
 struct sljit_jump *nohit;  
 struct sljit_jump *return_value;  
2409    
2410  if (common->mode == JIT_COMPILE)  if (common->mode == JIT_COMPILE)
2411    return CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);    {
2412      add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2413      return;
2414      }
2415    
2416  jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);  jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);
2417  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2418    {    {
2419    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));
2420    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);
2421    JUMPHERE(nohit);    add_jump(compiler, end_reached, JUMP(SLJIT_JUMP));
   return_value = JUMP(SLJIT_JUMP);  
2422    }    }
2423  else  else
2424    {    {
2425    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));
2426    if (common->partialmatchlabel != NULL)    if (common->partialmatchlabel != NULL)
2427      JUMPTO(SLJIT_JUMP, common->partialmatchlabel);      JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2428    else    else
2429      add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));      add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2430    }    }
2431  JUMPHERE(jump);  JUMPHERE(jump);
 return return_value;  
2432  }  }
2433    
2434  static void fallback_at_str_end(compiler_common *common, jump_list **fallbacks)  static void detect_partial_match(compiler_common *common, jump_list **backtracks)
2435  {  {
2436  DEFINE_COMPILER;  DEFINE_COMPILER;
2437  struct sljit_jump *jump;  struct sljit_jump *jump;
2438    
2439  if (common->mode == JIT_COMPILE)  if (common->mode == JIT_COMPILE)
2440    {    {
2441    add_jump(compiler, fallbacks, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));    add_jump(compiler, backtracks, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2442    return;    return;
2443    }    }
2444    
2445  /* Partial matching mode. */  /* Partial matching mode. */
2446  jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);  jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);
2447  add_jump(compiler, fallbacks, 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));
2448  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2449    {    {
2450    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);
2451    add_jump(compiler, fallbacks, JUMP(SLJIT_JUMP));    add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));
2452    }    }
2453  else  else
2454    {    {
# Line 1647  else Line 2460  else
2460  JUMPHERE(jump);  JUMPHERE(jump);
2461  }  }
2462    
2463  static void read_char(compiler_common *common)  static void peek_char(compiler_common *common)
2464  {  {
2465  /* Reads the character into TMP1, updates STR_PTR.  /* Reads the character into TMP1, keeps STR_PTR.
2466  Does not check STR_END. TMP2 Destroyed. */  Does not check STR_END. TMP2 Destroyed. */
2467  DEFINE_COMPILER;  DEFINE_COMPILER;
2468  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2469  struct sljit_jump *jump;  struct sljit_jump *jump;
2470  #endif  #endif
2471    
2472  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2473  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2474  if (common->utf)  if (common->utf)
2475    {    {
 #ifdef COMPILE_PCRE8  
2476    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2477  #else    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
 #ifdef COMPILE_PCRE16  
   jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);  
 #endif  
 #endif /* COMPILE_PCRE8 */  
2478    add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2479      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2480      JUMPHERE(jump);
2481      }
2482    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2483    
2484    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2485    if (common->utf)
2486      {
2487      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2488      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2489      /* TMP2 contains the high surrogate. */
2490      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2491      OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2492      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2493      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2494      OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2495    JUMPHERE(jump);    JUMPHERE(jump);
2496    }    }
2497  #endif  #endif
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  
2498  }  }
2499    
2500  static void peek_char(compiler_common *common)  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2501    
2502    static BOOL is_char7_bitset(const pcre_uint8* bitset, BOOL nclass)
2503  {  {
2504  /* Reads the character into TMP1, keeps STR_PTR.  /* Tells whether the character codes below 128 are enough
2505  Does not check STR_END. TMP2 Destroyed. */  to determine a match. */
2506  DEFINE_COMPILER;  const pcre_uint8 value = nclass ? 0xff : 0;
2507  #ifdef SUPPORT_UTF  const pcre_uint8* end = bitset + 32;
 struct sljit_jump *jump;  
 #endif  
2508    
2509  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  bitset += 16;
2510  #ifdef SUPPORT_UTF  do
 if (common->utf)  
2511    {    {
2512  #ifdef COMPILE_PCRE8    if (*bitset++ != value)
2513    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);      return FALSE;
 #else  
 #ifdef COMPILE_PCRE16  
   jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);  
 #endif  
 #endif /* COMPILE_PCRE8 */  
   add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));  
   OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);  
   JUMPHERE(jump);  
2514    }    }
2515  #endif  while (bitset < end);
2516    return TRUE;
2517  }  }
2518    
2519  static void read_char8_type(compiler_common *common)  static void read_char7_type(compiler_common *common, BOOL full_read)
2520  {  {
2521  /* Reads the character type into TMP1, updates STR_PTR. Does not check STR_END. */  /* Reads the precise character type of a character into TMP1, if the character
2522    is less than 128. Otherwise it returns with zero. Does not check STR_END. The
2523    full_read argument tells whether characters above max are accepted or not. */
2524  DEFINE_COMPILER;  DEFINE_COMPILER;
 #if defined SUPPORT_UTF || defined COMPILE_PCRE16  
2525  struct sljit_jump *jump;  struct sljit_jump *jump;
 #endif  
2526    
2527  #ifdef SUPPORT_UTF  SLJIT_ASSERT(common->utf);
2528  if (common->utf)  
2529    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2530    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2531    
2532    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2533    
2534    if (full_read)
2535    {    {
   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  
   /* This can be an extra read in some situations, but hopefully  
   it is needed in most cases. */  
   OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);  
2536    jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);    jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2537    add_jump(compiler, &common->utfreadtype8, JUMP(SLJIT_FAST_CALL));    OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
   JUMPHERE(jump);  
 #else  
 #ifdef COMPILE_PCRE16  
   OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);  
   jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);  
   OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);  
   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);  
2538    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2539  #endif    JUMPHERE(jump);
 #endif /* COMPILE_PCRE8 */  
   return;  
2540    }    }
 #endif  
 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_PCRE16  
 /* The ctypes array contains only 256 values. */  
 OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);  
 jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);  
 #endif  
 OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);  
 #ifdef COMPILE_PCRE16  
 JUMPHERE(jump);  
 #endif  
2541  }  }
2542    
2543  static void skip_char_back(compiler_common *common)  #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2544    
2545    static void read_char_max(compiler_common *common, pcre_uint32 max, BOOL full_read)
2546  {  {
2547  /* Goes one character back. Affects STR_PTR and TMP1. Does not check begin. */  /* Reads the precise value of a character into TMP1, if the character is
2548    less than or equal to max. Otherwise it returns with a value greater than max.
