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

Diff of /code/trunk/pcre_jit_compile.c

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

revision 993 by zherczeg, Tue Jul 10 04:33:00 2012 UTC revision 1549 by zherczeg, Tue Apr 21 07:13:00 2015 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
53  system files. */  system files. */
54    
55  #define SLJIT_MALLOC(size) (PUBL(malloc))(size)  #define SLJIT_MALLOC(size, allocator_data) (PUBL(malloc))(size)
56  #define SLJIT_FREE(ptr) (PUBL(free))(ptr)  #define SLJIT_FREE(ptr, allocator_data) (PUBL(free))(ptr)
57  #define SLJIT_CONFIG_AUTO 1  #define SLJIT_CONFIG_AUTO 1
58  #define SLJIT_CONFIG_STATIC 1  #define SLJIT_CONFIG_STATIC 1
59  #define SLJIT_VERBOSE 0  #define SLJIT_VERBOSE 0
# Line 65  system files. */ Line 65  system files. */
65  #error Unsupported architecture  #error Unsupported architecture
66  #endif  #endif
67    
68  /* Allocate memory on the stack. Fast, but limited size. */  /* Defines for debugging purposes. */
 #define LOCAL_SPACE_SIZE 32768  
69    
70    /* 1 - Use unoptimized capturing brackets.
71       2 - Enable capture_last_ptr (includes option 1). */
72    /* #define DEBUG_FORCE_UNOPTIMIZED_CBRAS 2 */
73    
74    /* 1 - Always have a control head. */
75    /* #define DEBUG_FORCE_CONTROL_HEAD 1 */
76    
77    /* Allocate memory for the regex stack on the real machine stack.
78    Fast, but limited size. */
79    #define MACHINE_STACK_SIZE 32768
80    
81    /* Growth rate for stack allocated by the OS. Should be the multiply
82    of page size. */
83  #define STACK_GROWTH_RATE 8192  #define STACK_GROWTH_RATE 8192
84    
85  /* Enable to check that the allocation could destroy temporaries. */  /* Enable to check that the allocation could destroy temporaries. */
# Line 89  vertical (sub-expression) (See struct ba Line 101  vertical (sub-expression) (See struct ba
101    
102  The condition checkers are boolean (true/false) checkers. Machine code is generated  The condition checkers are boolean (true/false) checkers. Machine code is generated
103  for the checker itself and for the actions depending on the result of the checker.  for the checker itself and for the actions depending on the result of the checker.
104  The 'true' case is called as the try path (expected path), and the other is called as  The 'true' case is called as the matching path (expected path), and the other is called as
105  the 'backtrack' path. Branch instructions are expesive for all CPUs, so we avoid taken  the 'backtrack' path. Branch instructions are expesive for all CPUs, so we avoid taken
106  branches on the try path.  branches on the matching path.
107    
108   Greedy star operator (*) :   Greedy star operator (*) :
109     Try path: match happens.     Matching path: match happens.
110     Backtrack path: match failed.     Backtrack path: match failed.
111   Non-greedy star operator (*?) :   Non-greedy star operator (*?) :
112     Try path: no need to perform a match.     Matching path: no need to perform a match.
113     Backtrack path: match is required.     Backtrack path: match is required.
114    
115  The following example shows how the code generated for a capturing bracket  The following example shows how the code generated for a capturing bracket
# Line 108  we have the following regular expression Line 120  we have the following regular expression
120    
121  The generated code will be the following:  The generated code will be the following:
122    
123   A try path   A matching path
124   '(' try path (pushing arguments to the stack)   '(' matching path (pushing arguments to the stack)
125   B try path   B matching path
126   ')' try path (pushing arguments to the stack)   ')' matching path (pushing arguments to the stack)
127   D try path   D matching path
128   return with successful match   return with successful match
129    
130   D backtrack path   D backtrack path
131   ')' backtrack path (If we arrived from "C" jump to the backtrack of "C")   ')' backtrack path (If we arrived from "C" jump to the backtrack of "C")
132   B backtrack path   B backtrack path
133   C expected path   C expected path
134   jump to D try path   jump to D matching path
135   C backtrack path   C backtrack path
136   A backtrack path   A backtrack path
137    
# Line 127  The generated code will be the following Line 139  The generated code will be the following
139   code paths. In this way the topmost value on the stack is always belong   code paths. In this way the topmost value on the stack is always belong
140   to the current backtrack code path. The backtrack path must check   to the current backtrack code path. The backtrack path must check
141   whether there is a next alternative. If so, it needs to jump back to   whether there is a next alternative. If so, it needs to jump back to
142   the try path eventually. Otherwise it needs to clear out its own stack   the matching path eventually. Otherwise it needs to clear out its own stack
143   frame and continue the execution on the backtrack code paths.   frame and continue the execution on the backtrack code paths.
144  */  */
145    
146  /*  /*
147  Saved stack frames:  Saved stack frames:
148    
149  Atomic blocks and asserts require reloading the values of local variables  Atomic blocks and asserts require reloading the values of private data
150  when the backtrack mechanism performed. Because of OP_RECURSE, the locals  when the backtrack mechanism performed. Because of OP_RECURSE, the data
151  are not necessarly known in compile time, thus we need a dynamic restore  are not necessarly known in compile time, thus we need a dynamic restore
152  mechanism.  mechanism.
153    
154  The stack frames are stored in a chain list, and have the following format:  The stack frames are stored in a chain list, and have the following format:
155  ([ capturing bracket offset ][ start value ][ end value ])+ ... [ 0 ] [ previous head ]  ([ capturing bracket offset ][ start value ][ end value ])+ ... [ 0 ] [ previous head ]
156    
157  Thus we can restore the locals to a particular point in the stack.  Thus we can restore the private data to a particular point in the stack.
158  */  */
159    
160  typedef struct jit_arguments {  typedef struct jit_arguments {
# Line 154  typedef struct jit_arguments { Line 166  typedef struct jit_arguments {
166    int *offsets;    int *offsets;
167    pcre_uchar *uchar_ptr;    pcre_uchar *uchar_ptr;
168    pcre_uchar *mark_ptr;    pcre_uchar *mark_ptr;
169      void *callout_data;
170    /* Everything else after. */    /* Everything else after. */
171    int offsetcount;    pcre_uint32 limit_match;
172    int calllimit;    int real_offset_count;
173      int offset_count;
174    pcre_uint8 notbol;    pcre_uint8 notbol;
175    pcre_uint8 noteol;    pcre_uint8 noteol;
176    pcre_uint8 notempty;    pcre_uint8 notempty;
# Line 165  typedef struct jit_arguments { Line 179  typedef struct jit_arguments {
179    
180  typedef struct executable_functions {  typedef struct executable_functions {
181    void *executable_funcs[JIT_NUMBER_OF_COMPILE_MODES];    void *executable_funcs[JIT_NUMBER_OF_COMPILE_MODES];
182      void *read_only_data_heads[JIT_NUMBER_OF_COMPILE_MODES];
183      sljit_uw executable_sizes[JIT_NUMBER_OF_COMPILE_MODES];
184    PUBL(jit_callback) callback;    PUBL(jit_callback) callback;
185    void *userdata;    void *userdata;
186    sljit_uw executable_sizes[JIT_NUMBER_OF_COMPILE_MODES];    pcre_uint32 top_bracket;
187      pcre_uint32 limit_match;
188  } executable_functions;  } executable_functions;
189    
190  typedef struct jump_list {  typedef struct jump_list {
# Line 175  typedef struct jump_list { Line 192  typedef struct jump_list {
192    struct jump_list *next;    struct jump_list *next;
193  } jump_list;  } jump_list;
194    
 enum stub_types { stack_alloc };  
   
195  typedef struct stub_list {  typedef struct stub_list {
   enum stub_types type;  
   int data;  
196    struct sljit_jump *start;    struct sljit_jump *start;
197    struct sljit_label *quit;    struct sljit_label *quit;
198    struct stub_list *next;    struct stub_list *next;
199  } stub_list;  } stub_list;
200    
201    typedef struct label_addr_list {
202      struct sljit_label *label;
203      sljit_uw *update_addr;
204      struct label_addr_list *next;
205    } label_addr_list;
206    
207    enum frame_types {
208      no_frame = -1,
209      no_stack = -2
210    };
211    
212    enum control_types {
213      type_mark = 0,
214      type_then_trap = 1
215    };
216    
217  typedef int (SLJIT_CALL *jit_function)(jit_arguments *args);  typedef int (SLJIT_CALL *jit_function)(jit_arguments *args);
218    
219  /* The following structure is the key data type for the recursive  /* The following structure is the key data type for the recursive
220  code generator. It is allocated by compile_trypath, and contains  code generator. It is allocated by compile_matchingpath, and contains
221  the aguments for compile_backtrackpath. Must be the first member  the arguments for compile_backtrackingpath. Must be the first member
222  of its descendants. */  of its descendants. */
223  typedef struct backtrack_common {  typedef struct backtrack_common {
224    /* Concatenation stack. */    /* Concatenation stack. */
# Line 205  typedef struct backtrack_common { Line 234  typedef struct backtrack_common {
234  typedef struct assert_backtrack {  typedef struct assert_backtrack {
235    backtrack_common common;    backtrack_common common;
236    jump_list *condfailed;    jump_list *condfailed;
237    /* Less than 0 (-1) if a frame is not needed. */    /* Less than 0 if a frame is not needed. */
238    int framesize;    int framesize;
239    /* Points to our private memory word on the stack. */    /* Points to our private memory word on the stack. */
240    int localptr;    int private_data_ptr;
241    /* For iterators. */    /* For iterators. */
242    struct sljit_label *trypath;    struct sljit_label *matchingpath;
243  } assert_backtrack;  } assert_backtrack;
244    
245  typedef struct bracket_backtrack {  typedef struct bracket_backtrack {
246    backtrack_common common;    backtrack_common common;
247    /* Where to coninue if an alternative is successfully matched. */    /* Where to coninue if an alternative is successfully matched. */
248    struct sljit_label *alttrypath;    struct sljit_label *alternative_matchingpath;
249    /* For rmin and rmax iterators. */    /* For rmin and rmax iterators. */
250    struct sljit_label *recursivetrypath;    struct sljit_label *recursive_matchingpath;
251    /* For greedy ? operator. */    /* For greedy ? operator. */
252    struct sljit_label *zerotrypath;    struct sljit_label *zero_matchingpath;
253    /* Contains the branches of a failed condition. */    /* Contains the branches of a failed condition. */
254    union {    union {
255      /* Both for OP_COND, OP_SCOND. */      /* Both for OP_COND, OP_SCOND. */
256      jump_list *condfailed;      jump_list *condfailed;
257      assert_backtrack *assert;      assert_backtrack *assert;
258      /* For OP_ONCE. -1 if not needed. */      /* For OP_ONCE. Less than 0 if not needed. */
259      int framesize;      int framesize;
260    } u;    } u;
261    /* Points to our private memory word on the stack. */    /* Points to our private memory word on the stack. */
262    int localptr;    int private_data_ptr;
263  } bracket_backtrack;  } bracket_backtrack;
264    
265  typedef struct bracketpos_backtrack {  typedef struct bracketpos_backtrack {
266    backtrack_common common;    backtrack_common common;
267    /* Points to our private memory word on the stack. */    /* Points to our private memory word on the stack. */
268    int localptr;    int private_data_ptr;
269    /* Reverting stack is needed. */    /* Reverting stack is needed. */
270    int framesize;    int framesize;
271    /* Allocated stack size. */    /* Allocated stack size. */
# Line 245  typedef struct bracketpos_backtrack { Line 274  typedef struct bracketpos_backtrack {
274    
275  typedef struct braminzero_backtrack {  typedef struct braminzero_backtrack {
276    backtrack_common common;    backtrack_common common;
277    struct sljit_label *trypath;    struct sljit_label *matchingpath;
278  } braminzero_backtrack;  } braminzero_backtrack;
279    
280  typedef struct iterator_backtrack {  typedef struct iterator_backtrack {
281    backtrack_common common;    backtrack_common common;
282    /* Next iteration. */    /* Next iteration. */
283    struct sljit_label *trypath;    struct sljit_label *matchingpath;
284  } iterator_backtrack;  } iterator_backtrack;
285    
286  typedef struct recurse_entry {  typedef struct recurse_entry {
# Line 261  typedef struct recurse_entry { Line 290  typedef struct recurse_entry {
290    /* Collects the calls until the function is not created. */    /* Collects the calls until the function is not created. */
291    jump_list *calls;    jump_list *calls;
292    /* Points to the starting opcode. */    /* Points to the starting opcode. */
293    int start;    sljit_sw start;
294  } recurse_entry;  } recurse_entry;
295    
296  typedef struct recurse_backtrack {  typedef struct recurse_backtrack {
297    backtrack_common common;    backtrack_common common;
298      BOOL inlined_pattern;
299  } recurse_backtrack;  } recurse_backtrack;
300    
301  #define MAX_RANGE_SIZE 6  #define OP_THEN_TRAP OP_TABLE_LENGTH
302    
303    typedef struct then_trap_backtrack {
304      backtrack_common common;
305      /* If then_trap is not NULL, this structure contains the real
306      then_trap for the backtracking path. */
307      struct then_trap_backtrack *then_trap;
308      /* Points to the starting opcode. */
309      sljit_sw start;
310      /* Exit point for the then opcodes of this alternative. */
311      jump_list *quit;
312      /* Frame size of the current alternative. */
313      int framesize;
314    } then_trap_backtrack;
315    
316    #define MAX_RANGE_SIZE 4
317    
318  typedef struct compiler_common {  typedef struct compiler_common {
319      /* The sljit ceneric compiler. */
320    struct sljit_compiler *compiler;    struct sljit_compiler *compiler;
321      /* First byte code. */
322    pcre_uchar *start;    pcre_uchar *start;
323      /* Maps private data offset to each opcode. */
324    /* Opcode local area direct map. */    sljit_si *private_data_ptrs;
325    int *localptrs;    /* Chain list of read-only data ptrs. */
326    int cbraptr;    void *read_only_data_head;
327    /* OVector starting point. Must be divisible by 2. */    /* Tells whether the capturing bracket is optimized. */
328      pcre_uint8 *optimized_cbracket;
329      /* Tells whether the starting offset is a target of then. */
330      pcre_uint8 *then_offsets;
331      /* Current position where a THEN must jump. */
332      then_trap_backtrack *then_trap;
333      /* Starting offset of private data for capturing brackets. */
334      int cbra_ptr;
335      /* Output vector starting point. Must be divisible by 2. */
336    int ovector_start;    int ovector_start;
337    /* Last known position of the requested byte. */    /* Last known position of the requested byte. */
338    int req_char_ptr;    int req_char_ptr;
339    /* Head of the last recursion. */    /* Head of the last recursion. */
340    int recursive_head;    int recursive_head_ptr;
341    /* First inspected character for partial matching. */    /* First inspected character for partial matching. */
342    int start_used_ptr;    int start_used_ptr;
343    /* Starting pointer for partial soft matches. */    /* Starting pointer for partial soft matches. */
# Line 291  typedef struct compiler_common { Line 346  typedef struct compiler_common {
346    int first_line_end;    int first_line_end;
347    /* Points to the marked string. */    /* Points to the marked string. */
348    int mark_ptr;    int mark_ptr;
349      /* Recursive control verb management chain. */
350      int control_head_ptr;
351      /* Points to the last matched capture block index. */
352      int capture_last_ptr;
353      /* Points to the starting position of the current match. */
354      int start_ptr;
355    
356    /* Flipped and lower case tables. */    /* Flipped and lower case tables. */
357    const pcre_uint8 *fcc;    const pcre_uint8 *fcc;
358    sljit_w lcc;    sljit_sw lcc;
359    /* Mode can be PCRE_STUDY_JIT_COMPILE and others. */    /* Mode can be PCRE_STUDY_JIT_COMPILE and others. */
360    int mode;    int mode;
361      /* TRUE, when minlength is greater than 0. */
362      BOOL might_be_empty;
363      /* \K is found in the pattern. */
364      BOOL has_set_som;
365      /* (*SKIP:arg) is found in the pattern. */
366      BOOL has_skip_arg;
367      /* (*THEN) is found in the pattern. */
368      BOOL has_then;
369      /* Needs to know the start position anytime. */
370      BOOL needs_start_ptr;
371      /* Currently in recurse or negative assert. */
372      BOOL local_exit;
373      /* Currently in a positive assert. */
374      BOOL positive_assert;
375    /* Newline control. */    /* Newline control. */
376    int nltype;    int nltype;
377      pcre_uint32 nlmax;
378      pcre_uint32 nlmin;
379    int newline;    int newline;
380    int bsr_nltype;    int bsr_nltype;
381      pcre_uint32 bsr_nlmax;
382      pcre_uint32 bsr_nlmin;
383    /* Dollar endonly. */    /* Dollar endonly. */
384    int endonly;    int endonly;
   BOOL has_set_som;  
385    /* Tables. */    /* Tables. */
386    sljit_w ctypes;    sljit_sw ctypes;
   int digits[2 + MAX_RANGE_SIZE];  
387    /* Named capturing brackets. */    /* Named capturing brackets. */
388    sljit_uw name_table;    pcre_uchar *name_table;
389    sljit_w name_count;    sljit_sw name_count;
390    sljit_w name_entry_size;    sljit_sw name_entry_size;
391    
392    /* Labels and jump lists. */    /* Labels and jump lists. */
393    struct sljit_label *partialmatchlabel;    struct sljit_label *partialmatchlabel;
394    struct sljit_label *quitlabel;    struct sljit_label *quit_label;
395    struct sljit_label *acceptlabel;    struct sljit_label *forced_quit_label;
396      struct sljit_label *accept_label;
397      struct sljit_label *ff_newline_shortcut;
398    stub_list *stubs;    stub_list *stubs;
399      label_addr_list *label_addrs;
400    recurse_entry *entries;    recurse_entry *entries;
401    recurse_entry *currententry;    recurse_entry *currententry;
402    jump_list *partialmatch;    jump_list *partialmatch;
403    jump_list *quit;    jump_list *quit;
404      jump_list *positive_assert_quit;
405      jump_list *forced_quit;
406    jump_list *accept;    jump_list *accept;
407    jump_list *calllimit;    jump_list *calllimit;
408    jump_list *stackalloc;    jump_list *stackalloc;
# Line 331  typedef struct compiler_common { Line 413  typedef struct compiler_common {
413    jump_list *vspace;    jump_list *vspace;
414    jump_list *casefulcmp;    jump_list *casefulcmp;
415    jump_list *caselesscmp;    jump_list *caselesscmp;
416      jump_list *reset_match;
417    BOOL jscript_compat;    BOOL jscript_compat;
418  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
419    BOOL utf;    BOOL utf;
420  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
421    BOOL use_ucp;    BOOL use_ucp;
422  #endif  #endif
   jump_list *utfreadchar;  
423  #ifdef COMPILE_PCRE8  #ifdef COMPILE_PCRE8
424      jump_list *utfreadchar;
425      jump_list *utfreadchar16;
426    jump_list *utfreadtype8;    jump_list *utfreadtype8;
427  #endif  #endif
428  #endif /* SUPPORT_UTF */  #endif /* SUPPORT_UTF */
# Line 355  typedef struct compare_context { Line 439  typedef struct compare_context {
439  #if defined SLJIT_UNALIGNED && SLJIT_UNALIGNED  #if defined SLJIT_UNALIGNED && SLJIT_UNALIGNED
440    int ucharptr;    int ucharptr;
441    union {    union {
442      sljit_i asint;      sljit_si asint;
443      sljit_uh asushort;      sljit_uh asushort;
444  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
445      sljit_ub asbyte;      sljit_ub asbyte;
446      sljit_ub asuchars[4];      sljit_ub asuchars[4];
447  #else  #elif defined COMPILE_PCRE16
 #ifdef COMPILE_PCRE16  
448      sljit_uh asuchars[2];      sljit_uh asuchars[2];
449  #endif  #elif defined COMPILE_PCRE32
450        sljit_ui asuchars[1];
451  #endif  #endif
452    } c;    } c;
453    union {    union {
454      sljit_i asint;      sljit_si asint;
455      sljit_uh asushort;      sljit_uh asushort;
456  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
457      sljit_ub asbyte;      sljit_ub asbyte;
458      sljit_ub asuchars[4];      sljit_ub asuchars[4];
459  #else  #elif defined COMPILE_PCRE16
 #ifdef COMPILE_PCRE16  
460      sljit_uh asuchars[2];      sljit_uh asuchars[2];
461  #endif  #elif defined COMPILE_PCRE32
462        sljit_ui asuchars[1];
463  #endif  #endif
464    } oc;    } oc;
465  #endif  #endif
466  } compare_context;  } compare_context;
467    
 enum {  
   frame_end = 0,  
   frame_setstrbegin = -1,  
   frame_setmark = -2  
 };  
   
