/[pcre]/code/tags/pcre-8.37/pcre_jit_compile.c
ViewVC logotype

Diff of /code/tags/pcre-8.37/pcre_jit_compile.c

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

revision 1221 by ph10, Sun Nov 11 20:27:03 2012 UTC revision 1437 by zherczeg, Fri Jan 10 08:52:20 2014 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 65  system files. */ Line 65  system files. */
65  #error Unsupported architecture  #error Unsupported architecture
66  #endif  #endif
67    
68    /* Defines for debugging purposes. */
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.  /* Allocate memory for the regex stack on the real machine stack.
78  Fast, but limited size. */  Fast, but limited size. */
79  #define MACHINE_STACK_SIZE 32768  #define MACHINE_STACK_SIZE 32768
# Line 157  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 168  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      sljit_uw *read_only_data[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    pcre_uint32 top_bracket;    pcre_uint32 top_bracket;
187    sljit_uw executable_sizes[JIT_NUMBER_OF_COMPILE_MODES];    pcre_uint32 limit_match;
188  } executable_functions;  } executable_functions;
189    
190  typedef struct jump_list {  typedef struct jump_list {
# Line 179  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 *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_matchingpath, and contains  code generator. It is allocated by compile_matchingpath, and contains
221  the aguments for compile_backtrackingpath. 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 209  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 private_data_ptr;    int private_data_ptr;
# Line 230  typedef struct bracket_backtrack { Line 255  typedef struct bracket_backtrack {
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. */
# Line 265  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. */    /* Maps private data offset to each opcode. */
324    int *private_data_ptrs;    sljit_si *private_data_ptrs;
325      /* This read-only data is available during runtime. */
326      sljit_uw *read_only_data;
327      /* The total size of the read-only data. */
328      sljit_uw read_only_data_size;
329      /* The next free entry of the read_only_data. */
330      sljit_uw *read_only_data_ptr;
331    /* Tells whether the capturing bracket is optimized. */    /* Tells whether the capturing bracket is optimized. */
332    pcre_uint8 *optimized_cbracket;    pcre_uint8 *optimized_cbracket;
333      /* Tells whether the starting offset is a target of then. */
334      pcre_uint8 *then_offsets;
335      /* Current position where a THEN must jump. */
336      then_trap_backtrack *then_trap;
337    /* Starting offset of private data for capturing brackets. */    /* Starting offset of private data for capturing brackets. */
338    int cbraptr;    int cbra_ptr;
339    /* OVector starting point. Must be divisible by 2. */    /* Output vector starting point. Must be divisible by 2. */
340    int ovector_start;    int ovector_start;
341    /* Last known position of the requested byte. */    /* Last known position of the requested byte. */
342    int req_char_ptr;    int req_char_ptr;
343    /* Head of the last recursion. */    /* Head of the last recursion. */
344    int recursive_head;    int recursive_head_ptr;
345    /* First inspected character for partial matching. */    /* First inspected character for partial matching. */
346    int start_used_ptr;    int start_used_ptr;
347    /* Starting pointer for partial soft matches. */    /* Starting pointer for partial soft matches. */
# Line 298  typedef struct compiler_common { Line 350  typedef struct compiler_common {
350    int first_line_end;    int first_line_end;
351    /* Points to the marked string. */    /* Points to the marked string. */
352    int mark_ptr;    int mark_ptr;
353      /* Recursive control verb management chain. */
354      int control_head_ptr;
355      /* Points to the last matched capture block index. */
356      int capture_last_ptr;
357      /* Points to the starting position of the current match. */
358      int start_ptr;
359    
360    /* Flipped and lower case tables. */    /* Flipped and lower case tables. */
361    const pcre_uint8 *fcc;    const pcre_uint8 *fcc;
362    sljit_sw lcc;    sljit_sw lcc;
363    /* Mode can be PCRE_STUDY_JIT_COMPILE and others. */    /* Mode can be PCRE_STUDY_JIT_COMPILE and others. */
364    int mode;    int mode;
365      /* TRUE, when minlength is greater than 0. */
366      BOOL might_be_empty;
367      /* \K is found in the pattern. */
368      BOOL has_set_som;
369      /* (*SKIP:arg) is found in the pattern. */
370      BOOL has_skip_arg;
371      /* (*THEN) is found in the pattern. */
372      BOOL has_then;
373      /* Needs to know the start position anytime. */
374      BOOL needs_start_ptr;
375      /* Currently in recurse or negative assert. */
376      BOOL local_exit;
377      /* Currently in a positive assert. */
378      BOOL positive_assert;
379    /* Newline control. */    /* Newline control. */
380    int nltype;    int nltype;
381      pcre_uint32 nlmax;
382      pcre_uint32 nlmin;
383    int newline;    int newline;
384    int bsr_nltype;    int bsr_nltype;
385      pcre_uint32 bsr_nlmax;
386      pcre_uint32 bsr_nlmin;
387    /* Dollar endonly. */    /* Dollar endonly. */
388    int endonly;    int endonly;
   BOOL has_set_som;  
389    /* Tables. */    /* Tables. */
390    sljit_sw ctypes;    sljit_sw ctypes;
   int digits[2 + MAX_RANGE_SIZE];  
391    /* Named capturing brackets. */    /* Named capturing brackets. */
392    sljit_uw name_table;    pcre_uchar *name_table;
393    sljit_sw name_count;    sljit_sw name_count;
394    sljit_sw name_entry_size;    sljit_sw name_entry_size;
395    
396    /* Labels and jump lists. */    /* Labels and jump lists. */
397    struct sljit_label *partialmatchlabel;    struct sljit_label *partialmatchlabel;
398    struct sljit_label *quitlabel;    struct sljit_label *quit_label;
399    struct sljit_label *acceptlabel;    struct sljit_label *forced_quit_label;
400      struct sljit_label *accept_label;
401    stub_list *stubs;    stub_list *stubs;
402      label_addr_list *label_addrs;
403    recurse_entry *entries;    recurse_entry *entries;
404    recurse_entry *currententry;    recurse_entry *currententry;
405    jump_list *partialmatch;    jump_list *partialmatch;
406    jump_list *quit;    jump_list *quit;
407      jump_list *positive_assert_quit;
408      jump_list *forced_quit;
409    jump_list *accept;    jump_list *accept;
410    jump_list *calllimit;    jump_list *calllimit;
411    jump_list *stackalloc;    jump_list *stackalloc;
# Line 338  typedef struct compiler_common { Line 416  typedef struct compiler_common {
416    jump_list *vspace;    jump_list *vspace;
417    jump_list *casefulcmp;    jump_list *casefulcmp;
418    jump_list *caselesscmp;    jump_list *caselesscmp;
419      jump_list *reset_match;
420    BOOL jscript_compat;    BOOL jscript_compat;
421  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
422    BOOL utf;    BOOL utf;
423  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
424    BOOL use_ucp;    BOOL use_ucp;
425  #endif  #endif
 #ifndef COMPILE_PCRE32  
   jump_list *utfreadchar;  
 #endif  
426  #ifdef COMPILE_PCRE8  #ifdef COMPILE_PCRE8
427      jump_list *utfreadchar;
428      jump_list *utfreadchar16;
429    jump_list *utfreadtype8;    jump_list *utfreadtype8;
430  #endif  #endif
431  #endif /* SUPPORT_UTF */  #endif /* SUPPORT_UTF */
# Line 390  typedef struct compare_context { Line 468  typedef struct compare_context {
468  #endif  #endif
469  } compare_context;  } compare_context;
470    
 enum {  
   frame_end = 0,  
   frame_setstrbegin = -1,  
   frame_setmark = -2  
 };  
   