2549    Does not check STR_END. The full_read argument tells whether characters above
2550    max are accepted or not. */
2551  DEFINE_COMPILER;  DEFINE_COMPILER;
2552  #if defined SUPPORT_UTF && defined COMPILE_PCRE8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2553  struct sljit_label *label;  struct sljit_jump *jump;
2554    #endif
2555    
2556  if (common->utf)  SLJIT_UNUSED_ARG(full_read);
2557    {  SLJIT_UNUSED_ARG(max);
2558    
2559    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2560    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2561    
2562    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2563    if (common->utf)
2564      {
2565      if (max < 128 && !full_read)
2566        return;
2567    
2568      jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2569      if (max >= 0x800)
2570        add_jump(compiler, (max < 0x10000) ? &common->utfreadchar16 : &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2571      else if (max < 128)
2572        {
2573        OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2574        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2575        }
2576      else
2577        {
2578        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2579        if (!full_read)
2580          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2581        else
2582          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2583        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2584        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2585        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2586        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2587        if (full_read)
2588          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2589        }
2590      JUMPHERE(jump);
2591      }
2592    #endif
2593    
2594    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2595    if (common->utf)
2596      {
2597      if (max >= 0x10000)
2598        {
2599        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2600        jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2601        /* TMP2 contains the high surrogate. */
2602        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2603        OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2604        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2605        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2606        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2607        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2608        JUMPHERE(jump);
2609        return;
2610        }
2611    
2612      if (max < 0xd800 && !full_read)
2613        return;
2614    
2615      /* Skip low surrogate if necessary. */
2616      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2617      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2618      if (full_read)
2619        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2620      if (max >= 0xd800)
2621        OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0x10000);
2622      JUMPHERE(jump);
2623      }
2624    #endif
2625    }
2626    
2627    static SLJIT_INLINE void read_char(compiler_common *common)
2628    {
2629    read_char_max(common, 0x7fffffff, TRUE);
2630    }
2631    
2632    static void read_char8_type(compiler_common *common, BOOL full_read)
2633    {
2634    /* Reads the character type into TMP1, updates STR_PTR. Does not check STR_END.
2635    The full_read argument tells whether characters above max are accepted or not. */
2636    DEFINE_COMPILER;
2637    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2638    struct sljit_jump *jump;
2639    #endif
2640    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2641    struct sljit_jump *jump2;
2642    #endif
2643    
2644    SLJIT_UNUSED_ARG(full_read);
2645    
2646    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2647    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2648    
2649    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2650    if (common->utf)
2651      {
2652      /* This can be an extra read in some situations, but hopefully
2653      it is needed in most cases. */
2654      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2655      jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2656      if (!full_read)
2657        {
2658        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2659        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2660        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2661        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2662        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2663        OP2(SLJIT_OR, TMP2, 0, TMP2, 0, TMP1, 0);
2664        OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2665        jump2 = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);
2666        OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2667        JUMPHERE(jump2);
2668        }
2669      else
2670        add_jump(compiler, &common->utfreadtype8, JUMP(SLJIT_FAST_CALL));
2671      JUMPHERE(jump);
2672      return;
2673      }
2674    #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2675    
2676    #if !defined COMPILE_PCRE8
2677    /* The ctypes array contains only 256 values. */
2678    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2679    jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);
2680    #endif
2681    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2682    #if !defined COMPILE_PCRE8
2683    JUMPHERE(jump);
2684    #endif
2685    
2686    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2687    if (common->utf && full_read)
2688      {
2689      /* Skip low surrogate if necessary. */
2690      OP2(SLJIT_SUB, TMP2, 0, TMP2, 0, SLJIT_IMM, 0xd800);
2691      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2692      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2693      JUMPHERE(jump);
2694      }
2695    #endif /* SUPPORT_UTF && COMPILE_PCRE16 */
2696    }
2697    
2698    static void skip_char_back(compiler_common *common)
2699    {
2700    /* Goes one character back. Affects STR_PTR and TMP1. Does not check begin. */
2701    DEFINE_COMPILER;
2702    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2703    #if defined COMPILE_PCRE8
2704    struct sljit_label *label;
2705    
2706    if (common->utf)
2707      {
2708    label = LABEL();    label = LABEL();
2709    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
2710    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));
# Line 1767  if (common->utf) Line 2712  if (common->utf)
2712    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, 0x80, label);    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, 0x80, label);
2713    return;    return;
2714    }    }
2715  #endif  #elif defined COMPILE_PCRE16
 #if defined SUPPORT_UTF && defined COMPILE_PCRE16  
2716  if (common->utf)  if (common->utf)
2717    {    {
2718    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 1776  if (common->utf) Line 2720  if (common->utf)
2720    /* Skip low surrogate if necessary. */    /* Skip low surrogate if necessary. */
2721    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
2722    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);
2723    COND_VALUE(SLJIT_MOV, TMP1, 0, SLJIT_C_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
2724    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
2725    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2726    return;    return;
2727    }    }
2728  #endif  #endif /* COMPILE_PCRE[8|16] */
2729    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2730  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));
2731  }  }
2732    
2733  static void check_newlinechar(compiler_common *common, int nltype, jump_list **fallbacks, BOOL jumpiftrue)  static void check_newlinechar(compiler_common *common, int nltype, jump_list **backtracks, BOOL jumpiftrue)
2734  {  {
2735  /* 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. */
2736  DEFINE_COMPILER;  DEFINE_COMPILER;
# Line 1793  DEFINE_COMPILER; Line 2738  DEFINE_COMPILER;
2738  if (nltype == NLTYPE_ANY)  if (nltype == NLTYPE_ANY)