468  /* Undefine sljit macros. */  /* Undefine sljit macros. */
469  #undef CMP  #undef CMP
470    
471  /* Used for accessing the elements of the stack. */  /* Used for accessing the elements of the stack. */
472  #define STACK(i)      ((-(i) - 1) * (int)sizeof(sljit_w))  #define STACK(i)      ((-(i) - 1) * (int)sizeof(sljit_sw))
473    
474  #define TMP1          SLJIT_TEMPORARY_REG1  #define TMP1          SLJIT_R0
475  #define TMP2          SLJIT_TEMPORARY_REG3  #define TMP2          SLJIT_R2
476  #define TMP3          SLJIT_TEMPORARY_EREG2  #define TMP3          SLJIT_R3
477  #define STR_PTR       SLJIT_SAVED_REG1  #define STR_PTR       SLJIT_S0
478  #define STR_END       SLJIT_SAVED_REG2  #define STR_END       SLJIT_S1
479  #define STACK_TOP     SLJIT_TEMPORARY_REG2  #define STACK_TOP     SLJIT_R1
480  #define STACK_LIMIT   SLJIT_SAVED_REG3  #define STACK_LIMIT   SLJIT_S2
481  #define ARGUMENTS     SLJIT_SAVED_EREG1  #define COUNT_MATCH   SLJIT_S3
482  #define CALL_COUNT    SLJIT_SAVED_EREG2  #define ARGUMENTS     SLJIT_S4
483  #define RETURN_ADDR   SLJIT_TEMPORARY_EREG1  #define RETURN_ADDR   SLJIT_R4
484    
485  /* Locals layout. */  /* Local space layout. */
486  /* These two locals can be used by the current opcode. */  /* These two locals can be used by the current opcode. */
487  #define LOCALS0          (0 * sizeof(sljit_w))  #define LOCALS0          (0 * sizeof(sljit_sw))
488  #define LOCALS1          (1 * sizeof(sljit_w))  #define LOCALS1          (1 * sizeof(sljit_sw))
489  /* Two local variables for possessive quantifiers (char1 cannot use them). */  /* Two local variables for possessive quantifiers (char1 cannot use them). */
490  #define POSSESSIVE0      (2 * sizeof(sljit_w))  #define POSSESSIVE0      (2 * sizeof(sljit_sw))
491  #define POSSESSIVE1      (3 * sizeof(sljit_w))  #define POSSESSIVE1      (3 * sizeof(sljit_sw))
492  /* Max limit of recursions. */  /* Max limit of recursions. */
493  #define CALL_LIMIT       (4 * sizeof(sljit_w))  #define LIMIT_MATCH      (4 * sizeof(sljit_sw))
494  /* The output vector is stored on the stack, and contains pointers  /* The output vector is stored on the stack, and contains pointers
495  to characters. The vector data is divided into two groups: the first  to characters. The vector data is divided into two groups: the first
496  group contains the start / end character pointers, and the second is  group contains the start / end character pointers, and the second is
497  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. */
498  #define OVECTOR_START    (common->ovector_start)  #define OVECTOR_START    (common->ovector_start)
499  #define OVECTOR(i)       (OVECTOR_START + (i) * sizeof(sljit_w))  #define OVECTOR(i)       (OVECTOR_START + (i) * (sljit_sw)sizeof(sljit_sw))
500  #define OVECTOR_PRIV(i)  (common->cbraptr + (i) * sizeof(sljit_w))  #define OVECTOR_PRIV(i)  (common->cbra_ptr + (i) * (sljit_sw)sizeof(sljit_sw))
501  #define PRIV_DATA(cc)    (common->localptrs[(cc) - common->start])  #define PRIVATE_DATA(cc) (common->private_data_ptrs[(cc) - common->start])
502    
503  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
504  #define MOV_UCHAR  SLJIT_MOV_UB  #define MOV_UCHAR  SLJIT_MOV_UB
505  #define MOVU_UCHAR SLJIT_MOVU_UB  #define MOVU_UCHAR SLJIT_MOVU_UB
506  #else  #elif defined COMPILE_PCRE16
 #ifdef COMPILE_PCRE16  
507  #define MOV_UCHAR  SLJIT_MOV_UH  #define MOV_UCHAR  SLJIT_MOV_UH
508  #define MOVU_UCHAR SLJIT_MOVU_UH  #define MOVU_UCHAR SLJIT_MOVU_UH
509    #elif defined COMPILE_PCRE32
510    #define MOV_UCHAR  SLJIT_MOV_UI
511    #define MOVU_UCHAR SLJIT_MOVU_UI
512  #else  #else
513  #error Unsupported compiling mode  #error Unsupported compiling mode
514  #endif  #endif
 #endif  
515    
516  /* Shortcuts. */  /* Shortcuts. */
517  #define DEFINE_COMPILER \  #define DEFINE_COMPILER \
# Line 449  the start pointers when the end of the c Line 528  the start pointers when the end of the c
528    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))
529  #define JUMPHERE(jump) \  #define JUMPHERE(jump) \
530    sljit_set_label((jump), sljit_emit_label(compiler))    sljit_set_label((jump), sljit_emit_label(compiler))
531    #define SET_LABEL(jump, label) \
532      sljit_set_label((jump), (label))
533  #define CMP(type, src1, src1w, src2, src2w) \  #define CMP(type, src1, src1w, src2, src2w) \
534    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))
535  #define CMPTO(type, src1, src1w, src2, src2w, label) \  #define CMPTO(type, src1, src1w, src2, src2w, label) \
536    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))
537  #define COND_VALUE(op, dst, dstw, type) \  #define OP_FLAGS(op, dst, dstw, src, srcw, type) \
538    sljit_emit_cond_value(compiler, (op), (dst), (dstw), (type))    sljit_emit_op_flags(compiler, (op), (dst), (dstw), (src), (srcw), (type))
539  #define GET_LOCAL_BASE(dst, dstw, offset) \  #define GET_LOCAL_BASE(dst, dstw, offset) \
540    sljit_get_local_base(compiler, (dst), (dstw), (offset))    sljit_get_local_base(compiler, (dst), (dstw), (offset))
541    
542  static pcre_uchar* bracketend(pcre_uchar* cc)  #define READ_CHAR_MAX 0x7fffffff
543    
544    static pcre_uchar *bracketend(pcre_uchar *cc)
545  {  {
546  SLJIT_ASSERT((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND));  SLJIT_ASSERT((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND));
547  do cc += GET(cc, 1); while (*cc == OP_ALT);  do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 467  cc += 1 + LINK_SIZE; Line 550  cc += 1 + LINK_SIZE;
550  return cc;  return cc;
551  }  }
552    
553    static int no_alternatives(pcre_uchar *cc)
554    {
555    int count = 0;
556    SLJIT_ASSERT((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND));
557    do
558      {
559      cc += GET(cc, 1);
560      count++;
561      }
562    while (*cc == OP_ALT);
563    SLJIT_ASSERT(*cc >= OP_KET && *cc <= OP_KETRPOS);
564    return count;
565    }
566    
567    static int ones_in_half_byte[16] = {
568      /* 0 */ 0, 1, 1, 2, /* 4 */ 1, 2, 2, 3,
569      /* 8 */ 1, 2, 2, 3, /* 12 */ 2, 3, 3, 4
570    };
571    
572  /* Functions whose might need modification for all new supported opcodes:  /* Functions whose might need modification for all new supported opcodes:
573   next_opcode   next_opcode
574   get_localspace   check_opcode_types
575   set_localptrs   set_private_data_ptrs
576   get_framesize   get_framesize
577   init_frame   init_frame
578   get_localsize   get_private_data_copy_length
579   copy_locals   copy_private_data
580   compile_trypath   compile_matchingpath
581   compile_backtrackpath   compile_backtrackingpath
582  */  */
583    
584  static pcre_uchar *next_opcode(compiler_common *common, pcre_uchar *cc)  static pcre_uchar *next_opcode(compiler_common *common, pcre_uchar *cc)
# Line 497  switch(*cc) Line 599  switch(*cc)
599    case OP_WORDCHAR:    case OP_WORDCHAR:
600    case OP_ANY:    case OP_ANY:
601    case OP_ALLANY:    case OP_ALLANY:
602      case OP_NOTPROP:
603      case OP_PROP:
604    case OP_ANYNL:    case OP_ANYNL:
605    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
606    case OP_HSPACE:    case OP_HSPACE:
# Line 509  switch(*cc) Line 613  switch(*cc)
613    case OP_CIRCM:    case OP_CIRCM:
614    case OP_DOLL:    case OP_DOLL:
615    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:  
616    case OP_CRSTAR:    case OP_CRSTAR:
617    case OP_CRMINSTAR:    case OP_CRMINSTAR:
618    case OP_CRPLUS:    case OP_CRPLUS:
619    case OP_CRMINPLUS:    case OP_CRMINPLUS:
620    case OP_CRQUERY:    case OP_CRQUERY:
621    case OP_CRMINQUERY:    case OP_CRMINQUERY:
622      case OP_CRRANGE:
623      case OP_CRMINRANGE:
624      case OP_CRPOSSTAR:
625      case OP_CRPOSPLUS:
626      case OP_CRPOSQUERY:
627      case OP_CRPOSRANGE:
628      case OP_CLASS:
629      case OP_NCLASS:
630      case OP_REF:
631      case OP_REFI:
632      case OP_DNREF:
633      case OP_DNREFI:
634      case OP_RECURSE:
635      case OP_CALLOUT:
636      case OP_ALT:
637      case OP_KET:
638      case OP_KETRMAX:
639      case OP_KETRMIN:
640      case OP_KETRPOS:
641      case OP_REVERSE:
642      case OP_ASSERT:
643      case OP_ASSERT_NOT:
644      case OP_ASSERTBACK:
645      case OP_ASSERTBACK_NOT:
646      case OP_ONCE:
647      case OP_ONCE_NC:
648      case OP_BRA:
649      case OP_BRAPOS:
650      case OP_CBRA:
651      case OP_CBRAPOS:
652      case OP_COND:
653      case OP_SBRA:
654      case OP_SBRAPOS:
655      case OP_SCBRA:
656      case OP_SCBRAPOS:
657      case OP_SCOND:
658      case OP_CREF:
659      case OP_DNCREF:
660      case OP_RREF:
661      case OP_DNRREF:
662    case OP_DEF:    case OP_DEF:
663    case OP_BRAZERO:    case OP_BRAZERO:
664    case OP_BRAMINZERO:    case OP_BRAMINZERO:
665    case OP_BRAPOSZERO:    case OP_BRAPOSZERO:
666      case OP_PRUNE:
667      case OP_SKIP:
668      case OP_THEN:
669    case OP_COMMIT:    case OP_COMMIT:
670    case OP_FAIL:    case OP_FAIL:
671    case OP_ACCEPT:    case OP_ACCEPT:
672    case OP_ASSERT_ACCEPT:    case OP_ASSERT_ACCEPT:
673      case OP_CLOSE:
674    case OP_SKIPZERO:    case OP_SKIPZERO:
675    return cc + 1;    return cc + PRIV(OP_lengths)[*cc];
   
   case OP_ANYBYTE:  
 #ifdef SUPPORT_UTF  
   if (common->utf) return NULL;  
 #endif  
   return cc + 1;  
676    
677    case OP_CHAR:    case OP_CHAR:
678    case OP_CHARI:    case OP_CHARI:
# Line 551  switch(*cc) Line 684  switch(*cc)
684    case OP_MINPLUS:    case OP_MINPLUS:
685    case OP_QUERY:    case OP_QUERY:
686    case OP_MINQUERY:    case OP_MINQUERY:
687      case OP_UPTO:
688      case OP_MINUPTO:
689      case OP_EXACT:
690    case OP_POSSTAR:    case OP_POSSTAR:
691    case OP_POSPLUS:    case OP_POSPLUS:
692    case OP_POSQUERY:    case OP_POSQUERY:
693      case OP_POSUPTO:
694    case OP_STARI:    case OP_STARI:
695    case OP_MINSTARI:    case OP_MINSTARI:
696    case OP_PLUSI:    case OP_PLUSI:
697    case OP_MINPLUSI:    case OP_MINPLUSI:
698    case OP_QUERYI:    case OP_QUERYI:
699    case OP_MINQUERYI:    case OP_MINQUERYI:
700      case OP_UPTOI:
701      case OP_MINUPTOI:
702      case OP_EXACTI:
703    case OP_POSSTARI:    case OP_POSSTARI:
704    case OP_POSPLUSI:    case OP_POSPLUSI:
705    case OP_POSQUERYI:    case OP_POSQUERYI:
706      case OP_POSUPTOI:
707    case OP_NOTSTAR:    case OP_NOTSTAR:
708    case OP_NOTMINSTAR:    case OP_NOTMINSTAR:
709    case OP_NOTPLUS:    case OP_NOTPLUS:
710    case OP_NOTMINPLUS:    case OP_NOTMINPLUS:
711    case OP_NOTQUERY:    case OP_NOTQUERY:
712    case OP_NOTMINQUERY:    case OP_NOTMINQUERY:
713      case OP_NOTUPTO:
714      case OP_NOTMINUPTO:
715      case OP_NOTEXACT:
716    case OP_NOTPOSSTAR:    case OP_NOTPOSSTAR:
717    case OP_NOTPOSPLUS:    case OP_NOTPOSPLUS:
718    case OP_NOTPOSQUERY:    case OP_NOTPOSQUERY:
719      case OP_NOTPOSUPTO:
720    case OP_NOTSTARI:    case OP_NOTSTARI:
721    case OP_NOTMINSTARI:    case OP_NOTMINSTARI:
722    case OP_NOTPLUSI:    case OP_NOTPLUSI:
723    case OP_NOTMINPLUSI:    case OP_NOTMINPLUSI:
724    case OP_NOTQUERYI:    case OP_NOTQUERYI:
725    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:  
726    case OP_NOTUPTOI:    case OP_NOTUPTOI:
727    case OP_NOTMINUPTOI:    case OP_NOTMINUPTOI:
728    case OP_NOTEXACTI:    case OP_NOTEXACTI:
729      case OP_NOTPOSSTARI:
730      case OP_NOTPOSPLUSI:
731      case OP_NOTPOSQUERYI:
732    case OP_NOTPOSUPTOI:    case OP_NOTPOSUPTOI:
733    cc += 2 + IMM2_SIZE;    cc += PRIV(OP_lengths)[*cc];
734  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
735    if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);    if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
736  #endif  #endif
737    return cc;    return cc;
738    
739    case OP_NOTPROP:    /* Special cases. */
740    case OP_PROP:    case OP_TYPESTAR:
741    return cc + 1 + 2;    case OP_TYPEMINSTAR:
742      case OP_TYPEPLUS:
743      case OP_TYPEMINPLUS:
744      case OP_TYPEQUERY:
745      case OP_TYPEMINQUERY:
746    case OP_TYPEUPTO:    case OP_TYPEUPTO:
747    case OP_TYPEMINUPTO:    case OP_TYPEMINUPTO:
748    case OP_TYPEEXACT:    case OP_TYPEEXACT:
749      case OP_TYPEPOSSTAR:
750      case OP_TYPEPOSPLUS:
751      case OP_TYPEPOSQUERY:
752    case OP_TYPEPOSUPTO:    case OP_TYPEPOSUPTO:
753    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;  
754    
755    case OP_CRRANGE:    case OP_ANYBYTE:
756    case OP_CRMINRANGE:  #ifdef SUPPORT_UTF
757    return cc + 1 + 2 * IMM2_SIZE;    if (common->utf) return NULL;
758    #endif
759    case OP_CLASS:    return cc + 1;
   case OP_NCLASS:  
   return cc + 1 + 32 / sizeof(pcre_uchar);  
760    
761  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
762    case OP_XCLASS:    case OP_XCLASS:
763    return cc + GET(cc, 1);    return cc + GET(cc, 1);
764  #endif  #endif
765    
   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;  
   
766    case OP_MARK:    case OP_MARK:
767      case OP_PRUNE_ARG:
768      case OP_SKIP_ARG:
769      case OP_THEN_ARG:
770    return cc + 1 + 2 + cc[1];    return cc + 1 + 2 + cc[1];
771    
772    default:    default:
773      /* All opcodes are supported now! */
774      SLJIT_ASSERT_STOP();
775    return NULL;    return NULL;
776    }    }
777  }  }
778    
779  #define CASE_ITERATOR_LOCAL1 \  static BOOL check_opcode_types(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend)
780      case OP_MINSTAR: \  {
781      case OP_MINPLUS: \  int count;
782      case OP_QUERY: \  pcre_uchar *slot;
     case OP_MINQUERY: \  
     case OP_MINSTARI: \  
     case OP_MINPLUSI: \  
     case OP_QUERYI: \  
     case OP_MINQUERYI: \  
     case OP_NOTMINSTAR: \  
     case OP_NOTMINPLUS: \  
     case OP_NOTQUERY: \  
     case OP_NOTMINQUERY: \  
     case OP_NOTMINSTARI: \  
     case OP_NOTMINPLUSI: \  
     case OP_NOTQUERYI: \  
     case OP_NOTMINQUERYI:  
783    
784  #define CASE_ITERATOR_LOCAL2A \  /* Calculate important variables (like stack size) and checks whether all opcodes are supported. */
785      case OP_STAR: \  while (cc < ccend)
786      case OP_PLUS: \    {
787      case OP_STARI: \    switch(*cc)
788      case OP_PLUSI: \      {
789      case OP_NOTSTAR: \      case OP_SET_SOM:
790      case OP_NOTPLUS: \      common->has_set_som = TRUE;
791      case OP_NOTSTARI: \      common->might_be_empty = TRUE;
792      case OP_NOTPLUSI:      cc += 1;
793        break;
794    
795  #define CASE_ITERATOR_LOCAL2B \      case OP_REF:
796      case OP_UPTO: \      case OP_REFI:
797      case OP_MINUPTO: \      common->optimized_cbracket[GET2(cc, 1)] = 0;
798      case OP_UPTOI: \      cc += 1 + IMM2_SIZE;
799      case OP_MINUPTOI: \      break;
     case OP_NOTUPTO: \  
     case OP_NOTMINUPTO: \  
     case OP_NOTUPTOI: \  
     case OP_NOTMINUPTOI:  
800    
801  #define CASE_ITERATOR_TYPE_LOCAL1 \      case OP_CBRAPOS:
802      case OP_TYPEMINSTAR: \      case OP_SCBRAPOS:
803      case OP_TYPEMINPLUS: \      common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] = 0;
804      case OP_TYPEQUERY: \      cc += 1 + LINK_SIZE + IMM2_SIZE;
805      case OP_TYPEMINQUERY:      break;
806    
807  #define CASE_ITERATOR_TYPE_LOCAL2A \      case OP_COND:
808      case OP_TYPESTAR: \      case OP_SCOND:
809      case OP_TYPEPLUS:      /* Only AUTO_CALLOUT can insert this opcode. We do
810           not intend to support this case. */
811        if (cc[1 + LINK_SIZE] == OP_CALLOUT)
812          return FALSE;
813        cc += 1 + LINK_SIZE;
814        break;
815    
816  #define CASE_ITERATOR_TYPE_LOCAL2B \      case OP_CREF:
817      case OP_TYPEUPTO: \      common->optimized_cbracket[GET2(cc, 1)] = 0;
818      case OP_TYPEMINUPTO:      cc += 1 + IMM2_SIZE;
819        break;
820    
821  static int get_class_iterator_size(pcre_uchar *cc)      case OP_DNREF:
822  {      case OP_DNREFI:
823  switch(*cc)      case OP_DNCREF:
824    {      count = GET2(cc, 1 + IMM2_SIZE);
825    case OP_CRSTAR:      slot = common->name_table + GET2(cc, 1) * common->name_entry_size;
826        while (count-- > 0)
827          {
828          common->optimized_cbracket[GET2(slot, 0)] = 0;
829          slot += common->name_entry_size;
830          }
831        cc += 1 + 2 * IMM2_SIZE;
832        break;
833    
834        case OP_RECURSE:
835        /* Set its value only once. */
836        if (common->recursive_head_ptr == 0)
837          {
838          common->recursive_head_ptr = common->ovector_start;
839          common->ovector_start += sizeof(sljit_sw);
840          }
841        cc += 1 + LINK_SIZE;
842        break;
843    
844        case OP_CALLOUT:
845        if (common->capture_last_ptr == 0)
846          {
847          common->capture_last_ptr = common->ovector_start;
848          common->ovector_start += sizeof(sljit_sw);
849          }
850        cc += 2 + 2 * LINK_SIZE;
851        break;
852    
853        case OP_THEN_ARG:
854        common->has_then = TRUE;
855        common->control_head_ptr = 1;
856        /* Fall through. */
857    
858        case OP_PRUNE_ARG:
859        common->needs_start_ptr = TRUE;
860        /* Fall through. */
861    
862        case OP_MARK:
863        if (common->mark_ptr == 0)
864          {
865          common->mark_ptr = common->ovector_start;
866          common->ovector_start += sizeof(sljit_sw);
867          }
868        cc += 1 + 2 + cc[1];
869        break;
870    
871        case OP_THEN:
872        common->has_then = TRUE;
873        common->control_head_ptr = 1;
874        /* Fall through. */
875    
876        case OP_PRUNE:
877        case OP_SKIP:
878        common->needs_start_ptr = TRUE;
879        cc += 1;
880        break;
881    
882        case OP_SKIP_ARG:
883        common->control_head_ptr = 1;
884        common->has_skip_arg = TRUE;
885        cc += 1 + 2 + cc[1];
886        break;
887    
888        default:
889        cc = next_opcode(common, cc);
890        if (cc == NULL)
891          return FALSE;
892        break;
893        }
894      }
895    return TRUE;
896    }
897    
898    static int get_class_iterator_size(pcre_uchar *cc)
899    {
900    switch(*cc)
901      {
902      case OP_CRSTAR:
903    case OP_CRPLUS:    case OP_CRPLUS:
904    return 2;    return 2;
905    
# Line 752  switch(*cc) Line 920  switch(*cc)
920    }    }
921  }  }
922    
923  static int get_localspace(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend)  static BOOL detect_repeat(compiler_common *common, pcre_uchar *begin)
924  {  {
925  int localspace = 0;  pcre_uchar *end = bracketend(begin);
926  pcre_uchar *alternative;  pcre_uchar *next;
927  pcre_uchar *end = NULL;  pcre_uchar *next_end;
928  int space, size, bracketlen;  pcre_uchar *max_end;
929    pcre_uchar type;
930  /* Calculate important variables (like stack size) and checks whether all opcodes are supported. */  sljit_sw length = end - begin;
931  while (cc < ccend)  int min, max, i;
   {  
   space = 0;  
   size = 0;  
   bracketlen = 0;  
   switch(*cc)  
     {  
     case OP_SET_SOM:  
     common->has_set_som = TRUE;  
     cc += 1;  
     break;  
   
     case OP_ASSERT:  
     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:  
     localspace += sizeof(sljit_w);  
     bracketlen = 1 + LINK_SIZE;  
     break;  
   
     case OP_CBRAPOS:  
     case OP_SCBRAPOS:  
     localspace += sizeof(sljit_w);  
     bracketlen = 1 + LINK_SIZE + IMM2_SIZE;  
     break;  
   