2739    {    {
2740    add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));
2741    add_jump(compiler, fallbacks, JUMP(jumpiftrue ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));    add_jump(compiler, backtracks, JUMP(jumpiftrue ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));
2742    }    }
2743  else if (nltype == NLTYPE_ANYCRLF)  else if (nltype == NLTYPE_ANYCRLF)
2744    {    {
2745    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_CR);    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_CR);
2746    COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
2747    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);
2748    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);
2749    add_jump(compiler, fallbacks, JUMP(jumpiftrue ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));    add_jump(compiler, backtracks, JUMP(jumpiftrue ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));
2750    }    }
2751  else  else
2752    {    {
2753    SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline < 256);    SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline < 256);
2754    add_jump(compiler, fallbacks, CMP(jumpiftrue ? SLJIT_C_EQUAL : SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, common->newline));    add_jump(compiler, backtracks, CMP(jumpiftrue ? SLJIT_C_EQUAL : SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, common->newline));
2755    }    }
2756  }  }
2757    
2758  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
2759    
2760  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
2761  static void do_utfreadchar(compiler_common *common)  static void do_utfreadchar(compiler_common *common)
2762  {  {
2763  /* Fast decoding a UTF-8 character. TMP1 contains the first byte  /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2764  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. */
2765  DEFINE_COMPILER;  DEFINE_COMPILER;
2766  struct sljit_jump *jump;  struct sljit_jump *jump;
2767    
2768  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2769    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2770    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2771    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2772    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2773    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2774    
2775  /* Searching for the first zero. */  /* Searching for the first zero. */
2776  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);
2777  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_C_NOT_ZERO);
2778  /* Two byte sequence. */  /* Two byte sequence. */
 OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  
2779  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));
2780  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x1f);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(2));
2781    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2782    
2783    JUMPHERE(jump);
2784    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2785    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
2786  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2787  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2788  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);  
2789    
2790  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);
2791  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_C_NOT_ZERO);
2792  /* 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));  
2793  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));
2794  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));  
2795  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
 JUMPHERE(jump);  
2796    
2797  /* Four byte sequence. */  /* Four byte sequence. */
2798  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  JUMPHERE(jump);
2799  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x07);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2800  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 18);  OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2801    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2802    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2803  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);  
2804  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2805  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(4));
2806    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2807    }
2808    
2809    static void do_utfreadchar16(compiler_common *common)
2810    {
2811    /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2812    of the character (>= 0xc0). Return value in TMP1. */
2813    DEFINE_COMPILER;
2814    struct sljit_jump *jump;
2815    
2816    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2817    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2818    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2819    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2820  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);  
2821  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2822  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(3));  
2823  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));  /* Searching for the first zero. */
2824    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
2825    jump = JUMP(SLJIT_C_NOT_ZERO);
2826    /* Two byte sequence. */
2827    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2828    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2829    
2830    JUMPHERE(jump);
2831    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2832    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
2833    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2834  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2835  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2836  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(3));  /* Three byte sequence. */
2837    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2838  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2839  }  }
2840    
# Line 1879  DEFINE_COMPILER; Line 2846  DEFINE_COMPILER;
2846  struct sljit_jump *jump;  struct sljit_jump *jump;
2847  struct sljit_jump *compare;  struct sljit_jump *compare;
2848    
2849  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2850    
2851  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0x20);  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0x20);
2852  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_C_NOT_ZERO);
# Line 1887  jump = JUMP(SLJIT_C_NOT_ZERO); Line 2854  jump = JUMP(SLJIT_C_NOT_ZERO);
2854  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2855  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));
2856  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);
2857    /* The upper 5 bits are known at this point. */
2858    compare = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0x3);
2859  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2860  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2861  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);  
2862  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2863  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2864    
2865  JUMPHERE(compare);  JUMPHERE(compare);
2866  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2867  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2868  JUMPHERE(jump);  
2869    /* We only have types for characters less than 256. */
2870    JUMPHERE(jump);
2871    OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2872    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2873    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2874    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2875    }
2876    
2877    #endif /* COMPILE_PCRE8 */
2878    
2879    #endif /* SUPPORT_UTF */
2880    
2881    #ifdef SUPPORT_UCP
2882    
2883    /* UCD_BLOCK_SIZE must be 128 (see the assert below). */
2884    #define UCD_BLOCK_MASK 127
2885    #define UCD_BLOCK_SHIFT 7
2886    
2887    static void do_getucd(compiler_common *common)
2888    {
2889    /* Search the UCD record for the character comes in TMP1.