     case OP_COND:  
     /* Might be a hidden SCOND. */  
     alternative = cc + GET(cc, 1);  
     if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)  
       localspace += sizeof(sljit_w);  
     bracketlen = 1 + LINK_SIZE;  
     break;  
932    
933      case OP_BRA:  /* Detect fixed iterations first. */
934      bracketlen = 1 + LINK_SIZE;  if (end[-(1 + LINK_SIZE)] != OP_KET)
935      break;    return FALSE;
936    
937      case OP_CBRA:  /* Already detected repeat. */
938      case OP_SCBRA:  if (common->private_data_ptrs[end - common->start - LINK_SIZE] != 0)
939      bracketlen = 1 + LINK_SIZE + IMM2_SIZE;    return TRUE;
     break;  
940    
941      CASE_ITERATOR_LOCAL1  next = end;
942      space = 1;  min = 1;
943      size = -2;  while (1)
944      {
945      if (*next != *begin)
946      break;      break;
947      next_end = bracketend(next);
948      CASE_ITERATOR_LOCAL2A    if (next_end - next != length || memcmp(begin, next, IN_UCHARS(length)) != 0)
     space = 2;  
     size = -2;  
949      break;      break;
950      next = next_end;
951      min++;
952      }
953    
954      CASE_ITERATOR_LOCAL2B  if (min == 2)
955      space = 2;    return FALSE;
     size = -(2 + IMM2_SIZE);  
     break;  
956    
957      CASE_ITERATOR_TYPE_LOCAL1  max = 0;
958      space = 1;  max_end = next;
959      size = 1;  if (*next == OP_BRAZERO || *next == OP_BRAMINZERO)
960      break;    {
961      type = *next;
962      while (1)
963        {
964        if (next[0] != type || next[1] != OP_BRA || next[2 + LINK_SIZE] != *begin)
965          break;
966        next_end = bracketend(next + 2 + LINK_SIZE);
967        if (next_end - next != (length + 2 + LINK_SIZE) || memcmp(begin, next + 2 + LINK_SIZE, IN_UCHARS(length)) != 0)
968          break;
969        next = next_end;
970        max++;
971        }
972    
973      CASE_ITERATOR_TYPE_LOCAL2A    if (next[0] == type && next[1] == *begin && max >= 1)
974      if (cc[1] != OP_ANYNL && cc[1] != OP_EXTUNI)      {
975        space = 2;      next_end = bracketend(next + 1);
976      size = 1;      if (next_end - next == (length + 1) && memcmp(begin, next + 1, IN_UCHARS(length)) == 0)
977      break;        {
978          for (i = 0; i < max; i++, next_end += 1 + LINK_SIZE)
979            if (*next_end != OP_KET)
980              break;
981    
982      CASE_ITERATOR_TYPE_LOCAL2B        if (i == max)
983      if (cc[1 + IMM2_SIZE] != OP_ANYNL && cc[1 + IMM2_SIZE] != OP_EXTUNI)          {
984        space = 2;          common->private_data_ptrs[max_end - common->start - LINK_SIZE] = next_end - max_end;
985      size = 1 + IMM2_SIZE;          common->private_data_ptrs[max_end - common->start - LINK_SIZE + 1] = (type == OP_BRAZERO) ? OP_UPTO : OP_MINUPTO;
986      break;          /* +2 the original and the last. */
987            common->private_data_ptrs[max_end - common->start - LINK_SIZE + 2] = max + 2;
988            if (min == 1)
989              return TRUE;
990            min--;
991            max_end -= (1 + LINK_SIZE) + GET(max_end, -LINK_SIZE);
992            }
993          }
994        }
995      }
996    
997      case OP_CLASS:  if (min >= 3)
998      case OP_NCLASS:    {
999      size += 1 + 32 / sizeof(pcre_uchar);    common->private_data_ptrs[end - common->start - LINK_SIZE] = max_end - end;
1000      space = get_class_iterator_size(cc + size);    common->private_data_ptrs[end - common->start - LINK_SIZE + 1] = OP_EXACT;
1001      break;    common->private_data_ptrs[end - common->start - LINK_SIZE + 2] = min;
1002      return TRUE;
1003      }
1004    
1005  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  return FALSE;
1006      case OP_XCLASS:  }
     size = GET(cc, 1);  
     space = get_class_iterator_size(cc + size);  
     break;  
 #endif  
1007    
1008      case OP_RECURSE:  #define CASE_ITERATOR_PRIVATE_DATA_1 \
1009      /* Set its value only once. */      case OP_MINSTAR: \
1010      if (common->recursive_head == 0)      case OP_MINPLUS: \
1011        {      case OP_QUERY: \
1012        common->recursive_head = common->ovector_start;      case OP_MINQUERY: \
1013        common->ovector_start += sizeof(sljit_w);      case OP_MINSTARI: \
1014        }      case OP_MINPLUSI: \
1015      cc += 1 + LINK_SIZE;      case OP_QUERYI: \
1016      break;      case OP_MINQUERYI: \
1017        case OP_NOTMINSTAR: \
1018        case OP_NOTMINPLUS: \
1019        case OP_NOTQUERY: \
1020        case OP_NOTMINQUERY: \
1021        case OP_NOTMINSTARI: \
1022        case OP_NOTMINPLUSI: \
1023        case OP_NOTQUERYI: \
1024        case OP_NOTMINQUERYI:
1025    
1026      case OP_MARK:  #define CASE_ITERATOR_PRIVATE_DATA_2A \
1027      if (common->mark_ptr == 0)      case OP_STAR: \
1028        {      case OP_PLUS: \
1029        common->mark_ptr = common->ovector_start;      case OP_STARI: \
1030        common->ovector_start += sizeof(sljit_w);      case OP_PLUSI: \
1031        }      case OP_NOTSTAR: \
1032      cc += 1 + 2 + cc[1];      case OP_NOTPLUS: \
1033      break;      case OP_NOTSTARI: \
1034        case OP_NOTPLUSI:
1035    
1036      default:  #define CASE_ITERATOR_PRIVATE_DATA_2B \
1037      cc = next_opcode(common, cc);      case OP_UPTO: \
1038      if (cc == NULL)      case OP_MINUPTO: \
1039        return -1;      case OP_UPTOI: \
1040      break;      case OP_MINUPTOI: \
1041      }      case OP_NOTUPTO: \
1042        case OP_NOTMINUPTO: \
1043        case OP_NOTUPTOI: \
1044        case OP_NOTMINUPTOI:
1045    
1046    if (space > 0 && cc >= end)  #define CASE_ITERATOR_TYPE_PRIVATE_DATA_1 \
1047      localspace += sizeof(sljit_w) * space;      case OP_TYPEMINSTAR: \
1048        case OP_TYPEMINPLUS: \
1049        case OP_TYPEQUERY: \
1050        case OP_TYPEMINQUERY:
1051    
1052    if (size != 0)  #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2A \
1053      {      case OP_TYPESTAR: \
1054      if (size < 0)      case OP_TYPEPLUS:
       {  
       cc += -size;  
 #ifdef SUPPORT_UTF  
       if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);  
 #endif  
       }  
     else  
       cc += size;  
     }  
1055    
1056    if (bracketlen > 0)  #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2B \
1057      {      case OP_TYPEUPTO: \
1058      if (cc >= end)      case OP_TYPEMINUPTO:
       {  
       end = bracketend(cc);  
       if (end[-1 - LINK_SIZE] == OP_KET)  
         end = NULL;  
       }  
     cc += bracketlen;  
     }  
   }  
 return localspace;  
 }  
1059    
1060  static void set_localptrs(compiler_common *common, int localptr, pcre_uchar *ccend)  static void set_private_data_ptrs(compiler_common *common, int *private_data_start, pcre_uchar *ccend)
1061  {  {
1062  pcre_uchar *cc = common->start;  pcre_uchar *cc = common->start;
1063  pcre_uchar *alternative;  pcre_uchar *alternative;
1064  pcre_uchar *end = NULL;  pcre_uchar *end = NULL;
1065    int private_data_ptr = *private_data_start;
1066  int space, size, bracketlen;  int space, size, bracketlen;
1067    
1068  while (cc < ccend)  while (cc < ccend)
# Line 922  while (cc < ccend) Line 1070  while (cc < ccend)
1070    space = 0;    space = 0;
1071    size = 0;    size = 0;
1072    bracketlen = 0;    bracketlen = 0;
1073      if (private_data_ptr > SLJIT_MAX_LOCAL_SIZE)
1074        return;
1075    
1076      if (*cc == OP_ONCE || *cc == OP_ONCE_NC || *cc == OP_BRA || *cc == OP_CBRA || *cc == OP_COND)
1077        if (detect_repeat(common, cc))
1078          {
1079          /* These brackets are converted to repeats, so no global
1080          based single character repeat is allowed. */
1081          if (cc >= end)
1082            end = bracketend(cc);
1083          }
1084    
1085    switch(*cc)    switch(*cc)
1086      {      {
1087        case OP_KET:
1088        if (common->private_data_ptrs[cc + 1 - common->start] != 0)
1089          {
1090          common->private_data_ptrs[cc - common->start] = private_data_ptr;
1091          private_data_ptr += sizeof(sljit_sw);
1092          cc += common->private_data_ptrs[cc + 1 - common->start];
1093          }
1094        cc += 1 + LINK_SIZE;
1095        break;
1096    
1097      case OP_ASSERT:      case OP_ASSERT:
1098      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1099      case OP_ASSERTBACK:      case OP_ASSERTBACK:
# Line 934  while (cc < ccend) Line 1104  while (cc < ccend)
1104      case OP_SBRA:      case OP_SBRA:
1105      case OP_SBRAPOS:      case OP_SBRAPOS:
1106      case OP_SCOND:      case OP_SCOND:
1107      common->localptrs[cc - common->start] = localptr;      common->private_data_ptrs[cc - common->start] = private_data_ptr;
1108      localptr += sizeof(sljit_w);      private_data_ptr += sizeof(sljit_sw);
1109      bracketlen = 1 + LINK_SIZE;      bracketlen = 1 + LINK_SIZE;
1110      break;      break;
1111    
1112      case OP_CBRAPOS:      case OP_CBRAPOS:
1113      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1114      common->localptrs[cc - common->start] = localptr;      common->private_data_ptrs[cc - common->start] = private_data_ptr;
1115      localptr += sizeof(sljit_w);      private_data_ptr += sizeof(sljit_sw);
1116      bracketlen = 1 + LINK_SIZE + IMM2_SIZE;      bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
1117      break;      break;
1118    
# Line 951  while (cc < ccend) Line 1121  while (cc < ccend)
1121      alternative = cc + GET(cc, 1);      alternative = cc + GET(cc, 1);
1122      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1123        {        {
1124        common->localptrs[cc - common->start] = localptr;        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1125        localptr += sizeof(sljit_w);        private_data_ptr += sizeof(sljit_sw);
1126        }        }
1127      bracketlen = 1 + LINK_SIZE;      bracketlen = 1 + LINK_SIZE;
1128      break;      break;
# Line 966  while (cc < ccend) Line 1136  while (cc < ccend)
1136      bracketlen = 1 + LINK_SIZE + IMM2_SIZE;      bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
1137      break;      break;
1138    
1139      CASE_ITERATOR_LOCAL1      CASE_ITERATOR_PRIVATE_DATA_1
1140      space = 1;      space = 1;
1141      size = -2;      size = -2;
1142      break;      break;
1143    
1144      CASE_ITERATOR_LOCAL2A      CASE_ITERATOR_PRIVATE_DATA_2A
1145      space = 2;      space = 2;
1146      size = -2;      size = -2;
1147      break;      break;
1148    
1149      CASE_ITERATOR_LOCAL2B      CASE_ITERATOR_PRIVATE_DATA_2B
1150      space = 2;      space = 2;
1151      size = -(2 + IMM2_SIZE);      size = -(2 + IMM2_SIZE);
1152      break;      break;
1153    
1154      CASE_ITERATOR_TYPE_LOCAL1      CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1155      space = 1;      space = 1;
1156      size = 1;      size = 1;
1157      break;      break;
1158    
1159      CASE_ITERATOR_TYPE_LOCAL2A      CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1160      if (cc[1] != OP_ANYNL && cc[1] != OP_EXTUNI)      if (cc[1] != OP_ANYNL && cc[1] != OP_EXTUNI)
1161        space = 2;        space = 2;
1162      size = 1;      size = 1;
1163      break;      break;
1164    
1165      CASE_ITERATOR_TYPE_LOCAL2B      CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1166      if (cc[1 + IMM2_SIZE] != OP_ANYNL && cc[1 + IMM2_SIZE] != OP_EXTUNI)      if (cc[1 + IMM2_SIZE] != OP_ANYNL && cc[1 + IMM2_SIZE] != OP_EXTUNI)
1167        space = 2;        space = 2;
1168      size = 1 + IMM2_SIZE;      size = 1 + IMM2_SIZE;
# Line 1017  while (cc < ccend) Line 1187  while (cc < ccend)
1187      break;      break;
1188      }      }
1189    
1190      /* Character iterators, which are not inside a repeated bracket,
1191         gets a private slot instead of allocating it on the stack. */
1192    if (space > 0 && cc >= end)    if (space > 0 && cc >= end)
1193      {      {
1194      common->localptrs[cc - common->start] = localptr;      common->private_data_ptrs[cc - common->start] = private_data_ptr;
1195      localptr += sizeof(sljit_w) * space;      private_data_ptr += sizeof(sljit_sw) * space;
1196      }      }
1197    
1198    if (size != 0)    if (size != 0)
# Line 1047  while (cc < ccend) Line 1219  while (cc < ccend)
1219      cc += bracketlen;      cc += bracketlen;
1220      }      }
1221    }    }
1222    *private_data_start = private_data_ptr;
1223  }  }
1224    
1225  /* Returns with -1 if no need for frame. */  /* Returns with a frame_types (always < 0) if no need for frame. */
1226  static int get_framesize(compiler_common *common, pcre_uchar *cc, BOOL recursive)  static int get_framesize(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, BOOL recursive, BOOL *needs_control_head)
1227  {  {
 pcre_uchar *ccend = bracketend(cc);  
1228  int length = 0;  int length = 0;
1229  BOOL possessive = FALSE;  int possessive = 0;
1230    BOOL stack_restore = FALSE;
1231  BOOL setsom_found = recursive;  BOOL setsom_found = recursive;
1232  BOOL setmark_found = recursive;  BOOL setmark_found = recursive;
1233    /* The last capture is a local variable even for recursions. */
1234    BOOL capture_last_found = FALSE;
1235    
1236    #if defined DEBUG_FORCE_CONTROL_HEAD && DEBUG_FORCE_CONTROL_HEAD
1237    SLJIT_ASSERT(common->control_head_ptr != 0);
1238    *needs_control_head = TRUE;
1239    #else
1240    *needs_control_head = FALSE;
1241    #endif
1242    
1243  if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))  if (ccend == NULL)
1244    {    {
1245    length = 3;    ccend = bracketend(cc) - (1 + LINK_SIZE);
1246    possessive = TRUE;    if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))
1247        {
1248        possessive = length = (common->capture_last_ptr != 0) ? 5 : 3;
1249        /* This is correct regardless of common->capture_last_ptr. */
1250        capture_last_found = TRUE;
1251        }
1252      cc = next_opcode(common, cc);
1253    }    }
1254    
 cc = next_opcode(common, cc);  
1255  SLJIT_ASSERT(cc != NULL);  SLJIT_ASSERT(cc != NULL);
1256  while (cc < ccend)  while (cc < ccend)
1257    switch(*cc)    switch(*cc)
1258      {      {
1259      case OP_SET_SOM:      case OP_SET_SOM:
1260      SLJIT_ASSERT(common->has_set_som);      SLJIT_ASSERT(common->has_set_som);
1261        stack_restore = TRUE;
1262      if (!setsom_found)      if (!setsom_found)
1263        {        {
1264        length += 2;        length += 2;
# Line 1080  while (cc < ccend) Line 1268  while (cc < ccend)
1268      break;      break;
1269    
1270      case OP_MARK:      case OP_MARK:
1271        case OP_PRUNE_ARG:
1272        case OP_THEN_ARG:
1273      SLJIT_ASSERT(common->mark_ptr != 0);      SLJIT_ASSERT(common->mark_ptr != 0);
1274        stack_restore = TRUE;
1275      if (!setmark_found)      if (!setmark_found)
1276        {        {
1277        length += 2;        length += 2;
1278        setmark_found = TRUE;        setmark_found = TRUE;
1279        }        }
1280        if (common->control_head_ptr != 0)
1281          *needs_control_head = TRUE;
1282      cc += 1 + 2 + cc[1];      cc += 1 + 2 + cc[1];
1283      break;      break;
1284    
1285      case OP_RECURSE:      case OP_RECURSE:
1286        stack_restore = TRUE;
1287      if (common->has_set_som && !setsom_found)      if (common->has_set_som && !setsom_found)
1288        {        {
1289        length += 2;        length += 2;
# Line 1100  while (cc < ccend) Line 1294  while (cc < ccend)
1294        length += 2;        length += 2;
1295        setmark_found = TRUE;        setmark_found = TRUE;
1296        }        }
1297        if (common->capture_last_ptr != 0 && !capture_last_found)
1298          {
1299          length += 2;
1300          capture_last_found = TRUE;
1301          }
1302      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1303      break;      break;
1304    
1305      case OP_CBRA:      case OP_CBRA:
1306      case OP_CBRAPOS:      case OP_CBRAPOS:
1307      case OP_SCBRA:      case OP_SCBRA:
1308      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1309      length += 3;      stack_restore = TRUE;
1310      cc += 1 + LINK_SIZE + IMM2_SIZE;      if (common->capture_last_ptr != 0 && !capture_last_found)
1311      break;        {
1312          length += 2;
1313          capture_last_found = TRUE;
1314          }
1315        length += 3;
1316        cc += 1 + LINK_SIZE + IMM2_SIZE;
1317        break;
1318    
1319        default:
1320        stack_restore = TRUE;
1321        /* Fall through. */
1322    
1323        case OP_NOT_WORD_BOUNDARY:
1324        case OP_WORD_BOUNDARY:
1325        case OP_NOT_DIGIT:
1326        case OP_DIGIT:
1327        case OP_NOT_WHITESPACE:
1328        case OP_WHITESPACE:
1329        case OP_NOT_WORDCHAR:
1330        case OP_WORDCHAR:
1331        case OP_ANY:
1332        case OP_ALLANY:
1333        case OP_ANYBYTE:
1334        case OP_NOTPROP:
1335        case OP_PROP:
1336        case OP_ANYNL:
1337        case OP_NOT_HSPACE:
1338        case OP_HSPACE:
1339        case OP_NOT_VSPACE:
1340        case OP_VSPACE:
1341        case OP_EXTUNI:
1342        case OP_EODN:
1343        case OP_EOD:
1344        case OP_CIRC:
1345        case OP_CIRCM:
1346        case OP_DOLL:
1347        case OP_DOLLM:
1348        case OP_CHAR:
1349        case OP_CHARI:
1350        case OP_NOT:
1351        case OP_NOTI:
1352    
1353        case OP_EXACT:
1354        case OP_POSSTAR:
1355        case OP_POSPLUS:
1356        case OP_POSQUERY:
1357        case OP_POSUPTO:
1358    
1359        case OP_EXACTI:
1360        case OP_POSSTARI:
1361        case OP_POSPLUSI:
1362        case OP_POSQUERYI:
1363        case OP_POSUPTOI:
1364    
1365        case OP_NOTEXACT:
1366        case OP_NOTPOSSTAR:
1367        case OP_NOTPOSPLUS:
1368        case OP_NOTPOSQUERY:
1369        case OP_NOTPOSUPTO:
1370    
1371        case OP_NOTEXACTI:
1372        case OP_NOTPOSSTARI:
1373        case OP_NOTPOSPLUSI:
1374        case OP_NOTPOSQUERYI:
1375        case OP_NOTPOSUPTOI:
1376    
1377        case OP_TYPEEXACT:
1378        case OP_TYPEPOSSTAR:
1379        case OP_TYPEPOSPLUS:
1380        case OP_TYPEPOSQUERY:
1381        case OP_TYPEPOSUPTO:
1382    
1383        case OP_CLASS:
1384        case OP_NCLASS:
1385        case OP_XCLASS:
1386    
     default:  
1387      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
1388      SLJIT_ASSERT(cc != NULL);      SLJIT_ASSERT(cc != NULL);
1389      break;      break;
1390      }      }
1391    
1392  /* Possessive quantifiers can use a special case. */  /* Possessive quantifiers can use a special case. */
1393  if (SLJIT_UNLIKELY(possessive) && length == 3)  if (SLJIT_UNLIKELY(possessive == length))
1394    return -1;    return stack_restore ? no_frame : no_stack;
1395    
1396  if (length > 0)  if (length > 0)
1397    return length + 1;    return length + 1;
1398  return -1;  return stack_restore ? no_frame : no_stack;
1399  }  }
1400    
1401  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)
1402  {  {
1403  DEFINE_COMPILER;  DEFINE_COMPILER;
 pcre_uchar *ccend = bracketend(cc);  
1404  BOOL setsom_found = recursive;  BOOL setsom_found = recursive;
1405  BOOL setmark_found = recursive;  BOOL setmark_found = recursive;
1406    /* The last capture is a local variable even for recursions. */
1407    BOOL capture_last_found = FALSE;
1408  int offset;  int offset;
1409    
1410  /* >= 1 + shortest item size (2) */  /* >= 1 + shortest item size (2) */
# Line 1139  SLJIT_UNUSED_ARG(stacktop); Line 1412  SLJIT_UNUSED_ARG(stacktop);
1412  SLJIT_ASSERT(stackpos >= stacktop + 2);  SLJIT_ASSERT(stackpos >= stacktop + 2);
1413    
1414  stackpos = STACK(stackpos);  stackpos = STACK(stackpos);
1415  if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))  if (ccend == NULL)
1416    cc = next_opcode(common, cc);    {
1417      ccend = bracketend(cc) - (1 + LINK_SIZE);
1418      if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))
1419        cc = next_opcode(common, cc);
1420      }
1421    
1422  SLJIT_ASSERT(cc != NULL);  SLJIT_ASSERT(cc != NULL);
1423  while (cc < ccend)  while (cc < ccend)
1424    switch(*cc)    switch(*cc)
# Line 1149  while (cc < ccend) Line 1427  while (cc < ccend)
1427      SLJIT_ASSERT(common->has_set_som);      SLJIT_ASSERT(common->has_set_som);
1428      if (!setsom_found)      if (!setsom_found)
1429        {        {
1430        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(0));
1431        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setstrbegin);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1432        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1433        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1434        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1435        setsom_found = TRUE;        setsom_found = TRUE;
1436        }        }
1437      cc += 1;      cc += 1;
1438      break;      break;
1439    
1440      case OP_MARK:      case OP_MARK:
1441        case OP_PRUNE_ARG:
1442        case OP_THEN_ARG:
1443      SLJIT_ASSERT(common->mark_ptr != 0);      SLJIT_ASSERT(common->mark_ptr != 0);
1444      if (!setmark_found)      if (!setmark_found)
1445        {        {
1446        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->mark_ptr);
1447        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);
1448        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1449        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1450        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1451        setmark_found = TRUE;        setmark_found = TRUE;
1452        }        }
1453      cc += 1 + 2 + cc[1];      cc += 1 + 2 + cc[1];
# Line 1176  while (cc < ccend) Line 1456  while (cc < ccend)
1456      case OP_RECURSE:      case OP_RECURSE:
1457      if (common->has_set_som && !setsom_found)      if (common->has_set_som && !setsom_found)
1458        {        {
1459        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(0));
1460        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setstrbegin);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1461        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1462        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1463        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1464        setsom_found = TRUE;        setsom_found = TRUE;
1465        }        }
1466      if (common->mark_ptr != 0 && !setmark_found)      if (common->mark_ptr != 0 && !setmark_found)
1467        {        {
1468        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->mark_ptr);
1469        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);
1470        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
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        setmark_found = TRUE;        setmark_found = TRUE;
1474        }        }
1475        if (common->capture_last_ptr != 0 && !capture_last_found)
1476          {
1477          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->capture_last_ptr);
1478          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1479          stackpos += (int)sizeof(sljit_sw);
1480          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1481          stackpos += (int)sizeof(sljit_sw);
1482          capture_last_found = TRUE;
1483          }
1484      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1485      break;      break;
1486    
# Line 1199  while (cc < ccend) Line 1488  while (cc < ccend)
1488      case OP_CBRAPOS:      case OP_CBRAPOS:
1489      case OP_SCBRA:      case OP_SCBRA:
1490      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1491        if (common->capture_last_ptr != 0 && !capture_last_found)
1492          {
1493          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->capture_last_ptr);
1494          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1495          stackpos += (int)sizeof(sljit_sw);
1496          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1497          stackpos += (int)sizeof(sljit_sw);
1498          capture_last_found = TRUE;
1499          }
1500      offset = (GET2(cc, 1 + LINK_SIZE)) << 1;      offset = (GET2(cc, 1 + LINK_SIZE)) << 1;
1501      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));
1502      stackpos += (int)sizeof(sljit_w);      stackpos += (int)sizeof(sljit_sw);
1503      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset));      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(offset));
1504      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset + 1));      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(offset + 1));
1505      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1506      stackpos += (int)sizeof(sljit_w);      stackpos += (int)sizeof(sljit_sw);
1507      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);
1508      stackpos += (int)sizeof(sljit_w);      stackpos += (int)sizeof(sljit_sw);
1509    
1510      cc += 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1511      break;      break;
# Line 1218  while (cc < ccend) Line 1516  while (cc < ccend)
1516      break;      break;
1517      }      }
1518    
1519  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_end);  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, 0);
1520  SLJIT_ASSERT(stackpos == STACK(stacktop));  SLJIT_ASSERT(stackpos == STACK(stacktop));
1521  }  }
1522    
1523  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)
1524  {  {
1525  int localsize = 2;  int private_data_length = needs_control_head ? 3 : 2;
1526  int size;  int size;
1527  pcre_uchar *alternative;  pcre_uchar *alternative;
1528  /* Calculate the sum of the local variables. */  /* Calculate the sum of the private machine words. */
1529  while (cc < ccend)  while (cc < ccend)
1530    {    {
1531    size = 0;    size = 0;
1532    switch(*cc)    switch(*cc)
1533      {      {
1534        case OP_KET:
1535        if (PRIVATE_DATA(cc) != 0)
1536          {
1537          private_data_length++;
1538          SLJIT_ASSERT(PRIVATE_DATA(cc + 1) != 0);
1539          cc += PRIVATE_DATA(cc + 1);
1540          }
1541        cc += 1 + LINK_SIZE;
1542        break;
1543    
1544      case OP_ASSERT:      case OP_ASSERT:
1545      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1546      case OP_ASSERTBACK:      case OP_ASSERTBACK:
# Line 1243  while (cc < ccend) Line 1551  while (cc < ccend)
1551      case OP_SBRA:      case OP_SBRA:
1552      case OP_SBRAPOS:      case OP_SBRAPOS:
1553      case OP_SCOND:      case OP_SCOND:
1554      localsize++;      private_data_length++;
1555        SLJIT_ASSERT(PRIVATE_DATA(cc) != 0);
1556      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1557      break;      break;
1558    
1559      case OP_CBRA:      case OP_CBRA:
1560      case OP_SCBRA:      case OP_SCBRA:
1561      localsize++;      if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1562          private_data_length++;
1563      cc += 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1564      break;      break;
1565    
1566      case OP_CBRAPOS:      case OP_CBRAPOS:
1567      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1568      localsize += 2;      private_data_length += 2;
1569      cc += 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1570      break;      break;
1571    
# Line 1263  while (cc < ccend) Line 1573  while (cc < ccend)
1573      /* Might be a hidden SCOND. */      /* Might be a hidden SCOND. */
1574      alternative = cc + GET(cc, 1);      alternative = cc + GET(cc, 1);
1575      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1576        localsize++;        private_data_length++;
1577      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1578      break;      break;
1579    
1580      CASE_ITERATOR_LOCAL1      CASE_ITERATOR_PRIVATE_DATA_1
1581      if (PRIV_DATA(cc))      if (PRIVATE_DATA(cc))
1582        localsize++;        private_data_length++;
1583      cc += 2;      cc += 2;
1584  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
1585      if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1586  #endif  #endif
1587      break;      break;
1588    
1589      CASE_ITERATOR_LOCAL2A      CASE_ITERATOR_PRIVATE_DATA_2A
1590      if (PRIV_DATA(cc))      if (PRIVATE_DATA(cc))
1591        localsize += 2;        private_data_length += 2;
1592      cc += 2;      cc += 2;
1593  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
1594      if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1595  #endif  #endif
1596      break;      break;
1597    
1598      CASE_ITERATOR_LOCAL2B      CASE_ITERATOR_PRIVATE_DATA_2B
1599      if (PRIV_DATA(cc))      if (PRIVATE_DATA(cc))
1600        localsize += 2;        private_data_length += 2;
1601      cc += 2 + IMM2_SIZE;      cc += 2 + IMM2_SIZE;
1602  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
1603      if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1604  #endif  #endif
1605      break;      break;
1606    
1607      CASE_ITERATOR_TYPE_LOCAL1      CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1608      if (PRIV_DATA(cc))      if (PRIVATE_DATA(cc))
1609        localsize++;        private_data_length++;
1610      cc += 1;      cc += 1;
1611      break;      break;
1612    
1613      CASE_ITERATOR_TYPE_LOCAL2A      CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1614      if (PRIV_DATA(cc))      if (PRIVATE_DATA(cc))
1615        localsize += 2;        private_data_length += 2;
1616      cc += 1;      cc += 1;
1617      break;      break;
1618    
1619      CASE_ITERATOR_TYPE_LOCAL2B      CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1620      if (PRIV_DATA(cc))      if (PRIVATE_DATA(cc))
1621        localsize += 2;        private_data_length += 2;
1622      cc += 1 + IMM2_SIZE;      cc += 1 + IMM2_SIZE;
1623      break;      break;
1624    
# Line 1320  while (cc < ccend) Line 1630  while (cc < ccend)
1630  #else  #else
1631      size = 1 + 32 / (int)sizeof(pcre_uchar);      size = 1 + 32 / (int)sizeof(pcre_uchar);
1632  #endif  #endif
1633      if (PRIV_DATA(cc))      if (PRIVATE_DATA(cc))
1634        localsize += get_class_iterator_size(cc + size);        private_data_length += get_class_iterator_size(cc + size);
1635      cc += size;      cc += size;
1636      break;      break;
1637    
# Line 1332  while (cc < ccend) Line 1642  while (cc < ccend)
1642      }      }
1643    }    }
1644  SLJIT_ASSERT(cc == ccend);  SLJIT_ASSERT(cc == ccend);
1645  return localsize;  return private_data_length;
1646  }  }
1647    
1648  static void copy_locals(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend,  static void copy_private_data(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend,
1649    BOOL save, int stackptr, int stacktop)    BOOL save, int stackptr, int stacktop, BOOL needs_control_head)
1650  {  {
1651  DEFINE_COMPILER;  DEFINE_COMPILER;
1652  int srcw[2];  int srcw[2];
# Line 1357  stacktop = STACK(stacktop - 1); Line 1667  stacktop = STACK(stacktop - 1);
1667    
1668  if (!save)  if (!save)
1669    {    {
1670    stackptr += sizeof(sljit_w);    stackptr += (needs_control_head ? 2 : 1) * sizeof(sljit_sw);
1671    if (stackptr < stacktop)    if (stackptr < stacktop)
1672      {      {
1673      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1674      stackptr += sizeof(sljit_w);      stackptr += sizeof(sljit_sw);
1675      tmp1empty = FALSE;      tmp1empty = FALSE;
1676      }      }
1677    if (stackptr < stacktop)    if (stackptr < stacktop)
1678      {      {
1679      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1680      stackptr += sizeof(sljit_w);      stackptr += sizeof(sljit_sw);
1681      tmp2empty = FALSE;      tmp2empty = FALSE;
1682      }      }
1683    /* The tmp1next must be TRUE in either way. */    /* The tmp1next must be TRUE in either way. */
1684    }    }
1685    
1686  while (status != end)  do
1687    {    {
1688    count = 0;    count = 0;
1689    switch(status)    switch(status)
1690      {      {
1691      case start:      case start:
1692      SLJIT_ASSERT(save && common->recursive_head != 0);      SLJIT_ASSERT(save && common->recursive_head_ptr != 0);
1693      count = 1;      count = 1;
1694      srcw[0] = common->recursive_head;      srcw[0] = common->recursive_head_ptr;
1695        if (needs_control_head)
1696          {
1697          SLJIT_ASSERT(common->control_head_ptr != 0);
1698          count = 2;
1699          srcw[1] = common->control_head_ptr;
1700          }
1701      status = loop;      status = loop;
1702      break;      break;
1703    
# Line 1394  while (status != end) Line 1710  while (status != end)
1710    
1711      switch(*cc)      switch(*cc)
1712        {        {
1713          case OP_KET:
1714          if (PRIVATE_DATA(cc) != 0)
1715            {
1716            count = 1;
1717            srcw[0] = PRIVATE_DATA(cc);
1718            SLJIT_ASSERT(PRIVATE_DATA(cc + 1) != 0);
1719            cc += PRIVATE_DATA(cc + 1);
1720            }
1721          cc += 1 + LINK_SIZE;
1722          break;
1723    
1724        case OP_ASSERT:        case OP_ASSERT:
1725        case OP_ASSERT_NOT:        case OP_ASSERT_NOT:
1726        case OP_ASSERTBACK:        case OP_ASSERTBACK:
# Line 1405  while (status != end) Line 1732  while (status != end)
1732        case OP_SBRAPOS:        case OP_SBRAPOS:
1733        case OP_SCOND:        case OP_SCOND:
1734        count = 1;        count = 1;
1735        srcw[0] = PRIV_DATA(cc);        srcw[0] = PRIVATE_DATA(cc);
1736        SLJIT_ASSERT(srcw[0] != 0);        SLJIT_ASSERT(srcw[0] != 0);
1737        cc += 1 + LINK_SIZE;        cc += 1 + LINK_SIZE;
1738        break;        break;
1739    
1740        case OP_CBRA:        case OP_CBRA:
1741        case OP_SCBRA:        case OP_SCBRA:
1742        count = 1;        if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1743        srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));          {
1744            count = 1;
1745            srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1746            }
1747        cc += 1 + LINK_SIZE + IMM2_SIZE;        cc += 1 + LINK_SIZE + IMM2_SIZE;
1748        break;        break;
1749    
1750        case OP_CBRAPOS:        case OP_CBRAPOS:
1751        case OP_SCBRAPOS:        case OP_SCBRAPOS:
1752        count = 2;        count = 2;
1753        srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));        srcw[0] = PRIVATE_DATA(cc);
1754        srcw[1] = PRIV_DATA(cc);        srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1755        SLJIT_ASSERT(srcw[0] != 0);        SLJIT_ASSERT(srcw[0] != 0 && srcw[1] != 0);
1756        cc += 1 + LINK_SIZE + IMM2_SIZE;        cc += 1 + LINK_SIZE + IMM2_SIZE;
1757        break;        break;
1758    
# Line 1432  while (status != end) Line 1762  while (status != end)
1762        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1763          {          {
1764          count = 1;          count = 1;
1765          srcw[0] = PRIV_DATA(cc);          srcw[0] = PRIVATE_DATA(cc);
1766          SLJIT_ASSERT(srcw[0] != 0);          SLJIT_ASSERT(srcw[0] != 0);
1767          }          }
1768        cc += 1 + LINK_SIZE;        cc += 1 + LINK_SIZE;
1769        break;        break;
1770    
1771        CASE_ITERATOR_LOCAL1        CASE_ITERATOR_PRIVATE_DATA_1
1772        if (PRIV_DATA(cc))        if (PRIVATE_DATA(cc))
1773          {          {
1774          count = 1;          count = 1;
1775          srcw[0] = PRIV_DATA(cc);          srcw[0] = PRIVATE_DATA(cc);
1776          }          }
1777        cc += 2;        cc += 2;
1778  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
# Line 1450  while (status != end) Line 1780  while (status != end)
1780  #endif  #endif
1781        break;        break;
1782    
1783        CASE_ITERATOR_LOCAL2A        CASE_ITERATOR_PRIVATE_DATA_2A
1784        if (PRIV_DATA(cc))        if (PRIVATE_DATA(cc))
1785          {          {
1786          count = 2;          count = 2;
1787          srcw[0] = PRIV_DATA(cc);          srcw[0] = PRIVATE_DATA(cc);
1788          srcw[1] = PRIV_DATA(cc) + sizeof(sljit_w);          srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1789          }          }
1790        cc += 2;        cc += 2;
1791  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
# Line 1463  while (status != end) Line 1793  while (status != end)
1793  #endif  #endif
1794        break;        break;
1795    
1796        CASE_ITERATOR_LOCAL2B        CASE_ITERATOR_PRIVATE_DATA_2B
1797        if (PRIV_DATA(cc))        if (PRIVATE_DATA(cc))
1798          {          {
1799          count = 2;          count = 2;
1800          srcw[0] = PRIV_DATA(cc);          srcw[0] = PRIVATE_DATA(cc);
1801          srcw[1] = PRIV_DATA(cc) + sizeof(sljit_w);          srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1802          }          }
1803        cc += 2 + IMM2_SIZE;        cc += 2 + IMM2_SIZE;
1804  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
# Line 1476  while (status != end) Line 1806  while (status != end)
1806  #endif  #endif
1807        break;        break;
1808    
1809        CASE_ITERATOR_TYPE_LOCAL1        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1810        if (PRIV_DATA(cc))        if (PRIVATE_DATA(cc))
1811          {          {
1812          count = 1;          count = 1;
1813          srcw[0] = PRIV_DATA(cc);          srcw[0] = PRIVATE_DATA(cc);
1814          }          }
1815        cc += 1;        cc += 1;
1816        break;        break;
1817    
1818        CASE_ITERATOR_TYPE_LOCAL2A        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1819        if (PRIV_DATA(cc))        if (PRIVATE_DATA(cc))
1820          {          {
1821          count = 2;          count = 2;
1822          srcw[0] = PRIV_DATA(cc);          srcw[0] = PRIVATE_DATA(cc);
1823          srcw[1] = srcw[0] + sizeof(sljit_w);          srcw[1] = srcw[0] + sizeof(sljit_sw);
1824          }          }
1825        cc += 1;        cc += 1;
1826        break;        break;
1827    
1828        CASE_ITERATOR_TYPE_LOCAL2B        CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1829        if (PRIV_DATA(cc))        if (PRIVATE_DATA(cc))
1830          {          {
1831          count = 2;          count = 2;
1832          srcw[0] = PRIV_DATA(cc);          srcw[0] = PRIVATE_DATA(cc);
1833          srcw[1] = srcw[0] + sizeof(sljit_w);          srcw[1] = srcw[0] + sizeof(sljit_sw);
1834          }          }
1835        cc += 1 + IMM2_SIZE;        cc += 1 + IMM2_SIZE;
1836        break;        break;
# Line 1513  while (status != end) Line 1843  while (status != end)
1843  #else  #else
1844        size = 1 + 32 / (int)sizeof(pcre_uchar);        size = 1 + 32 / (int)sizeof(pcre_uchar);
1845  #endif  #endif
1846        if (PRIV_DATA(cc))        if (PRIVATE_DATA(cc))
1847          switch(get_class_iterator_size(cc + size))          switch(get_class_iterator_size(cc + size))
1848            {            {
1849            case 1:            case 1:
1850            count = 1;            count = 1;
1851            srcw[0] = PRIV_DATA(cc);            srcw[0] = PRIVATE_DATA(cc);
1852            break;            break;
1853    
1854            case 2:            case 2:
1855            count = 2;            count = 2;
1856            srcw[0] = PRIV_DATA(cc);            srcw[0] = PRIVATE_DATA(cc);
1857            srcw[1] = srcw[0] + sizeof(sljit_w);            srcw[1] = srcw[0] + sizeof(sljit_sw);
1858            break;            break;
1859    
1860            default:            default:
# Line 1556  while (status != end) Line 1886  while (status != end)
1886          if (!tmp1empty)          if (!tmp1empty)
1887            {            {
1888            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1889            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1890            }            }
1891          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), srcw[count]);
1892          tmp1empty = FALSE;          tmp1empty = FALSE;
1893          tmp1next = FALSE;          tmp1next = FALSE;
1894          }          }
# Line 1567  while (status != end) Line 1897  while (status != end)
1897          if (!tmp2empty)          if (!tmp2empty)
1898            {            {
1899            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1900            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1901            }            }
1902          OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);          OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_SP), srcw[count]);
1903          tmp2empty = FALSE;          tmp2empty = FALSE;
1904          tmp1next = TRUE;          tmp1next = TRUE;
1905          }          }
# Line 1579  while (status != end) Line 1909  while (status != end)
1909        if (tmp1next)        if (tmp1next)
1910          {          {
1911          SLJIT_ASSERT(!tmp1empty);          SLJIT_ASSERT(!tmp1empty);
1912          OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count], TMP1, 0);          OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), srcw[count], TMP1, 0);
1913          tmp1empty = stackptr >= stacktop;          tmp1empty = stackptr >= stacktop;
1914          if (!tmp1empty)          if (!tmp1empty)
1915            {            {
1916            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1917            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1918            }            }
1919          tmp1next = FALSE;          tmp1next = FALSE;
1920          }          }
1921        else        else
1922          {          {
1923          SLJIT_ASSERT(!tmp2empty);          SLJIT_ASSERT(!tmp2empty);
1924          OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count], TMP2, 0);          OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), srcw[count], TMP2, 0);
1925          tmp2empty = stackptr >= stacktop;          tmp2empty = stackptr >= stacktop;
1926          if (!tmp2empty)          if (!tmp2empty)