2890    Returns chartype in TMP1 and UCD offset in TMP2. */
2891    DEFINE_COMPILER;
2892    
2893    SLJIT_ASSERT(UCD_BLOCK_SIZE == 128 && sizeof(ucd_record) == 8);
2894    
2895    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2896    OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
2897    OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(ucd_stage1));
2898    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);
2899    OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
2900    OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, TMP2, 0);
2901    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_stage2));
2902    OP1(SLJIT_MOV_UH, TMP2, 0, SLJIT_MEM2(TMP2, TMP1), 1);
2903    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_records) + SLJIT_OFFSETOF(ucd_record, chartype));
2904    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM2(TMP1, TMP2), 3);
2905    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2906    }
2907    #endif
2908    
2909    static SLJIT_INLINE struct sljit_label *mainloop_entry(compiler_common *common, BOOL hascrorlf, BOOL firstline)
2910    {
2911    DEFINE_COMPILER;
2912    struct sljit_label *mainloop;
2913    struct sljit_label *newlinelabel = NULL;
2914    struct sljit_jump *start;
2915    struct sljit_jump *end = NULL;
2916    struct sljit_jump *nl = NULL;
2917    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2918    struct sljit_jump *singlechar;
2919    #endif
2920    jump_list *newline = NULL;
2921    BOOL newlinecheck = FALSE;
2922    BOOL readuchar = FALSE;
2923    
2924    if (!(hascrorlf || firstline) && (common->nltype == NLTYPE_ANY ||
2925        common->nltype == NLTYPE_ANYCRLF || common->newline > 255))
2926      newlinecheck = TRUE;
2927    
2928    if (firstline)
2929      {
2930      /* Search for the end of the first line. */
2931      SLJIT_ASSERT(common->first_line_end != 0);
2932      OP1(SLJIT_MOV, TMP3, 0, STR_PTR, 0);
2933    
2934      if (common->nltype == NLTYPE_FIXED && common->newline > 255)
2935        {
2936        mainloop = LABEL();
2937        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2938        end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
2939        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(-1));
2940        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2941        CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, mainloop);
2942        CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, mainloop);
2943        JUMPHERE(end);
2944        OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2945        }
2946      else
2947        {
2948        end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
2949        mainloop = LABEL();
2950        /* Continual stores does not cause data dependency. */
2951        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);
2952        read_char(common);
2953        check_newlinechar(common, common->nltype, &newline, TRUE);
2954        CMPTO(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0, mainloop);
2955        JUMPHERE(end);
2956        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);
2957        set_jumps(newline, LABEL());
2958        }
2959    
2960      OP1(SLJIT_MOV, STR_PTR, 0, TMP3, 0);
2961      }
2962    
2963    start = JUMP(SLJIT_JUMP);
2964    
2965    if (newlinecheck)
2966      {
2967      newlinelabel = LABEL();
2968      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2969      end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
2970      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2971      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, common->newline & 0xff);
2972      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
2973    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2974      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, UCHAR_SHIFT);
2975    #endif
2976      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2977      nl = JUMP(SLJIT_JUMP);
2978      }
2979    
2980    mainloop = LABEL();
2981    
2982    /* Increasing the STR_PTR here requires one less jump in the most common case. */
2983    #ifdef SUPPORT_UTF
2984    if (common->utf) readuchar = TRUE;
2985    #endif
2986    if (newlinecheck) readuchar = TRUE;
2987    
2988    if (readuchar)
2989      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2990    
2991    if (newlinecheck)
2992      CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, newlinelabel);
2993    
2994    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2995    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2996    #if defined COMPILE_PCRE8
2997    if (common->utf)
2998      {
2999      singlechar = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
3000      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
3001      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3002      JUMPHERE(singlechar);
3003      }
3004    #elif defined COMPILE_PCRE16
3005    if (common->utf)
3006      {
3007      singlechar = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);
3008      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
3009      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);
3010      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
3011      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
3012      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3013      JUMPHERE(singlechar);
3014      }
3015    #endif /* COMPILE_PCRE[8|16] */
3016    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
3017    JUMPHERE(start);
3018    
3019    if (newlinecheck)
3020      {
3021      JUMPHERE(end);
3022      JUMPHERE(nl);
3023      }
3024    
3025    return mainloop;
3026    }
3027    
3028    static int scan_prefix(compiler_common *common, pcre_uchar *cc, pcre_uint32 *chars, int max_chars)
3029    {
3030    /* Recursive function, which scans prefix literals. */
3031    int len, repeat, len_save, consumed = 0;
3032    pcre_uint32 caseless, chr, mask;
3033    pcre_uchar *alternative, *cc_save;
3034    BOOL last, any;
3035    
3036    repeat = 1;
3037    while (TRUE)
3038      {
3039      last = TRUE;
3040      any = FALSE;
3041      caseless = 0;
3042      switch (*cc)
3043        {
3044        case OP_CHARI:
3045        caseless = 1;
3046        case OP_CHAR:
3047        last = FALSE;
3048        cc++;
3049        break;
3050    
3051        case OP_SOD:
3052        case OP_SOM:
3053        case OP_SET_SOM:
3054        case OP_NOT_WORD_BOUNDARY:
3055        case OP_WORD_BOUNDARY:
3056        case OP_EODN:
3057        case OP_EOD:
3058        case OP_CIRC:
3059        case OP_CIRCM:
3060        case OP_DOLL:
3061        case OP_DOLLM:
3062        /* Zero width assertions. */
3063        cc++;
3064        continue;
3065    
3066        case OP_PLUS:
3067        case OP_MINPLUS:
3068        case OP_POSPLUS:
3069        cc++;
3070        break;
3071    
3072        case OP_EXACTI:
3073        caseless = 1;
3074        case OP_EXACT:
3075        repeat = GET2(cc, 1);
3076        last = FALSE;
3077        cc += 1 + IMM2_SIZE;
3078        break;
3079    
3080        case OP_PLUSI:
3081        case OP_MINPLUSI:
3082        case OP_POSPLUSI:
3083        caseless = 1;
3084        cc++;
3085        break;
3086    
3087        case OP_KET:
3088        cc += 1 + LINK_SIZE;
3089        continue;
3090    
3091        case OP_ALT:
3092        cc += GET(cc, 1);
3093        continue;
3094    
3095        case OP_ONCE:
3096        case OP_ONCE_NC:
3097        case OP_BRA:
3098        case OP_BRAPOS:
3099        case OP_CBRA:
3100        case OP_CBRAPOS:
3101        alternative = cc + GET(cc, 1);
3102        while (*alternative == OP_ALT)
3103          {
3104          max_chars = scan_prefix(common, alternative + 1 + LINK_SIZE, chars, max_chars);
3105          if (max_chars == 0)
3106            return consumed;
3107          alternative += GET(alternative, 1);
3108          }
3109    
3110        if (*cc == OP_CBRA || *cc == OP_CBRAPOS)
3111          cc += IMM2_SIZE;
3112        cc += 1 + LINK_SIZE;
3113        continue;
3114    
3115        case OP_CLASS:
3116        case OP_NCLASS:
3117        any = TRUE;
3118        cc += 1 + 32 / sizeof(pcre_uchar);
3119        break;
3120    
3121    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3122        case OP_XCLASS:
3123        any = TRUE;
3124        cc += GET(cc, 1);
3125        break;
3126    #endif
3127    
3128        case OP_NOT_DIGIT:
3129        case OP_DIGIT:
3130        case OP_NOT_WHITESPACE:
3131        case OP_WHITESPACE:
3132        case OP_NOT_WORDCHAR:
3133        case OP_WORDCHAR:
3134        case OP_ANY:
3135        case OP_ALLANY:
3136        any = TRUE;
3137        cc++;
3138        break;
3139    
3140    #ifdef SUPPORT_UCP
3141        case OP_NOTPROP:
3142        case OP_PROP:
3143        any = TRUE;
3144        cc += 1 + 2;
3145        break;
3146    #endif
3147    
3148        case OP_TYPEEXACT:
3149        repeat = GET2(cc, 1);
3150        cc += 1 + IMM2_SIZE;
3151        continue;
3152    
3153        default:
3154        return consumed;
3155        }
3156    
3157      if (any)
3158        {
3159    #ifdef SUPPORT_UTF
3160        if (common->utf) return consumed;
3161    #endif
3162    #if defined COMPILE_PCRE8
3163        mask = 0xff;
3164    #elif defined COMPILE_PCRE16
3165        mask = 0xffff;
3166    #elif defined COMPILE_PCRE32
3167        mask = 0xffffffff;
3168    #else
3169        SLJIT_ASSERT_STOP();
3170    #endif
3171    
3172        do
3173          {
3174          chars[0] = mask;
3175          chars[1] = mask;
3176    
3177          if (--max_chars == 0)
3178            return consumed;
3179          consumed++;
3180          chars += 2;
3181          }
3182        while (--repeat > 0);
3183    
3184        repeat = 1;
3185        continue;
3186        }
3187    
3188      len = 1;
3189    #ifdef SUPPORT_UTF
3190      if (common->utf && HAS_EXTRALEN(*cc)) len += GET_EXTRALEN(*cc);
3191    #endif
3192    
3193      if (caseless != 0 && char_has_othercase(common, cc))
3194        {
3195        caseless = char_get_othercase_bit(common, cc);
3196        if (caseless == 0)
3197          return consumed;
3198    #ifdef COMPILE_PCRE8
3199        caseless = ((caseless & 0xff) << 8) | (len - (caseless >> 8));
3200    #else
3201        if ((caseless & 0x100) != 0)
3202          caseless = ((caseless & 0xff) << 16) | (len - (caseless >> 9));
3203        else
3204          caseless = ((caseless & 0xff) << 8) | (len - (caseless >> 9));
3205    #endif
3206        }
3207      else
3208        caseless = 0;
3209    
3210      len_save = len;
3211      cc_save = cc;
3212      while (TRUE)
3213        {
3214        do
3215          {
3216          chr = *cc;
3217    #ifdef COMPILE_PCRE32
3218          if (SLJIT_UNLIKELY(chr == NOTACHAR))
3219            return consumed;
3220    #endif
3221          mask = 0;
3222          if (len == (caseless & 0xff))
3223            {
3224            mask = caseless >> 8;
3225            chr |= mask;
3226            }
3227    
3228  /* We only have types for characters less than 256. */        if (chars[0] == NOTACHAR)
3229  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), (sljit_w)PRIV(utf8_table4) - 0xc0);          {
3230  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);          chars[0] = chr;
3231  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);          chars[1] = mask;
3232  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);          }
3233  }        else
3234            {
3235            mask |= chars[0] ^ chr;
3236            chr |= mask;
3237            chars[0] = chr;
3238            chars[1] |= mask;
3239            }
3240    
3241  #else /* COMPILE_PCRE8 */        len--;
3242          if (--max_chars == 0)
3243            return consumed;
3244          consumed++;
3245          chars += 2;
3246          cc++;
3247          }
3248        while (len > 0);
3249    
3250  #ifdef COMPILE_PCRE16      if (--repeat == 0)
3251  static void do_utfreadchar(compiler_common *common)        break;
 {  
 /* 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;  
3252    
3253  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);      len = len_save;
3254  jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xdc00);      cc = cc_save;
3255  /* Do nothing, only return. */      }
 sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  
3256    
3257  JUMPHERE(jump);    repeat = 1;
3258  /* Combine two 16 bit characters. */    if (last)
3259  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));      return consumed;
3260  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));    }
 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);  
 sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  
3261  }  }
 #endif /* COMPILE_PCRE16 */  
3262    
3263  #endif /* COMPILE_PCRE8 */  #define MAX_N_CHARS 16
3264    
3265  #endif /* SUPPORT_UTF */  static SLJIT_INLINE BOOL fast_forward_first_n_chars(compiler_common *common, BOOL firstline)
3266    {
3267    DEFINE_COMPILER;
3268    struct sljit_label *start;
3269    struct sljit_jump *quit;
3270    pcre_uint32 chars[MAX_N_CHARS * 2];
3271    pcre_uint8 ones[MAX_N_CHARS];
3272    pcre_uint32 mask;
3273    int i, max;
3274    int offsets[3];
3275    
3276  #ifdef SUPPORT_UCP  for (i = 0; i < MAX_N_CHARS; i++)
3277      {
3278      chars[i << 1] = NOTACHAR;
3279      chars[(i << 1) + 1] = 0;
3280      }
3281    
3282  /* UCD_BLOCK_SIZE must be 128 (see the assert below). */  max = scan_prefix(common, common->start, chars, MAX_N_CHARS);
 #define UCD_BLOCK_MASK 127  
 #define UCD_BLOCK_SHIFT 7  
3283    
3284  static void do_getucd(compiler_common *common)  if (max <= 1)
3285  {    return FALSE;
 /* Search the UCD record for the character comes in TMP1.  