1927            {            {
1928            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1929            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1930            }            }
1931          tmp1next = TRUE;          tmp1next = TRUE;
1932          }          }
1933        }        }
1934      }      }
1935    }    }
1936    while (status != end);
1937    
1938  if (save)  if (save)
1939    {    {
# Line 1611  if (save) Line 1942  if (save)
1942      if (!tmp1empty)      if (!tmp1empty)
1943        {        {
1944        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1945        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1946        }        }
1947      if (!tmp2empty)      if (!tmp2empty)
1948        {        {
1949        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1950        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1951        }        }
1952      }      }
1953    else    else
# Line 1624  if (save) Line 1955  if (save)
1955      if (!tmp2empty)      if (!tmp2empty)
1956        {        {
1957        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1958        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1959        }        }
1960      if (!tmp1empty)      if (!tmp1empty)
1961        {        {
1962        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1963        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1964        }        }
1965      }      }
1966    }    }
1967  SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));  SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));
1968  }  }
1969    
1970  #undef CASE_ITERATOR_LOCAL1  static SLJIT_INLINE pcre_uchar *set_then_offsets(compiler_common *common, pcre_uchar *cc, pcre_uint8 *current_offset)
1971  #undef CASE_ITERATOR_LOCAL2A  {
1972  #undef CASE_ITERATOR_LOCAL2B  pcre_uchar *end = bracketend(cc);
1973  #undef CASE_ITERATOR_TYPE_LOCAL1  BOOL has_alternatives = cc[GET(cc, 1)] == OP_ALT;
1974  #undef CASE_ITERATOR_TYPE_LOCAL2A  
1975  #undef CASE_ITERATOR_TYPE_LOCAL2B  /* Assert captures then. */
1976    if (*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT)
1977      current_offset = NULL;
1978    /* Conditional block does not. */
1979    if (*cc == OP_COND || *cc == OP_SCOND)
1980      has_alternatives = FALSE;
1981    
1982    cc = next_opcode(common, cc);
1983    if (has_alternatives)
1984      current_offset = common->then_offsets + (cc - common->start);
1985    
1986    while (cc < end)
1987      {
1988      if ((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND))
1989        cc = set_then_offsets(common, cc, current_offset);
1990      else
1991        {
1992        if (*cc == OP_ALT && has_alternatives)
1993          current_offset = common->then_offsets + (cc + 1 + LINK_SIZE - common->start);
1994        if (*cc >= OP_THEN && *cc <= OP_THEN_ARG && current_offset != NULL)
1995          *current_offset = 1;
1996        cc = next_opcode(common, cc);
1997        }
1998      }
1999    
2000    return end;
2001    }
2002    
2003    #undef CASE_ITERATOR_PRIVATE_DATA_1
2004    #undef CASE_ITERATOR_PRIVATE_DATA_2A
2005    #undef CASE_ITERATOR_PRIVATE_DATA_2B
2006    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_1
2007    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
2008    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
2009    
2010  static SLJIT_INLINE BOOL ispowerof2(unsigned int value)  static SLJIT_INLINE BOOL is_powerof2(unsigned int value)
2011  {  {
2012  return (value & (value - 1)) == 0;  return (value & (value - 1)) == 0;
2013  }  }
# Line 1653  static SLJIT_INLINE void set_jumps(jump_ Line 2017  static SLJIT_INLINE void set_jumps(jump_
2017  while (list)  while (list)
2018    {    {
2019    /* sljit_set_label is clever enough to do nothing    /* sljit_set_label is clever enough to do nothing
2020    if either the jump or the label is NULL */    if either the jump or the label is NULL. */
2021    sljit_set_label(list->jump, label);    SET_LABEL(list->jump, label);
2022    list = list->next;    list = list->next;
2023    }    }
2024  }  }
2025    
2026  static SLJIT_INLINE void add_jump(struct sljit_compiler *compiler, jump_list **list, struct sljit_jump* jump)  static SLJIT_INLINE void add_jump(struct sljit_compiler *compiler, jump_list **list, struct sljit_jump *jump)
2027  {  {
2028  jump_list *list_item = sljit_alloc_memory(compiler, sizeof(jump_list));  jump_list *list_item = sljit_alloc_memory(compiler, sizeof(jump_list));
2029  if (list_item)  if (list_item)
# Line 1670  if (list_item) Line 2034  if (list_item)
2034    }    }
2035  }  }
2036    
2037  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)
2038  {  {
2039  DEFINE_COMPILER;  DEFINE_COMPILER;
2040  stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));  stub_list *list_item = sljit_alloc_memory(compiler, sizeof(stub_list));
2041    
2042  if (list_item)  if (list_item)
2043    {    {
   list_item->type = type;  
   list_item->data = data;  
2044    list_item->start = start;    list_item->start = start;
2045    list_item->quit = LABEL();    list_item->quit = LABEL();
2046    list_item->next = common->stubs;    list_item->next = common->stubs;
# Line 1689  if (list_item) Line 2051  if (list_item)
2051  static void flush_stubs(compiler_common *common)  static void flush_stubs(compiler_common *common)
2052  {  {
2053  DEFINE_COMPILER;  DEFINE_COMPILER;
2054  stub_list* list_item = common->stubs;  stub_list *list_item = common->stubs;
2055    
2056  while (list_item)  while (list_item)
2057    {    {
2058    JUMPHERE(list_item->start);    JUMPHERE(list_item->start);
2059    switch(list_item->type)    add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));
     {  
     case stack_alloc:  
     add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));  
     break;  
     }  
2060    JUMPTO(SLJIT_JUMP, list_item->quit);    JUMPTO(SLJIT_JUMP, list_item->quit);
2061    list_item = list_item->next;    list_item = list_item->next;
2062    }    }
2063  common->stubs = NULL;  common->stubs = NULL;
2064  }  }
2065    
2066  static SLJIT_INLINE void decrease_call_count(compiler_common *common)  static void add_label_addr(compiler_common *common, sljit_uw *update_addr)
2067    {
2068    DEFINE_COMPILER;
2069    label_addr_list *label_addr;
2070    
2071    label_addr = sljit_alloc_memory(compiler, sizeof(label_addr_list));
2072    if (label_addr == NULL)
2073      return;
2074    label_addr->label = LABEL();
2075    label_addr->update_addr = update_addr;
2076    label_addr->next = common->label_addrs;
2077    common->label_addrs = label_addr;
2078    }
2079    
2080    static SLJIT_INLINE void count_match(compiler_common *common)
2081  {  {
2082  DEFINE_COMPILER;  DEFINE_COMPILER;
2083    
2084  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);
2085  add_jump(compiler, &common->calllimit, JUMP(SLJIT_C_ZERO));  add_jump(compiler, &common->calllimit, JUMP(SLJIT_ZERO));
2086  }  }
2087    
2088  static SLJIT_INLINE void allocate_stack(compiler_common *common, int size)  static SLJIT_INLINE void allocate_stack(compiler_common *common, int size)
# Line 1719  static SLJIT_INLINE void allocate_stack( Line 2090  static SLJIT_INLINE void allocate_stack(
2090  /* May destroy all locals and registers except TMP2. */  /* May destroy all locals and registers except TMP2. */
2091  DEFINE_COMPILER;  DEFINE_COMPILER;
2092    
2093  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));
2094  #ifdef DESTROY_REGISTERS  #ifdef DESTROY_REGISTERS
2095  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);
2096  OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);  OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);
2097  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);
2098  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), LOCALS0, TMP1, 0);
2099  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), LOCALS1, TMP1, 0);
2100  #endif  #endif
2101  add_stub(common, stack_alloc, 0, CMP(SLJIT_C_GREATER, STACK_TOP, 0, STACK_LIMIT, 0));  add_stub(common, CMP(SLJIT_GREATER, STACK_TOP, 0, STACK_LIMIT, 0));
2102  }  }
2103    
2104  static SLJIT_INLINE void free_stack(compiler_common *common, int size)  static SLJIT_INLINE void free_stack(compiler_common *common, int size)
2105  {  {
2106  DEFINE_COMPILER;  DEFINE_COMPILER;
2107  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));
2108    }
2109    
2110    static sljit_uw * allocate_read_only_data(compiler_common *common, sljit_uw size)
2111    {
2112    DEFINE_COMPILER;
2113    sljit_uw *result;
2114    
2115    if (SLJIT_UNLIKELY(sljit_get_compiler_error(compiler)))
2116      return NULL;
2117    
2118    result = (sljit_uw *)SLJIT_MALLOC(size + sizeof(sljit_uw), compiler->allocator_data);
2119    if (SLJIT_UNLIKELY(result == NULL))
2120      {
2121      sljit_set_compiler_memory_error(compiler);
2122      return NULL;
2123      }
2124    
2125    *(void**)result = common->read_only_data_head;
2126    common->read_only_data_head = (void *)result;
2127    return result + 1;
2128    }
2129    
2130    static void free_read_only_data(void *current, void *allocator_data)
2131    {
2132    void *next;
2133    
2134    SLJIT_UNUSED_ARG(allocator_data);
2135    
2136    while (current != NULL)
2137      {
2138      next = *(void**)current;
2139      SLJIT_FREE(current, allocator_data);
2140      current = next;
2141      }
2142  }  }
2143    
2144  static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)  static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)
# Line 1741  static SLJIT_INLINE void reset_ovector(c Line 2146  static SLJIT_INLINE void reset_ovector(c
2146  DEFINE_COMPILER;  DEFINE_COMPILER;
2147  struct sljit_label *loop;  struct sljit_label *loop;
2148  int i;  int i;
2149    
2150  /* At this point we can freely use all temporary registers. */  /* At this point we can freely use all temporary registers. */
2151    SLJIT_ASSERT(length > 1);
2152  /* TMP1 returns with begin - 1. */  /* TMP1 returns with begin - 1. */
2153  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_R0, 0, SLJIT_MEM1(SLJIT_S0), SLJIT_OFFSETOF(jit_arguments, begin), SLJIT_IMM, IN_UCHARS(1));
2154    if (length < 8)
2155      {
2156      for (i = 1; i < length; i++)
2157        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), OVECTOR(i), SLJIT_R0, 0);
2158      }
2159    else
2160      {
2161      GET_LOCAL_BASE(SLJIT_R1, 0, OVECTOR_START);
2162      OP1(SLJIT_MOV, SLJIT_R2, 0, SLJIT_IMM, length - 1);
2163      loop = LABEL();
2164      OP1(SLJIT_MOVU, SLJIT_MEM1(SLJIT_R1), sizeof(sljit_sw), SLJIT_R0, 0);
2165      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_IMM, 1);
2166      JUMPTO(SLJIT_NOT_ZERO, loop);
2167      }
2168    }
2169    
2170    static SLJIT_INLINE void do_reset_match(compiler_common *common, int length)
2171    {
2172    DEFINE_COMPILER;
2173    struct sljit_label *loop;
2174    int i;
2175    
2176    SLJIT_ASSERT(length > 1);
2177    /* OVECTOR(1) contains the "string begin - 1" constant. */
2178    if (length > 2)
2179      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(1));
2180  if (length < 8)  if (length < 8)
2181    {    {
2182    for (i = 0; i < length; i++)    for (i = 2; i < length; i++)
2183      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), SLJIT_TEMPORARY_REG1, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), OVECTOR(i), TMP1, 0);
2184    }    }
2185  else  else
2186    {    {
2187    GET_LOCAL_BASE(SLJIT_TEMPORARY_REG2, 0, OVECTOR_START - sizeof(sljit_w));    GET_LOCAL_BASE(TMP2, 0, OVECTOR_START + sizeof(sljit_sw));
2188    OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, length);    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_IMM, length - 2);
2189    loop = LABEL();    loop = LABEL();
2190    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);
2191    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);
2192    JUMPTO(SLJIT_C_NOT_ZERO, loop);    JUMPTO(SLJIT_NOT_ZERO, loop);
2193      }
2194    
2195    OP1(SLJIT_MOV, STACK_TOP, 0, ARGUMENTS, 0);
2196    if (common->mark_ptr != 0)
2197      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->mark_ptr, SLJIT_IMM, 0);
2198    if (common->control_head_ptr != 0)
2199      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->control_head_ptr, SLJIT_IMM, 0);
2200    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(jit_arguments, stack));
2201    OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->start_ptr);
2202    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(struct sljit_stack, base));
2203    }
2204    
2205    static sljit_sw SLJIT_CALL do_search_mark(sljit_sw *current, const pcre_uchar *skip_arg)
2206    {
2207    while (current != NULL)
2208      {
2209      switch (current[-2])
2210        {
2211        case type_then_trap:
2212        break;
2213    
2214        case type_mark:
2215        if (STRCMP_UC_UC(skip_arg, (pcre_uchar *)current[-3]) == 0)
2216          return current[-4];
2217        break;
2218    
2219        default:
2220        SLJIT_ASSERT_STOP();
2221        break;
2222        }
2223      current = (sljit_sw*)current[-1];
2224    }    }
2225    return -1;
2226  }  }
2227    
2228  static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)  static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)
2229  {  {
2230  DEFINE_COMPILER;  DEFINE_COMPILER;
2231  struct sljit_label *loop;  struct sljit_label *loop;
2232  struct sljit_jump *earlyexit;  struct sljit_jump *early_quit;
2233    
2234  /* At this point we can freely use all registers. */  /* At this point we can freely use all registers. */
2235  OP1(SLJIT_MOV, SLJIT_SAVED_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));  OP1(SLJIT_MOV, SLJIT_S2, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(1));
2236  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1), STR_PTR, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), OVECTOR(1), STR_PTR, 0);
2237    
2238  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, ARGUMENTS, 0);  OP1(SLJIT_MOV, SLJIT_R0, 0, ARGUMENTS, 0);
2239  if (common->mark_ptr != 0)  if (common->mark_ptr != 0)
2240    OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);    OP1(SLJIT_MOV, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_SP), common->mark_ptr);
2241  OP1(SLJIT_MOV_SI, SLJIT_TEMPORARY_REG2, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, offsetcount));  OP1(SLJIT_MOV_SI, SLJIT_R1, 0, SLJIT_MEM1(SLJIT_R0), SLJIT_OFFSETOF(jit_arguments, offset_count));
2242  if (common->mark_ptr != 0)  if (common->mark_ptr != 0)
2243    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, mark_ptr), SLJIT_TEMPORARY_REG3, 0);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_R0), SLJIT_OFFSETOF(jit_arguments, mark_ptr), SLJIT_R2, 0);
2244  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_R2, 0, SLJIT_MEM1(SLJIT_R0), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));
2245  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, begin));  OP1(SLJIT_MOV, SLJIT_R0, 0, SLJIT_MEM1(SLJIT_R0), SLJIT_OFFSETOF(jit_arguments, begin));
2246  GET_LOCAL_BASE(SLJIT_SAVED_REG1, 0, OVECTOR_START);  GET_LOCAL_BASE(SLJIT_S0, 0, OVECTOR_START);
2247  /* Unlikely, but possible */  /* Unlikely, but possible */
2248  earlyexit = CMP(SLJIT_C_EQUAL, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, 0);  early_quit = CMP(SLJIT_EQUAL, SLJIT_R1, 0, SLJIT_IMM, 0);
2249  loop = LABEL();  loop = LABEL();
2250  OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, SLJIT_MEM1(SLJIT_SAVED_REG1), 0, SLJIT_TEMPORARY_REG1, 0);  OP2(SLJIT_SUB, SLJIT_S1, 0, SLJIT_MEM1(SLJIT_S0), 0, SLJIT_R0, 0);
2251  OP2(SLJIT_ADD, SLJIT_SAVED_REG1, 0, SLJIT_SAVED_REG1, 0, SLJIT_IMM, sizeof(sljit_w));  OP2(SLJIT_ADD, SLJIT_S0, 0, SLJIT_S0, 0, SLJIT_IMM, sizeof(sljit_sw));
2252  /* Copy the integer value to the output buffer */  /* Copy the integer value to the output buffer */
2253  #ifdef COMPILE_PCRE16  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2254  OP2(SLJIT_ASHR, SLJIT_SAVED_REG2, 0, SLJIT_SAVED_REG2, 0, SLJIT_IMM, 1);  OP2(SLJIT_ASHR, SLJIT_S1, 0, SLJIT_S1, 0, SLJIT_IMM, UCHAR_SHIFT);
2255  #endif  #endif
2256  OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_TEMPORARY_REG3), sizeof(int), SLJIT_SAVED_REG2, 0);  OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_R2), sizeof(int), SLJIT_S1, 0);
2257  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_TEMPORARY_REG2, 0, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, 1);  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_R1, 0, SLJIT_R1, 0, SLJIT_IMM, 1);
2258  JUMPTO(SLJIT_C_NOT_ZERO, loop);  JUMPTO(SLJIT_NOT_ZERO, loop);
2259  JUMPHERE(earlyexit);  JUMPHERE(early_quit);
2260    
2261  /* Calculate the return value, which is the maximum ovector value. */  /* Calculate the return value, which is the maximum ovector value. */
2262  if (topbracket > 1)  if (topbracket > 1)
2263    {    {
2264    GET_LOCAL_BASE(SLJIT_TEMPORARY_REG1, 0, OVECTOR_START + topbracket * 2 * sizeof(sljit_w));    GET_LOCAL_BASE(SLJIT_R0, 0, OVECTOR_START + topbracket * 2 * sizeof(sljit_sw));
2265    OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, topbracket + 1);    OP1(SLJIT_MOV, SLJIT_R1, 0, SLJIT_IMM, topbracket + 1);
2266    
2267    /* OVECTOR(0) is never equal to SLJIT_SAVED_REG3. */    /* OVECTOR(0) is never equal to SLJIT_S2. */
2268    loop = LABEL();    loop = LABEL();
2269    OP1(SLJIT_MOVU, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), -(2 * (sljit_w)sizeof(sljit_w)));    OP1(SLJIT_MOVU, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_R0), -(2 * (sljit_sw)sizeof(sljit_sw)));
2270    OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG2, 0, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, 1);    OP2(SLJIT_SUB, SLJIT_R1, 0, SLJIT_R1, 0, SLJIT_IMM, 1);
2271    CMPTO(SLJIT_C_EQUAL, SLJIT_TEMPORARY_REG3, 0, SLJIT_SAVED_REG3, 0, loop);    CMPTO(SLJIT_EQUAL, SLJIT_R2, 0, SLJIT_S2, 0, loop);
2272    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_TEMPORARY_REG2, 0);    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_R1, 0);
2273    }    }
2274  else  else
2275    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, 1);    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, 1);
# Line 1813  else Line 2278  else
2278  static SLJIT_INLINE void return_with_partial_match(compiler_common *common, struct sljit_label *quit)  static SLJIT_INLINE void return_with_partial_match(compiler_common *common, struct sljit_label *quit)
2279  {  {
2280  DEFINE_COMPILER;  DEFINE_COMPILER;
2281    struct sljit_jump *jump;
2282    
2283  SLJIT_COMPILE_ASSERT(STR_END == SLJIT_SAVED_REG2, str_end_must_be_saved_reg2);  SLJIT_COMPILE_ASSERT(STR_END == SLJIT_S1, str_end_must_be_saved_reg2);
2284  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
2285      && (common->mode == JIT_PARTIAL_SOFT_COMPILE ? common->hit_start != 0 : common->hit_start == 0));
2286    
2287  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG2, 0, ARGUMENTS, 0);  OP1(SLJIT_MOV, SLJIT_R1, 0, ARGUMENTS, 0);
2288  OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, PCRE_ERROR_PARTIAL);  OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, PCRE_ERROR_PARTIAL);
2289  OP1(SLJIT_MOV_SI, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), SLJIT_OFFSETOF(jit_arguments, offsetcount));  OP1(SLJIT_MOV_SI, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_R1), SLJIT_OFFSETOF(jit_arguments, real_offset_count));
2290  CMPTO(SLJIT_C_LESS, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, 2, quit);  CMPTO(SLJIT_SIG_LESS, SLJIT_R2, 0, SLJIT_IMM, 2, quit);
2291    
2292  /* Store match begin and end. */  /* Store match begin and end. */
2293  OP1(SLJIT_MOV, SLJIT_SAVED_REG1, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), SLJIT_OFFSETOF(jit_arguments, begin));  OP1(SLJIT_MOV, SLJIT_S0, 0, SLJIT_MEM1(SLJIT_R1), SLJIT_OFFSETOF(jit_arguments, begin));
2294  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG2, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), SLJIT_OFFSETOF(jit_arguments, offsets));  OP1(SLJIT_MOV, SLJIT_R1, 0, SLJIT_MEM1(SLJIT_R1), SLJIT_OFFSETOF(jit_arguments, offsets));
2295  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);  
2296  OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, STR_END, 0, SLJIT_SAVED_REG1, 0);  jump = CMP(SLJIT_SIG_LESS, SLJIT_R2, 0, SLJIT_IMM, 3);
2297  #ifdef COMPILE_PCRE16  OP2(SLJIT_SUB, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_SP), common->mode == JIT_PARTIAL_HARD_COMPILE ? common->start_ptr : (common->hit_start + (int)sizeof(sljit_sw)), SLJIT_S0, 0);
2298  OP2(SLJIT_ASHR, SLJIT_SAVED_REG2, 0, SLJIT_SAVED_REG2, 0, SLJIT_IMM, 1);  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2299  #endif  OP2(SLJIT_ASHR, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_IMM, UCHAR_SHIFT);
 OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), sizeof(int), SLJIT_SAVED_REG2, 0);  
   
 OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG3, 0, SLJIT_TEMPORARY_REG3, 0, SLJIT_SAVED_REG1, 0);  
 #ifdef COMPILE_PCRE16  
 OP2(SLJIT_ASHR, SLJIT_TEMPORARY_REG3, 0, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, 1);  
2300  #endif  #endif
2301  OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), 0, SLJIT_TEMPORARY_REG3, 0);  OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_R1), 2 * sizeof(int), SLJIT_R2, 0);
2302    JUMPHERE(jump);
2303    
2304    OP1(SLJIT_MOV, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_SP), common->mode == JIT_PARTIAL_HARD_COMPILE ? common->start_used_ptr : common->hit_start);
2305    OP2(SLJIT_SUB, SLJIT_S1, 0, STR_END, 0, SLJIT_S0, 0);
2306    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2307    OP2(SLJIT_ASHR, SLJIT_S1, 0, SLJIT_S1, 0, SLJIT_IMM, UCHAR_SHIFT);
2308    #endif
2309    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_R1), sizeof(int), SLJIT_S1, 0);
2310    
2311    OP2(SLJIT_SUB, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_S0, 0);
2312    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2313    OP2(SLJIT_ASHR, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_IMM, UCHAR_SHIFT);
2314    #endif
2315    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_R1), 0, SLJIT_R2, 0);
2316    
2317  JUMPTO(SLJIT_JUMP, quit);  JUMPTO(SLJIT_JUMP, quit);
2318  }  }
# Line 1850  struct sljit_jump *jump; Line 2326  struct sljit_jump *jump;
2326  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2327    {    {
2328    /* The value of -1 must be kept for start_used_ptr! */    /* The value of -1 must be kept for start_used_ptr! */
2329    OP2(SLJIT_ADD, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, SLJIT_IMM, 1);    OP2(SLJIT_ADD, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, SLJIT_IMM, 1);
2330    /* Jumps if start_used_ptr < STR_PTR, or start_used_ptr == -1. Although overwriting    /* Jumps if start_used_ptr < STR_PTR, or start_used_ptr == -1. Although overwriting
2331    is not necessary if start_used_ptr == STR_PTR, it does not hurt as well. */    is not necessary if start_used_ptr == STR_PTR, it does not hurt as well. */
2332    jump = CMP(SLJIT_C_LESS_EQUAL, TMP1, 0, STR_PTR, 0);    jump = CMP(SLJIT_LESS_EQUAL, TMP1, 0, STR_PTR, 0);
2333    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0);
2334    JUMPHERE(jump);    JUMPHERE(jump);
2335    }    }
2336  else if (common->mode == JIT_PARTIAL_HARD_COMPILE)  else if (common->mode == JIT_PARTIAL_HARD_COMPILE)
2337    {    {
2338    jump = CMP(SLJIT_C_LESS_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);    jump = CMP(SLJIT_LESS_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0);
2339    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0);
2340    JUMPHERE(jump);    JUMPHERE(jump);
2341    }    }
2342  }  }
2343    
2344  static SLJIT_INLINE BOOL char_has_othercase(compiler_common *common, pcre_uchar* cc)  static SLJIT_INLINE BOOL char_has_othercase(compiler_common *common, pcre_uchar *cc)
2345  {  {
2346  /* Detects if the character has an othercase. */  /* Detects if the character has an othercase. */
2347  unsigned int c;  unsigned int c;
# Line 1908  if (common->utf && c > 127) Line 2384  if (common->utf && c > 127)
2384  return TABLE_GET(c, common->fcc, c);  return TABLE_GET(c, common->fcc, c);
2385  }  }
2386    
2387  static unsigned int char_get_othercase_bit(compiler_common *common, pcre_uchar* cc)  static unsigned int char_get_othercase_bit(compiler_common *common, pcre_uchar *cc)
2388  {  {
2389  /* Detects if the character and its othercase has only 1 bit difference. */  /* Detects if the character and its othercase has only 1 bit difference. */
2390  unsigned int c, oc, bit;  unsigned int c, oc, bit;
# Line 1949  if (c <= 127 && bit == 0x20) Line 2425  if (c <= 127 && bit == 0x20)
2425    return (0 << 8) | 0x20;    return (0 << 8) | 0x20;
2426    
2427  /* Since c != oc, they must have at least 1 bit difference. */  /* Since c != oc, they must have at least 1 bit difference. */
2428  if (!ispowerof2(bit))  if (!is_powerof2(bit))
2429    return 0;    return 0;
2430    
2431  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
2432    
2433  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
2434  if (common->utf && c > 127)  if (common->utf && c > 127)
# Line 1968  if (common->utf && c > 127) Line 2444  if (common->utf && c > 127)
2444  #endif /* SUPPORT_UTF */  #endif /* SUPPORT_UTF */
2445  return (0 << 8) | bit;  return (0 << 8) | bit;
2446    
2447  #else /* COMPILE_PCRE8 */  #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2448    
 #ifdef COMPILE_PCRE16  
2449  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
2450  if (common->utf && c > 65535)  if (common->utf && c > 65535)
2451    {    {
# Line 1981  if (common->utf && c > 65535) Line 2456  if (common->utf && c > 65535)
2456    }    }
2457  #endif /* SUPPORT_UTF */  #endif /* SUPPORT_UTF */
2458  return (bit < 256) ? ((0 << 8) | bit) : ((1 << 8) | (bit >> 8));  return (bit < 256) ? ((0 << 8) | bit) : ((1 << 8) | (bit >> 8));
 #endif /* COMPILE_PCRE16 */  
2459    
2460  #endif /* COMPILE_PCRE8 */  #endif /* COMPILE_PCRE[8|16|32] */
2461  }  }
2462    
2463  static void check_partial(compiler_common *common, BOOL force)  static void check_partial(compiler_common *common, BOOL force)
2464  {  {
2465  /* Checks whether a partial matching is occured. Does not modify registers. */  /* Checks whether a partial matching is occurred. Does not modify registers. */
2466  DEFINE_COMPILER;  DEFINE_COMPILER;
2467  struct sljit_jump *jump = NULL;  struct sljit_jump *jump = NULL;
2468    
# Line 1998  if (common->mode == JIT_COMPILE) Line 2472  if (common->mode == JIT_COMPILE)
2472    return;    return;
2473    
2474  if (!force)  if (!force)
2475    jump = CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);    jump = CMP(SLJIT_GREATER_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0);
2476  else if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  else if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2477    jump = CMP(SLJIT_C_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, SLJIT_IMM, -1);    jump = CMP(SLJIT_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, SLJIT_IMM, -1);
2478    
2479  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2480    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, -1);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->hit_start, SLJIT_IMM, 0);
2481  else  else
2482    {    {
2483    if (common->partialmatchlabel != NULL)    if (common->partialmatchlabel != NULL)
# Line 2016  if (jump != NULL) Line 2490  if (jump != NULL)
2490    JUMPHERE(jump);    JUMPHERE(jump);
2491  }  }
2492    
2493  static struct sljit_jump *check_str_end(compiler_common *common)  static void check_str_end(compiler_common *common, jump_list **end_reached)
2494  {  {
2495  /* Does not affect registers. Usually used in a tight spot. */  /* Does not affect registers. Usually used in a tight spot. */
2496  DEFINE_COMPILER;  DEFINE_COMPILER;
2497  struct sljit_jump *jump;  struct sljit_jump *jump;
 struct sljit_jump *nohit;  
 struct sljit_jump *return_value;  
2498    
2499  if (common->mode == JIT_COMPILE)  if (common->mode == JIT_COMPILE)
2500    return CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);    {
2501      add_jump(compiler, end_reached, CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2502      return;
2503      }
2504    
2505  jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);  jump = CMP(SLJIT_LESS, STR_PTR, 0, STR_END, 0);
2506  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2507    {    {
2508    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_GREATER_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0));
2509    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, -1);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->hit_start, SLJIT_IMM, 0);
2510    JUMPHERE(nohit);    add_jump(compiler, end_reached, JUMP(SLJIT_JUMP));
   return_value = JUMP(SLJIT_JUMP);  
2511    }    }
2512  else  else
2513    {    {
2514    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_GREATER_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0));
2515    if (common->partialmatchlabel != NULL)    if (common->partialmatchlabel != NULL)
2516      JUMPTO(SLJIT_JUMP, common->partialmatchlabel);      JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2517    else    else
2518      add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));      add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2519    }    }
2520  JUMPHERE(jump);  JUMPHERE(jump);
 return return_value;  
2521  }  }
2522    
2523  static void detect_partial_match(compiler_common *common, jump_list **backtracks)  static void detect_partial_match(compiler_common *common, jump_list **backtracks)
# Line 2054  struct sljit_jump *jump; Line 2527  struct sljit_jump *jump;
2527    
2528  if (common->mode == JIT_COMPILE)  if (common->mode == JIT_COMPILE)
2529    {    {
2530    add_jump(compiler, backtracks, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));    add_jump(compiler, backtracks, CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2531    return;    return;
2532    }    }
2533    
2534  /* Partial matching mode. */  /* Partial matching mode. */
2535  jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);  jump = CMP(SLJIT_LESS, STR_PTR, 0, STR_END, 0);
2536  add_jump(compiler, backtracks, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));  add_jump(compiler, backtracks, CMP(SLJIT_GREATER_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0));
2537  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2538    {    {
2539    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, -1);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->hit_start, SLJIT_IMM, 0);
2540    add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));    add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));
2541    }    }
2542  else  else
# Line 2076  else Line 2549  else
2549  JUMPHERE(jump);  JUMPHERE(jump);
2550  }  }
2551    
2552  static void read_char(compiler_common *common)  static void peek_char(compiler_common *common, pcre_uint32 max)
2553  {  {
2554  /* Reads the character into TMP1, updates STR_PTR.  /* Reads the character into TMP1, keeps STR_PTR.
2555  Does not check STR_END. TMP2 Destroyed. */  Does not check STR_END. TMP2 Destroyed. */
2556  DEFINE_COMPILER;  DEFINE_COMPILER;
2557  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2558  struct sljit_jump *jump;  struct sljit_jump *jump;
2559  #endif  #endif
2560    
2561    SLJIT_UNUSED_ARG(max);
2562    
2563  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2564  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2565  if (common->utf)  if (common->utf)
2566    {    {
2567  #ifdef COMPILE_PCRE8    if (max < 128) return;
2568    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);  
2569  #else    jump = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2570  #ifdef COMPILE_PCRE16    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
   jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);  
 #endif  
 #endif /* COMPILE_PCRE8 */  
2571    add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2572      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2573      JUMPHERE(jump);
2574      }
2575    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2576    
2577    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2578    if (common->utf)
2579      {
2580      if (max < 0xd800) return;
2581    
2582      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2583      jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2584      /* TMP2 contains the high surrogate. */
2585      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2586      OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2587      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2588      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2589      OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2590    JUMPHERE(jump);    JUMPHERE(jump);
2591    }    }
2592  #endif  #endif
2593    }
2594    
2595    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2596    
2597    static BOOL is_char7_bitset(const pcre_uint8 *bitset, BOOL nclass)
2598    {
2599    /* Tells whether the character codes below 128 are enough
2600    to determine a match. */
2601    const pcre_uint8 value = nclass ? 0xff : 0;
2602    const pcre_uint8 *end = bitset + 32;
2603    
2604    bitset += 16;
2605    do
2606      {
2607      if (*bitset++ != value)
2608        return FALSE;
2609      }
2610    while (bitset < end);
2611    return TRUE;
2612    }
2613    
2614    static void read_char7_type(compiler_common *common, BOOL full_read)
2615    {
2616    /* Reads the precise character type of a character into TMP1, if the character
2617    is less than 128. Otherwise it returns with zero. Does not check STR_END. The
2618    full_read argument tells whether characters above max are accepted or not. */
2619    DEFINE_COMPILER;
2620    struct sljit_jump *jump;
2621    
2622    SLJIT_ASSERT(common->utf);
2623    
2624    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2625  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));
2626    
2627    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2628    
2629    if (full_read)
2630      {
2631      jump = CMP(SLJIT_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2632      OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2633      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2634      JUMPHERE(jump);
2635      }
2636  }  }
2637    
2638  static void peek_char(compiler_common *common)  #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2639    
2640    static void read_char_range(compiler_common *common, pcre_uint32 min, pcre_uint32 max, BOOL update_str_ptr)
2641  {  {
2642  /* Reads the character into TMP1, keeps STR_PTR.  /* Reads the precise value of a character into TMP1, if the character is
2643  Does not check STR_END. TMP2 Destroyed. */  between min and max (c >= min && c <= max). Otherwise it returns with a value
2644    outside the range. Does not check STR_END. */
2645  DEFINE_COMPILER;  DEFINE_COMPILER;
2646  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2647  struct sljit_jump *jump;  struct sljit_jump *jump;
2648  #endif  #endif
2649    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2650    struct sljit_jump *jump2;
2651    #endif
2652    
2653  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  SLJIT_UNUSED_ARG(update_str_ptr);
2654  #ifdef SUPPORT_UTF  SLJIT_UNUSED_ARG(min);
2655    SLJIT_UNUSED_ARG(max);
2656    SLJIT_ASSERT(min <= max);
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    
2661    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2662  if (common->utf)  if (common->utf)
2663    {    {
2664  #ifdef COMPILE_PCRE8    if (max < 128 && !update_str_ptr) return;
2665    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);  
2666  #else    jump = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2667  #ifdef COMPILE_PCRE16    if (min >= 0x10000)
2668    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);      {
2669        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xf0);
2670        if (update_str_ptr)
2671          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2672        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2673        jump2 = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0x7);
2674        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2675        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2676        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2677        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2678        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2679        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2680        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2681        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2682        if (!update_str_ptr)
2683          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2684        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2685        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2686        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2687        JUMPHERE(jump2);
2688        if (update_str_ptr)
2689          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2690        }
2691      else if (min >= 0x800 && max <= 0xffff)
2692        {
2693        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xe0);
2694        if (update_str_ptr)
2695          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2696        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2697        jump2 = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xf);
2698        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2699        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2700        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2701        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2702        if (!update_str_ptr)
2703          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2704        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2705        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2706        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2707        JUMPHERE(jump2);
2708        if (update_str_ptr)
2709          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2710        }
2711      else if (max >= 0x800)
2712        add_jump(compiler, (max < 0x10000) ? &common->utfreadchar16 : &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2713      else if (max < 128)
2714        {
2715        OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2716        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2717        }
2718      else
2719        {
2720        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2721        if (!update_str_ptr)
2722          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2723        else
2724          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2725        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2726        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2727        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2728        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2729        if (update_str_ptr)
2730          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2731        }
2732      JUMPHERE(jump);
2733      }
2734  #endif  #endif
2735  #endif /* COMPILE_PCRE8 */  
2736    add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));  #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2737    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);  if (common->utf)
2738      {
2739      if (max >= 0x10000)
2740        {
2741        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2742        jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2743        /* TMP2 contains the high surrogate. */
2744        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2745        OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2746        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2747        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2748        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2749        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2750        JUMPHERE(jump);
2751        return;
2752        }
2753    
2754      if (max < 0xd800 && !update_str_ptr) return;
2755    
2756      /* Skip low surrogate if necessary. */
2757      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2758      jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2759      if (update_str_ptr)
2760        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2761      if (max >= 0xd800)
2762        OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0x10000);
2763    JUMPHERE(jump);    JUMPHERE(jump);
2764    }    }
2765  #endif  #endif
2766  }  }
2767    
2768  static void read_char8_type(compiler_common *common)  static SLJIT_INLINE void read_char(compiler_common *common)
2769    {
2770    read_char_range(common, 0, READ_CHAR_MAX, TRUE);
2771    }
2772    
2773    static void read_char8_type(compiler_common *common, BOOL update_str_ptr)
2774  {  {
2775  /* Reads the character type into TMP1, updates STR_PTR. Does not check STR_END. */  /* Reads the character type into TMP1, updates STR_PTR. Does not check STR_END. */
2776  DEFINE_COMPILER;  DEFINE_COMPILER;
2777  #if defined SUPPORT_UTF || defined COMPILE_PCRE16  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2778  struct sljit_jump *jump;  struct sljit_jump *jump;
2779  #endif  #endif
2780    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2781    struct sljit_jump *jump2;
2782    #endif
2783    
2784  #ifdef SUPPORT_UTF  SLJIT_UNUSED_ARG(update_str_ptr);
2785    
2786    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2787    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2788    
2789    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2790  if (common->utf)  if (common->utf)
2791    {    {
   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  
2792    /* This can be an extra read in some situations, but hopefully    /* This can be an extra read in some situations, but hopefully
2793    it is needed in most cases. */    it is needed in most cases. */
2794    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2795    jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);    jump = CMP(SLJIT_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2796    add_jump(compiler, &common->utfreadtype8, JUMP(SLJIT_FAST_CALL));    if (!update_str_ptr)
2797    JUMPHERE(jump);      {
2798  #else      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2799  #ifdef COMPILE_PCRE16      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2800    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);      OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2801    jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2802    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2803        OP2(SLJIT_OR, TMP2, 0, TMP2, 0, TMP1, 0);
2804        OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2805        jump2 = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 255);
2806        OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2807        JUMPHERE(jump2);
2808        }
2809      else
2810        add_jump(compiler, &common->utfreadtype8, JUMP(SLJIT_FAST_CALL));
2811    JUMPHERE(jump);    JUMPHERE(jump);
   /* Skip low surrogate if necessary. */  
   OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0xfc00);  
   OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0xd800);  
   COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_EQUAL);  
   OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 1);  
   OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);  
 #endif  
 #endif /* COMPILE_PCRE8 */  
2812    return;    return;
2813    }    }
2814  #endif  #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2815  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);  
2816  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  #if !defined COMPILE_PCRE8
 #ifdef COMPILE_PCRE16  
2817  /* The ctypes array contains only 256 values. */  /* The ctypes array contains only 256 values. */
2818  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2819  jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);  jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 255);
2820  #endif  #endif
2821  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2822  #ifdef COMPILE_PCRE16  #if !defined COMPILE_PCRE8
2823  JUMPHERE(jump);  JUMPHERE(jump);
2824  #endif  #endif
2825    
2826    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2827    if (common->utf && update_str_ptr)
2828      {
2829      /* Skip low surrogate if necessary. */
2830      OP2(SLJIT_SUB, TMP2, 0, TMP2, 0, SLJIT_IMM, 0xd800);
2831      jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2832      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2833      JUMPHERE(jump);
2834      }
2835    #endif /* SUPPORT_UTF && COMPILE_PCRE16 */
2836  }  }
2837    
2838  static void skip_char_back(compiler_common *common)  static void skip_char_back(compiler_common *common)
2839  {  {
2840  /* Goes one character back. Affects STR_PTR and TMP1. Does not check begin. */  /* Goes one character back. Affects STR_PTR and TMP1. Does not check begin. */
2841  DEFINE_COMPILER;  DEFINE_COMPILER;
2842  #if defined SUPPORT_UTF && defined COMPILE_PCRE8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2843    #if defined COMPILE_PCRE8
2844  struct sljit_label *label;  struct sljit_label *label;
2845    
2846  if (common->utf)  if (common->utf)
# Line 2193  if (common->utf) Line 2849  if (common->utf)
2849    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
2850    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));
2851    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xc0);    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xc0);
2852    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, 0x80, label);    CMPTO(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, 0x80, label);
2853    return;    return;
2854    }    }
2855  #endif  #elif defined COMPILE_PCRE16
 #if defined SUPPORT_UTF && defined COMPILE_PCRE16  
2856  if (common->utf)  if (common->utf)
2857    {    {
2858    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
# Line 2205  if (common->utf) Line 2860  if (common->utf)
2860    /* Skip low surrogate if necessary. */    /* Skip low surrogate if necessary. */
2861    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
2862    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);
2863    COND_VALUE(SLJIT_MOV, TMP1, 0, SLJIT_C_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
2864    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
2865    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2866    return;    return;
2867    }    }
2868  #endif  #endif /* COMPILE_PCRE[8|16] */
2869    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2870  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));
2871  }  }
2872    
2873  static void check_newlinechar(compiler_common *common, int nltype, jump_list **backtracks, BOOL jumpiftrue)  static void check_newlinechar(compiler_common *common, int nltype, jump_list **backtracks, BOOL jumpifmatch)
2874  {  {
2875  /* 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. */
2876  DEFINE_COMPILER;  DEFINE_COMPILER;
2877    struct sljit_jump *jump;
2878    
2879  if (nltype == NLTYPE_ANY)  if (nltype == NLTYPE_ANY)
2880    {    {
2881    add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));
2882    add_jump(compiler, backtracks, JUMP(jumpiftrue ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));    add_jump(compiler, backtracks, JUMP(jumpifmatch ? SLJIT_NOT_ZERO : SLJIT_ZERO));
2883    }    }
2884  else if (nltype == NLTYPE_ANYCRLF)  else if (nltype == NLTYPE_ANYCRLF)
2885    {    {
2886    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_CR);    if (jumpifmatch)
2887    COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_EQUAL);      {
2888    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_NL);      add_jump(compiler, backtracks, CMP(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR));
2889    COND_VALUE(SLJIT_OR | SLJIT_SET_E, TMP2, 0, SLJIT_C_EQUAL);      add_jump(compiler, backtracks, CMP(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL));
2890    add_jump(compiler, backtracks, JUMP(jumpiftrue ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));      }
2891      else
2892        {
2893        jump = CMP(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR);
2894        add_jump(compiler, backtracks, CMP(SLJIT_NOT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL));
2895        JUMPHERE(jump);
2896        }
2897    }    }
2898  else  else
2899    {    {
2900    SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline < 256);    SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline < 256);
2901    add_jump(compiler, backtracks, CMP(jumpiftrue ? SLJIT_C_EQUAL : SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, common->newline));    add_jump(compiler, backtracks, CMP(jumpifmatch ? SLJIT_EQUAL : SLJIT_NOT_EQUAL, TMP1, 0, SLJIT_IMM, common->newline));
2902    }    }
2903  }  }
2904    
2905  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
2906    
2907  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
2908  static void do_utfreadchar(compiler_common *common)  static void do_utfreadchar(compiler_common *common)
2909  {  {
2910  /* Fast decoding a UTF-8 character. TMP1 contains the first byte  /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2911  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. */
2912  DEFINE_COMPILER;  DEFINE_COMPILER;
2913  struct sljit_jump *jump;  struct sljit_jump *jump;
2914    
2915  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2916  /* Searching for the first zero. */  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2917  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x20);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
 jump = JUMP(SLJIT_C_NOT_ZERO);  
 /* Two byte sequence. */  
 OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  
 OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x1f);  
2918  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2919  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2920  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2921  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(1));  
2922    /* Searching for the first zero. */
2923    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
2924    jump = JUMP(SLJIT_NOT_ZERO);
2925    /* Two byte sequence. */
2926    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2927    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(2));
2928  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
 JUMPHERE(jump);  
2929    
2930  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x10);  JUMPHERE(jump);
 jump = JUMP(SLJIT_C_NOT_ZERO);  
 /* Three byte sequence. */  
2931  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2932  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x0f);  OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
2933  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 12);  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2934  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);  
2935  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2936  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));  
2937    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2938    jump = JUMP(SLJIT_NOT_ZERO);
2939    /* Three byte sequence. */
2940  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));
2941  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));  
2942  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
 JUMPHERE(jump);  
2943    
2944  /* Four byte sequence. */  /* Four byte sequence. */
2945  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  JUMPHERE(jump);
2946  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x07);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2947  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 18);  OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2948    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2949    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2950  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);  
2951  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2952  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(4));
2953    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2954    }
2955    
2956    static void do_utfreadchar16(compiler_common *common)
2957    {
2958    /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2959    of the character (>= 0xc0). Return value in TMP1. */
2960    DEFINE_COMPILER;
2961    struct sljit_jump *jump;
2962    
2963    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2964    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2965    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2966    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2967  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);  
2968  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2969  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(3));  
2970  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));  /* Searching for the first zero. */
2971    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
2972    jump = JUMP(SLJIT_NOT_ZERO);
2973    /* Two byte sequence. */
2974    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2975    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2976    
2977    JUMPHERE(jump);
2978    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x400);
2979    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_NOT_ZERO);
2980    /* This code runs only in 8 bit mode. No need to shift the value. */
2981    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2982    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2983    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
2984    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2985  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2986  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2987  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(3));  /* Three byte sequence. */
2988    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2989  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2990  }  }
2991    
# Line 2311  struct sljit_jump *compare; Line 3000  struct sljit_jump *compare;
3000  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
3001    
3002  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);
3003  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_NOT_ZERO);
3004  /* Two byte sequence. */  /* Two byte sequence. */
3005  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
3006  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));
3007  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);
3008    /* The upper 5 bits are known at this point. */
3009    compare = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0x3);
3010  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
3011  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
3012  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);  
3013  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
3014  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3015    
3016  JUMPHERE(compare);  JUMPHERE(compare);
3017  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
3018  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
 JUMPHERE(jump);  
3019    
3020  /* We only have types for characters less than 256. */  /* We only have types for characters less than 256. */
 OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), (sljit_w)PRIV(utf8_table4) - 0xc0);  
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);  
 OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);  
 sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  
 }  
   