 Returns chartype in TMP1 and UCD offset in TMP2. */  
 DEFINE_COMPILER;  
3286    
3287  SLJIT_ASSERT(UCD_BLOCK_SIZE == 128 && sizeof(ucd_record) == 8);  for (i = 0; i < max; i++)
3288      {
3289      mask = chars[(i << 1) + 1];
3290      ones[i] = ones_in_half_byte[mask & 0xf];
3291      mask >>= 4;
3292      while (mask != 0)
3293        {
3294        ones[i] += ones_in_half_byte[mask & 0xf];
3295        mask >>= 4;
3296        }
3297      }
3298    
3299  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  offsets[0] = -1;
3300  OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);  /* Scan forward. */
3301  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_w)PRIV(ucd_stage1));  for (i = 0; i < max; i++)
3302  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);    if (ones[i] <= 2) {
3303  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);      offsets[0] = i;
3304  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, TMP2, 0);      break;
3305  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, (sljit_w)PRIV(ucd_stage2));    }
 OP1(SLJIT_MOV_UH, TMP2, 0, SLJIT_MEM2(TMP2, TMP1), 1);  
 OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, (sljit_w)PRIV(ucd_records) + SLJIT_OFFSETOF(ucd_record, chartype));  
 OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM2(TMP1, TMP2), 3);  
 sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  
 }  
 #endif  
3306    
3307  static SLJIT_INLINE struct sljit_label *mainloop_entry(compiler_common *common, BOOL hascrorlf, BOOL firstline)  if (offsets[0] == -1)
3308  {    return FALSE;
 DEFINE_COMPILER;  
 struct sljit_label *mainloop;  
 struct sljit_label *newlinelabel = NULL;  
 struct sljit_jump *start;  
 struct sljit_jump *end = NULL;  
 struct sljit_jump *nl = NULL;  
 #ifdef SUPPORT_UTF  
 struct sljit_jump *singlechar;  
 #endif  
 jump_list *newline = NULL;  
 BOOL newlinecheck = FALSE;  
 BOOL readuchar = FALSE;  
3309    
3310  if (!(hascrorlf || firstline) && (common->nltype == NLTYPE_ANY ||  /* Scan backward. */
3311      common->nltype == NLTYPE_ANYCRLF || common->newline > 255))  offsets[1] = -1;
3312    newlinecheck = TRUE;  for (i = max - 1; i > offsets[0]; i--)
3313      if (ones[i] <= 2) {
3314        offsets[1] = i;
3315        break;
3316      }
3317    
3318  if (firstline)  offsets[2] = -1;
3319    if (offsets[1] >= 0)
3320    {    {
3321    /* Search for the end of the first line. */    /* Scan from middle. */
3322    SLJIT_ASSERT(common->first_line_end != 0);    for (i = (offsets[0] + offsets[1]) / 2 + 1; i < offsets[1]; i++)
3323    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, STR_PTR, 0);      if (ones[i] <= 2)
3324    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_END, 0);        {
3325          offsets[2] = i;
3326          break;
3327          }
3328    
3329    if (common->nltype == NLTYPE_FIXED && common->newline > 255)    if (offsets[2] == -1)
     {  
     mainloop = LABEL();  
     OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  
     end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);  
     OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(-1));  
     OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));  
     CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, mainloop);  
     CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, mainloop);  
     OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  
     }  
   else  
3330      {      {
3331      end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);      for (i = (offsets[0] + offsets[1]) / 2; i > offsets[0]; i--)
3332      mainloop = LABEL();        if (ones[i] <= 2)
3333      /* Continual stores does not cause data dependency. */          {
3334      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);          offsets[2] = i;
3335      read_char(common);          break;
3336      check_newlinechar(common, common->nltype, &newline, TRUE);          }
     CMPTO(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0, mainloop);  
     OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);  
     set_jumps(newline, LABEL());  
3337      }      }
   
   JUMPHERE(end);  
   OP1(SLJIT_MOV, STR_PTR, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0);  
3338    }    }
3339    
3340  start = JUMP(SLJIT_JUMP);  SLJIT_ASSERT(offsets[1] == -1 || (offsets[0] < offsets[1]));
3341    SLJIT_ASSERT(offsets[2] == -1 || (offsets[0] < offsets[2] && offsets[1] > offsets[2]));
3342    
3343  if (newlinecheck)  chars[0] = chars[offsets[0] << 1];
3344    chars[1] = chars[(offsets[0] << 1) + 1];
3345    if (offsets[2] >= 0)
3346    {    {
3347    newlinelabel = LABEL();    chars[2] = chars[offsets[2] << 1];
3348    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));    chars[3] = chars[(offsets[2] << 1) + 1];
3349    end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);    }
3350    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  if (offsets[1] >= 0)
3351    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, common->newline & 0xff);    {
3352    COND_VALUE(SLJIT_MOV, TMP1, 0, SLJIT_C_EQUAL);    chars[4] = chars[offsets[1] << 1];
3353  #ifdef COMPILE_PCRE16    chars[5] = chars[(offsets[1] << 1) + 1];
   OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);  
 #endif  
   OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);  
   nl = JUMP(SLJIT_JUMP);  
3354    }    }
3355    
3356  mainloop = LABEL();  max -= 1;
3357    if (firstline)
3358      {
3359      SLJIT_ASSERT(common->first_line_end != 0);
3360      OP1(SLJIT_MOV, TMP3, 0, STR_END, 0);
3361      OP2(SLJIT_SUB, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, SLJIT_IMM, IN_UCHARS(max));
3362      }
3363    else
3364      OP2(SLJIT_SUB, STR_END, 0, STR_END, 0, SLJIT_IMM, IN_UCHARS(max));
3365    
3366  /* Increasing the STR_PTR here requires one less jump in the most common case. */  start = LABEL();
3367  #ifdef SUPPORT_UTF  quit = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
 if (common->utf) readuchar = TRUE;  
 #endif  
 if (newlinecheck) readuchar = TRUE;  
3368    
3369  if (readuchar)  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(offsets[0]));
3370    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  if (offsets[1] >= 0)
3371      OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(offsets[1]));
3372    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3373    
3374  if (newlinecheck)  if (chars[1] != 0)
3375    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, newlinelabel);    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, chars[1]);
3376    CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, chars[0], start);
3377    if (offsets[2] >= 0)
3378      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(offsets[2] - 1));
3379    
3380  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  if (offsets[1] >= 0)
 #if defined SUPPORT_UTF && defined COMPILE_PCRE8  
 if (common->utf)  
   {  
   singlechar = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);  
   OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_w)PRIV(utf8_table4) - 0xc0);  
   OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);  
   JUMPHERE(singlechar);  
   }  
 #endif  
 #if defined SUPPORT_UTF && defined COMPILE_PCRE16  
 if (common->utf)  
3381    {    {
3382    singlechar = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);    if (chars[5] != 0)
3383    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);      OP2(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_IMM, chars[5]);
3384    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);    CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, chars[4], start);
   COND_VALUE(SLJIT_MOV, TMP1, 0, SLJIT_C_EQUAL);  
   OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);  
   OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);  
   JUMPHERE(singlechar);  
3385    }    }
 #endif  
 JUMPHERE(start);  
3386    
3387  if (newlinecheck)  if (offsets[2] >= 0)
3388    {    {
3389    JUMPHERE(end);    if (chars[3] != 0)
3390    JUMPHERE(nl);      OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, chars[3]);
3391      CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, chars[2], start);
3392    }    }
3393    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3394    
3395  return mainloop;  JUMPHERE(quit);
3396    
3397    if (firstline)
3398      OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);
3399    else
3400      OP2(SLJIT_ADD, STR_END, 0, STR_END, 0, SLJIT_IMM, IN_UCHARS(max));
3401    return TRUE;
3402  }  }
3403    
3404    #undef MAX_N_CHARS
3405    
3406  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)
3407  {  {
3408  DEFINE_COMPILER;  DEFINE_COMPILER;
3409  struct sljit_label *start;  struct sljit_label *start;
3410  struct sljit_jump *leave;  struct sljit_jump *quit;
3411  struct sljit_jump *found;  struct sljit_jump *found;
3412  pcre_uchar oc, bit;  pcre_uchar oc, bit;
3413    
3414  if (firstline)  if (firstline)
3415    {    {
3416    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), POSSESSIVE0, STR_END, 0);    SLJIT_ASSERT(common->first_line_end != 0);
3417      OP1(SLJIT_MOV, TMP3, 0, STR_END, 0);
3418    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);
3419    }    }
3420    
3421  start = LABEL();  start = LABEL();
3422  leave = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);  quit = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3423  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3424    
3425  oc = first_char;  oc = first_char;
# Line 2117  if (first_char == oc) Line 3436  if (first_char == oc)
3436  else  else
3437    {    {
3438    bit = first_char ^ oc;    bit = first_char ^ oc;
3439    if (ispowerof2(bit))    if (is_powerof2(bit))
3440      {      {
3441      OP2(SLJIT_OR, TMP2, 0, TMP1, 0, SLJIT_IMM, bit);      OP2(SLJIT_OR, TMP2, 0, TMP1, 0, SLJIT_IMM, bit);
3442      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 2125  else Line 3444  else
3444    else    else
3445      {      {
3446      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);
3447      COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_EQUAL);      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
3448      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);
3449      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);
3450      found = JUMP(SLJIT_C_NOT_ZERO);      found = JUMP(SLJIT_C_NOT_ZERO);
3451      }      }
3452    }    }
3453    
3454  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));
 #if defined SUPPORT_UTF && defined COMPILE_PCRE8  
 if (common->utf)  
   {  
   CMPTO(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0, start);  
   OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_w)PRIV(utf8_table4) - 0xc0);  
   OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);  
   }  
 #endif  
 #if defined SUPPORT_UTF && defined COMPILE_PCRE16  
 if (common->utf)  
   {  
   CMPTO(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800, start);  
   OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);  
   OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);  
   COND_VALUE(SLJIT_MOV, TMP1, 0, SLJIT_C_EQUAL);  
   OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);  
   OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);  
   }  
 #endif  
3455  JUMPTO(SLJIT_JUMP, start);  JUMPTO(SLJIT_JUMP, start);