 #else /* COMPILE_PCRE8 */  
   
 #ifdef COMPILE_PCRE16  
 static void do_utfreadchar(compiler_common *common)  
 {  
 /* Fast decoding a UTF-16 character. TMP1 contains the first 16 bit char  
 of the character (>= 0xd800). Return char value in TMP1, length - 1 in TMP2. */  
 DEFINE_COMPILER;  
 struct sljit_jump *jump;  
   
 sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);  
 jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xdc00);  
 /* Do nothing, only return. */  
 sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  
   
3021  JUMPHERE(jump);  JUMPHERE(jump);
3022  /* Combine two 16 bit characters. */  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
3023  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
3024  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
 OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);  
 OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 10);  
 OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3ff);  
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
 OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(1));  
 OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x10000);  
3025  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3026  }  }
 #endif /* COMPILE_PCRE16 */  
3027    
3028  #endif /* COMPILE_PCRE8 */  #endif /* COMPILE_PCRE8 */
3029    
# Line 2384  SLJIT_ASSERT(UCD_BLOCK_SIZE == 128 && si Line 3045  SLJIT_ASSERT(UCD_BLOCK_SIZE == 128 && si
3045    
3046  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
3047  OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);  OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
3048  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_w)PRIV(ucd_stage1));  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(ucd_stage1));
3049  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);
3050  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
3051  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, TMP2, 0);
3052  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, (sljit_w)PRIV(ucd_stage2));  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_stage2));
3053  OP1(SLJIT_MOV_UH, TMP2, 0, SLJIT_MEM2(TMP2, TMP1), 1);  OP1(SLJIT_MOV_UH, TMP2, 0, SLJIT_MEM2(TMP2, TMP1), 1);
3054  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, (sljit_w)PRIV(ucd_records) + SLJIT_OFFSETOF(ucd_record, chartype));  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_records) + SLJIT_OFFSETOF(ucd_record, chartype));
3055  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM2(TMP1, TMP2), 3);  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM2(TMP1, TMP2), 3);
3056  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3057  }  }
# Line 2404  struct sljit_label *newlinelabel = NULL; Line 3065  struct sljit_label *newlinelabel = NULL;
3065  struct sljit_jump *start;  struct sljit_jump *start;
3066  struct sljit_jump *end = NULL;  struct sljit_jump *end = NULL;
3067  struct sljit_jump *nl = NULL;  struct sljit_jump *nl = NULL;
3068  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3069  struct sljit_jump *singlechar;  struct sljit_jump *singlechar;
3070  #endif  #endif
3071  jump_list *newline = NULL;  jump_list *newline = NULL;
# Line 2420  if (firstline) Line 3081  if (firstline)
3081    /* Search for the end of the first line. */    /* Search for the end of the first line. */
3082    SLJIT_ASSERT(common->first_line_end != 0);    SLJIT_ASSERT(common->first_line_end != 0);
3083    OP1(SLJIT_MOV, TMP3, 0, STR_PTR, 0);    OP1(SLJIT_MOV, TMP3, 0, STR_PTR, 0);
   OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_END, 0);  
3084    
3085    if (common->nltype == NLTYPE_FIXED && common->newline > 255)    if (common->nltype == NLTYPE_FIXED && common->newline > 255)
3086      {      {
3087      mainloop = LABEL();      mainloop = LABEL();
3088      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));
3089      end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);      end = CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3090      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(-1));      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(-1));
3091      OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));      OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
3092      CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, mainloop);      CMPTO(SLJIT_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, mainloop);
3093      CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, mainloop);      CMPTO(SLJIT_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, mainloop);
3094      OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));      JUMPHERE(end);
3095        OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_SP), common->first_line_end, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3096      }      }
3097    else    else
3098      {      {
3099      end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);      end = CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3100      mainloop = LABEL();      mainloop = LABEL();
3101      /* Continual stores does not cause data dependency. */      /* Continual stores does not cause data dependency. */
3102      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->first_line_end, STR_PTR, 0);
3103      read_char(common);      read_char_range(common, common->nlmin, common->nlmax, TRUE);
3104      check_newlinechar(common, common->nltype, &newline, TRUE);      check_newlinechar(common, common->nltype, &newline, TRUE);
3105      CMPTO(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0, mainloop);      CMPTO(SLJIT_LESS, STR_PTR, 0, STR_END, 0, mainloop);
3106      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);      JUMPHERE(end);
3107        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->first_line_end, STR_PTR, 0);
3108      set_jumps(newline, LABEL());      set_jumps(newline, LABEL());
3109      }      }
3110    
   JUMPHERE(end);  
3111    OP1(SLJIT_MOV, STR_PTR, 0, TMP3, 0);    OP1(SLJIT_MOV, STR_PTR, 0, TMP3, 0);
3112    }    }
3113    
# Line 2456  if (newlinecheck) Line 3117  if (newlinecheck)
3117    {    {
3118    newlinelabel = LABEL();    newlinelabel = LABEL();
3119    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));
3120    end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);    end = CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3121    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3122    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, common->newline & 0xff);    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, common->newline & 0xff);
3123    COND_VALUE(SLJIT_MOV, TMP1, 0, SLJIT_C_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
3124  #ifdef COMPILE_PCRE16  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
3125    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, UCHAR_SHIFT);
3126  #endif  #endif
3127    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3128    nl = JUMP(SLJIT_JUMP);    nl = JUMP(SLJIT_JUMP);
# Line 2479  if (readuchar) Line 3140  if (readuchar)
3140    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3141    
3142  if (newlinecheck)  if (newlinecheck)
3143    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, newlinelabel);    CMPTO(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, newlinelabel);
3144    
3145  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));
3146  #if defined SUPPORT_UTF && defined COMPILE_PCRE8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3147    #if defined COMPILE_PCRE8
3148  if (common->utf)  if (common->utf)
3149    {    {
3150    singlechar = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);    singlechar = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
3151    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_w)PRIV(utf8_table4) - 0xc0);    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
3152    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3153    JUMPHERE(singlechar);    JUMPHERE(singlechar);
3154    }    }
3155  #endif  #elif defined COMPILE_PCRE16
 #if defined SUPPORT_UTF && defined COMPILE_PCRE16  
3156  if (common->utf)  if (common->utf)
3157    {    {
3158    singlechar = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);    singlechar = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xd800);
3159    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
3160    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);
3161    COND_VALUE(SLJIT_MOV, TMP1, 0, SLJIT_C_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
3162    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
3163    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3164    JUMPHERE(singlechar);    JUMPHERE(singlechar);
3165    }    }
3166  #endif  #endif /* COMPILE_PCRE[8|16] */
3167    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
3168  JUMPHERE(start);  JUMPHERE(start);
3169    
3170  if (newlinecheck)  if (newlinecheck)
# Line 2514  if (newlinecheck) Line 3176  if (newlinecheck)
3176  return mainloop;  return mainloop;
3177  }  }
3178    
3179  static SLJIT_INLINE BOOL fast_forward_first_two_chars(compiler_common *common, BOOL firstline)  #define MAX_N_CHARS 16
3180    #define MAX_N_BYTES 8
3181    
3182    static SLJIT_INLINE void add_prefix_byte(pcre_uint8 byte, pcre_uint8 *bytes)
3183  {  {
3184  DEFINE_COMPILER;  pcre_uint8 len = bytes[0];
3185  struct sljit_label *start;  int i;
 struct sljit_jump *quit;  
 struct sljit_jump *found;  
 pcre_int32 chars[4];  
 pcre_uchar *cc = common->start + 1 + IMM2_SIZE;  
 int location = 0;  
 pcre_int32 len, c, bit, caseless;  
 BOOL must_end;  
3186    
3187  #ifdef COMPILE_PCRE8  if (len == 255)
3188  union {    return;
3189      sljit_uh ascombined;  
3190      sljit_ub asuchars[2];  if (len == 0)
3191  } pair;    {
3192      bytes[0] = 1;
3193      bytes[1] = byte;
3194      return;
3195      }
3196    
3197    for (i = len; i > 0; i--)
3198      if (bytes[i] == byte)
3199        return;
3200    
3201    if (len >= MAX_N_BYTES - 1)
3202      {
3203      bytes[0] = 255;
3204      return;
3205      }
3206    
3207    len++;
3208    bytes[len] = byte;
3209    bytes[0] = len;
3210    }
3211    
3212    static int scan_prefix(compiler_common *common, pcre_uchar *cc, pcre_uint32 *chars, pcre_uint8 *bytes, int max_chars)
3213    {
3214    /* Recursive function, which scans prefix literals. */
3215    BOOL last, any, caseless;
3216    int len, repeat, len_save, consumed = 0;
3217    pcre_uint32 chr, mask;
3218    pcre_uchar *alternative, *cc_save, *oc;
3219    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3220    pcre_uchar othercase[8];
3221    #elif defined SUPPORT_UTF && defined COMPILE_PCRE16
3222    pcre_uchar othercase[2];
3223  #else  #else
3224  union {  pcre_uchar othercase[1];
     sljit_ui ascombined;  
     sljit_uh asuchars[2];  
 } pair;  
3225  #endif  #endif
3226    
3227  if (*(common->start + GET(common->start, 1)) == OP_ALT)  repeat = 1;
   return FALSE;  
   
3228  while (TRUE)  while (TRUE)
3229    {    {
3230    caseless = 0;    last = TRUE;
3231    must_end = TRUE;    any = FALSE;
3232    switch(*cc)    caseless = FALSE;
3233      switch (*cc)
3234      {      {
     case OP_CHAR:  
     must_end = FALSE;  
     cc++;  
     break;  
   
3235      case OP_CHARI:      case OP_CHARI:
3236      caseless = 1;      caseless = TRUE;
3237      must_end = FALSE;      case OP_CHAR:
3238        last = FALSE;
3239      cc++;      cc++;
3240      break;      break;
3241    
# Line 2573  while (TRUE) Line 3254  while (TRUE)
3254      cc++;      cc++;
3255      continue;      continue;
3256    
3257        case OP_ASSERT:
3258        case OP_ASSERT_NOT:
3259        case OP_ASSERTBACK:
3260        case OP_ASSERTBACK_NOT:
3261        cc = bracketend(cc);
3262        continue;
3263    
3264        case OP_PLUSI:
3265        case OP_MINPLUSI:
3266        case OP_POSPLUSI:
3267        caseless = TRUE;
3268      case OP_PLUS:      case OP_PLUS:
3269      case OP_MINPLUS:      case OP_MINPLUS:
3270      case OP_POSPLUS:      case OP_POSPLUS:
3271      cc++;      cc++;
3272      break;      break;
3273    
3274        case OP_EXACTI:
3275        caseless = TRUE;
3276      case OP_EXACT:      case OP_EXACT:
3277        repeat = GET2(cc, 1);
3278        last = FALSE;
3279      cc += 1 + IMM2_SIZE;      cc += 1 + IMM2_SIZE;
3280      break;      break;
3281    
3282      case OP_PLUSI:      case OP_QUERYI:
3283      case OP_MINPLUSI:      case OP_MINQUERYI:
3284      case OP_POSPLUSI:      case OP_POSQUERYI:
3285      caseless = 1;      caseless = TRUE;
3286        case OP_QUERY:
3287        case OP_MINQUERY:
3288        case OP_POSQUERY:
3289        len = 1;
3290      cc++;      cc++;
3291    #ifdef SUPPORT_UTF
3292        if (common->utf && HAS_EXTRALEN(*cc)) len += GET_EXTRALEN(*cc);
3293    #endif
3294        max_chars = scan_prefix(common, cc + len, chars, bytes, max_chars);
3295        if (max_chars == 0)
3296          return consumed;
3297        last = FALSE;
3298      break;      break;
3299    
3300      case OP_EXACTI:      case OP_KET:
3301      caseless = 1;      cc += 1 + LINK_SIZE;
3302        continue;
3303    
3304        case OP_ALT:
3305        cc += GET(cc, 1);
3306        continue;
3307    
3308        case OP_ONCE:
3309        case OP_ONCE_NC:
3310        case OP_BRA:
3311        case OP_BRAPOS:
3312        case OP_CBRA:
3313        case OP_CBRAPOS:
3314        alternative = cc + GET(cc, 1);
3315        while (*alternative == OP_ALT)
3316          {
3317          max_chars = scan_prefix(common, alternative + 1 + LINK_SIZE, chars, bytes, max_chars);
3318          if (max_chars == 0)
3319            return consumed;
3320          alternative += GET(alternative, 1);
3321          }
3322    
3323        if (*cc == OP_CBRA || *cc == OP_CBRAPOS)
3324          cc += IMM2_SIZE;
3325        cc += 1 + LINK_SIZE;
3326        continue;
3327    
3328        case OP_CLASS:
3329    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3330        if (common->utf && !is_char7_bitset((const pcre_uint8 *)(cc + 1), FALSE)) return consumed;
3331    #endif
3332        any = TRUE;
3333        cc += 1 + 32 / sizeof(pcre_uchar);
3334        break;
3335    
3336        case OP_NCLASS:
3337    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3338        if (common->utf) return consumed;
3339    #endif
3340        any = TRUE;
3341        cc += 1 + 32 / sizeof(pcre_uchar);
3342        break;
3343    
3344    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3345        case OP_XCLASS:
3346    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3347        if (common->utf) return consumed;
3348    #endif
3349        any = TRUE;
3350        cc += GET(cc, 1);
3351        break;
3352    #endif
3353    
3354        case OP_DIGIT:
3355    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3356        if (common->utf && !is_char7_bitset((const pcre_uint8 *)common->ctypes - cbit_length + cbit_digit, FALSE))
3357          return consumed;
3358    #endif
3359        any = TRUE;
3360        cc++;
3361        break;
3362    
3363        case OP_WHITESPACE:
3364    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3365        if (common->utf && !is_char7_bitset((const pcre_uint8 *)common->ctypes - cbit_length + cbit_space, FALSE))
3366          return consumed;
3367    #endif
3368        any = TRUE;
3369        cc++;
3370        break;
3371    
3372        case OP_WORDCHAR:
3373    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3374        if (common->utf && !is_char7_bitset((const pcre_uint8 *)common->ctypes - cbit_length + cbit_word, FALSE))
3375          return consumed;
3376    #endif
3377        any = TRUE;
3378        cc++;
3379        break;
3380    
3381        case OP_NOT:
3382        case OP_NOTI:
3383        cc++;
3384        /* Fall through. */
3385        case OP_NOT_DIGIT:
3386        case OP_NOT_WHITESPACE:
3387        case OP_NOT_WORDCHAR:
3388        case OP_ANY:
3389        case OP_ALLANY:
3390    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3391        if (common->utf) return consumed;
3392    #endif
3393        any = TRUE;
3394        cc++;
3395        break;
3396    
3397    #ifdef SUPPORT_UCP
3398        case OP_NOTPROP:
3399        case OP_PROP:
3400    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3401        if (common->utf) return consumed;
3402    #endif
3403        any = TRUE;
3404        cc += 1 + 2;
3405        break;
3406    #endif
3407    
3408        case OP_TYPEEXACT:
3409        repeat = GET2(cc, 1);
3410      cc += 1 + IMM2_SIZE;      cc += 1 + IMM2_SIZE;
3411        continue;
3412    
3413        case OP_NOTEXACT:
3414        case OP_NOTEXACTI:
3415    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3416        if (common->utf) return consumed;
3417    #endif
3418        any = TRUE;
3419        repeat = GET2(cc, 1);
3420        cc += 1 + IMM2_SIZE + 1;
3421      break;      break;
3422    
3423      default:      default:
3424      return FALSE;      return consumed;
3425        }
3426    
3427      if (any)
3428        {
3429    #if defined COMPILE_PCRE8
3430        mask = 0xff;
3431    #elif defined COMPILE_PCRE16
3432        mask = 0xffff;
3433    #elif defined COMPILE_PCRE32
3434        mask = 0xffffffff;
3435    #else
3436        SLJIT_ASSERT_STOP();
3437    #endif
3438    
3439        do
3440          {
3441          chars[0] = mask;
3442          chars[1] = mask;
3443          bytes[0] = 255;
3444    
3445          consumed++;
3446          if (--max_chars == 0)
3447            return consumed;
3448          chars += 2;
3449          bytes += MAX_N_BYTES;
3450          }
3451        while (--repeat > 0);
3452    
3453        repeat = 1;
3454        continue;
3455      }      }
3456    
3457    len = 1;    len = 1;
3458  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
3459    if (common->utf && HAS_EXTRALEN(cc[0])) len += GET_EXTRALEN(cc[0]);    if (common->utf && HAS_EXTRALEN(*cc)) len += GET_EXTRALEN(*cc);
3460  #endif  #endif
3461    
3462    if (caseless && char_has_othercase(common, cc))    if (caseless && char_has_othercase(common, cc))
3463      {      {
3464      caseless = char_get_othercase_bit(common, cc);  #ifdef SUPPORT_UTF
3465      if (caseless == 0)      if (common->utf)
3466        return FALSE;        {
3467  #ifdef COMPILE_PCRE8        GETCHAR(chr, cc);
3468      caseless = ((caseless & 0xff) << 8) | (len - (caseless >> 8));        if ((int)PRIV(ord2utf)(char_othercase(common, chr), othercase) != len)
3469  #else          return consumed;
3470      if ((caseless & 0x100) != 0)        }
       caseless = ((caseless & 0xff) << 16) | (len - (caseless >> 9));  
3471      else      else
       caseless = ((caseless & 0xff) << 8) | (len - (caseless >> 9));  
3472  #endif  #endif
3473          {
3474          chr = *cc;
3475          othercase[0] = TABLE_GET(chr, common->fcc, chr);
3476          }
3477      }      }
3478    else    else
3479      caseless = 0;      caseless = FALSE;
3480    
3481    while (len > 0 && location < 2 * 2)    len_save = len;
3482      cc_save = cc;
3483      while (TRUE)
3484      {      {
3485      c = *cc;      oc = othercase;
3486      bit = 0;      do
     if (len == (caseless & 0xff))  
3487        {        {
3488        bit = caseless >> 8;        chr = *cc;
3489        c |= bit;  #ifdef COMPILE_PCRE32
3490          if (SLJIT_UNLIKELY(chr == NOTACHAR))
3491            return consumed;
3492    #endif
3493          add_prefix_byte((pcre_uint8)chr, bytes);
3494    
3495          mask = 0;
3496          if (caseless)
3497            {
3498            add_prefix_byte((pcre_uint8)*oc, bytes);
3499            mask = *cc ^ *oc;
3500            chr |= mask;
3501            }
3502    
3503    #ifdef COMPILE_PCRE32
3504          if (chars[0] == NOTACHAR && chars[1] == 0)
3505    #else
3506          if (chars[0] == NOTACHAR)
3507    #endif
3508            {
3509            chars[0] = chr;
3510            chars[1] = mask;
3511            }
3512          else
3513            {
3514            mask |= chars[0] ^ chr;
3515            chr |= mask;
3516            chars[0] = chr;
3517            chars[1] |= mask;
3518            }
3519    
3520          len--;
3521          consumed++;
3522          if (--max_chars == 0)
3523            return consumed;
3524          chars += 2;
3525          bytes += MAX_N_BYTES;
3526          cc++;
3527          oc++;
3528        }        }
3529        while (len > 0);
3530    
3531      chars[location] = c;      if (--repeat == 0)
3532      chars[location + 1] = bit;        break;
3533    
3534      len--;      len = len_save;
3535      location += 2;      cc = cc_save;
3536      cc++;      }
3537    
3538      repeat = 1;
3539      if (last)
3540        return consumed;
3541      }
3542    }
3543    
3544    static SLJIT_INLINE BOOL fast_forward_first_n_chars(compiler_common *common, BOOL firstline)
3545    {
3546    DEFINE_COMPILER;
3547    struct sljit_label *start;
3548    struct sljit_jump *quit;
3549    pcre_uint32 chars[MAX_N_CHARS * 2];
3550    pcre_uint8 bytes[MAX_N_CHARS * MAX_N_BYTES];
3551    pcre_uint8 ones[MAX_N_CHARS];
3552    int offsets[3];
3553    pcre_uint32 mask;
3554    pcre_uint8 *byte_set, *byte_set_end;
3555    int i, max, from;
3556    int range_right = -1, range_len = 3 - 1;
3557    sljit_ub *update_table = NULL;
3558    BOOL in_range;
3559    
3560    for (i = 0; i < MAX_N_CHARS; i++)
3561      {
3562      chars[i << 1] = NOTACHAR;
3563      chars[(i << 1) + 1] = 0;
3564      bytes[i * MAX_N_BYTES] = 0;
3565      }
3566    
3567    max = scan_prefix(common, common->start, chars, bytes, MAX_N_CHARS);
3568    
3569    if (max <= 1)
3570      return FALSE;
3571    
3572    for (i = 0; i < max; i++)
3573      {
3574      mask = chars[(i << 1) + 1];
3575      ones[i] = ones_in_half_byte[mask & 0xf];
3576      mask >>= 4;
3577      while (mask != 0)
3578        {
3579        ones[i] += ones_in_half_byte[mask & 0xf];
3580        mask >>= 4;
3581        }
3582      }
3583    
3584    in_range = FALSE;
3585    from = 0;   /* Prevent compiler "uninitialized" warning */
3586    for (i = 0; i <= max; i++)
3587      {
3588      if (in_range && (i - from) > range_len && (bytes[(i - 1) * MAX_N_BYTES] <= 4))
3589        {
3590        range_len = i - from;
3591        range_right = i - 1;
3592        }
3593    
3594      if (i < max && bytes[i * MAX_N_BYTES] < 255)
3595        {
3596        if (!in_range)
3597          {
3598          in_range = TRUE;
3599          from = i;
3600          }