3456  JUMPHERE(found);  JUMPHERE(found);
3457  JUMPHERE(leave);  JUMPHERE(quit);
3458    
3459  if (firstline)  if (firstline)
3460    OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), POSSESSIVE0);    OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);
3461  }  }
3462    
3463  static SLJIT_INLINE void fast_forward_newline(compiler_common *common, BOOL firstline)  static SLJIT_INLINE void fast_forward_newline(compiler_common *common, BOOL firstline)
# Line 2166  DEFINE_COMPILER; Line 3466  DEFINE_COMPILER;
3466  struct sljit_label *loop;  struct sljit_label *loop;
3467  struct sljit_jump *lastchar;  struct sljit_jump *lastchar;
3468  struct sljit_jump *firstchar;  struct sljit_jump *firstchar;
3469  struct sljit_jump *leave;  struct sljit_jump *quit;
3470  struct sljit_jump *foundcr = NULL;  struct sljit_jump *foundcr = NULL;
3471  struct sljit_jump *notfoundnl;  struct sljit_jump *notfoundnl;
3472  jump_list *newline = NULL;  jump_list *newline = NULL;
3473    
3474  if (firstline)  if (firstline)
3475    {    {
3476    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), POSSESSIVE0, STR_END, 0);    SLJIT_ASSERT(common->first_line_end != 0);
3477      OP1(SLJIT_MOV, TMP3, 0, STR_END, 0);
3478    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);
3479    }    }
3480    
# Line 2187  if (common->nltype == NLTYPE_FIXED && co Line 3488  if (common->nltype == NLTYPE_FIXED && co
3488    
3489    OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, IN_UCHARS(2));    OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, IN_UCHARS(2));
3490    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);
3491    COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_GREATER_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_GREATER_EQUAL);
3492  #ifdef COMPILE_PCRE16  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
3493    OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 1);    OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCHAR_SHIFT);
3494  #endif  #endif
3495    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
3496    
3497    loop = LABEL();    loop = LABEL();
3498    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));
3499    leave = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);    quit = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3500    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(-2));    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(-2));
3501    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(-1));    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(-1));
3502    CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, loop);    CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, loop);
3503    CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, loop);    CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, loop);
3504    
3505    JUMPHERE(leave);    JUMPHERE(quit);
3506    JUMPHERE(firstchar);    JUMPHERE(firstchar);
3507    JUMPHERE(lastchar);    JUMPHERE(lastchar);
3508    
# Line 2225  set_jumps(newline, loop); Line 3526  set_jumps(newline, loop);
3526    
3527  if (common->nltype == NLTYPE_ANY || common->nltype == NLTYPE_ANYCRLF)  if (common->nltype == NLTYPE_ANY || common->nltype == NLTYPE_ANYCRLF)
3528    {    {
3529    leave = JUMP(SLJIT_JUMP);    quit = JUMP(SLJIT_JUMP);
3530    JUMPHERE(foundcr);    JUMPHERE(foundcr);
3531    notfoundnl = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);    notfoundnl = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3532    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3533    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);
3534    COND_VALUE(SLJIT_MOV, TMP1, 0, SLJIT_C_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
3535  #ifdef COMPILE_PCRE16  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
3536    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, UCHAR_SHIFT);
3537  #endif  #endif
3538    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3539    JUMPHERE(notfoundnl);    JUMPHERE(notfoundnl);
3540    JUMPHERE(leave);    JUMPHERE(quit);
3541    }    }
3542  JUMPHERE(lastchar);  JUMPHERE(lastchar);
3543  JUMPHERE(firstchar);  JUMPHERE(firstchar);
3544    
3545  if (firstline)  if (firstline)
3546    OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), POSSESSIVE0);    OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);
3547  }  }
3548    
3549  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);
3550    
3551    static SLJIT_INLINE void fast_forward_start_bits(compiler_common *common, pcre_uint8 *start_bits, BOOL firstline)
3552  {  {
3553  DEFINE_COMPILER;  DEFINE_COMPILER;
3554  struct sljit_label *start;  struct sljit_label *start;
3555  struct sljit_jump *leave;  struct sljit_jump *quit;
3556  struct sljit_jump *found;  struct sljit_jump *found = NULL;
3557    jump_list *matches = NULL;
3558  #ifndef COMPILE_PCRE8  #ifndef COMPILE_PCRE8
3559  struct sljit_jump *jump;  struct sljit_jump *jump;
3560  #endif  #endif
3561    
3562  if (firstline)  if (firstline)
3563    {    {
3564    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), POSSESSIVE0, STR_END, 0);    SLJIT_ASSERT(common->first_line_end != 0);
3565      OP1(SLJIT_MOV, RETURN_ADDR, 0, STR_END, 0);
3566    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);
3567    }    }
3568    
3569  start = LABEL();  start = LABEL();
3570  leave = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);  quit = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3571  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3572  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
3573  if (common->utf)  if (common->utf)