471  /* Undefine sljit macros. */  /* Undefine sljit macros. */
472  #undef CMP  #undef CMP
473    
# Line 410  enum { Line 482  enum {
482  #define STACK_TOP     SLJIT_SCRATCH_REG2  #define STACK_TOP     SLJIT_SCRATCH_REG2
483  #define STACK_LIMIT   SLJIT_SAVED_REG3  #define STACK_LIMIT   SLJIT_SAVED_REG3
484  #define ARGUMENTS     SLJIT_SAVED_EREG1  #define ARGUMENTS     SLJIT_SAVED_EREG1
485  #define CALL_COUNT    SLJIT_SAVED_EREG2  #define COUNT_MATCH   SLJIT_SAVED_EREG2
486  #define RETURN_ADDR   SLJIT_TEMPORARY_EREG1  #define RETURN_ADDR   SLJIT_TEMPORARY_EREG1
487    
488  /* Local space layout. */  /* Local space layout. */
# Line 421  enum { Line 493  enum {
493  #define POSSESSIVE0      (2 * sizeof(sljit_sw))  #define POSSESSIVE0      (2 * sizeof(sljit_sw))
494  #define POSSESSIVE1      (3 * sizeof(sljit_sw))  #define POSSESSIVE1      (3 * sizeof(sljit_sw))
495  /* Max limit of recursions. */  /* Max limit of recursions. */
496  #define CALL_LIMIT       (4 * sizeof(sljit_sw))  #define LIMIT_MATCH      (4 * sizeof(sljit_sw))
497  /* The output vector is stored on the stack, and contains pointers  /* The output vector is stored on the stack, and contains pointers
498  to characters. The vector data is divided into two groups: the first  to characters. The vector data is divided into two groups: the first
499  group contains the start / end character pointers, and the second is  group contains the start / end character pointers, and the second is
500  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. */
501  #define OVECTOR_START    (common->ovector_start)  #define OVECTOR_START    (common->ovector_start)
502  #define OVECTOR(i)       (OVECTOR_START + (i) * sizeof(sljit_sw))  #define OVECTOR(i)       (OVECTOR_START + (i) * (sljit_sw)sizeof(sljit_sw))
503  #define OVECTOR_PRIV(i)  (common->cbraptr + (i) * sizeof(sljit_sw))  #define OVECTOR_PRIV(i)  (common->cbra_ptr + (i) * (sljit_sw)sizeof(sljit_sw))
504  #define PRIVATE_DATA(cc) (common->private_data_ptrs[(cc) - common->start])  #define PRIVATE_DATA(cc) (common->private_data_ptrs[(cc) - common->start])
505    
506  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
# Line 459  the start pointers when the end of the c Line 531  the start pointers when the end of the c
531    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))
532  #define JUMPHERE(jump) \  #define JUMPHERE(jump) \
533    sljit_set_label((jump), sljit_emit_label(compiler))    sljit_set_label((jump), sljit_emit_label(compiler))
534    #define SET_LABEL(jump, label) \
535      sljit_set_label((jump), (label))
536  #define CMP(type, src1, src1w, src2, src2w) \  #define CMP(type, src1, src1w, src2, src2w) \
537    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))
538  #define CMPTO(type, src1, src1w, src2, src2w, label) \  #define CMPTO(type, src1, src1w, src2, src2w, label) \
# Line 468  the start pointers when the end of the c Line 542  the start pointers when the end of the c
542  #define GET_LOCAL_BASE(dst, dstw, offset) \  #define GET_LOCAL_BASE(dst, dstw, offset) \
543    sljit_get_local_base(compiler, (dst), (dstw), (offset))    sljit_get_local_base(compiler, (dst), (dstw), (offset))
544    
545    #define READ_CHAR_MAX 0x7fffffff
546    
547  static pcre_uchar* bracketend(pcre_uchar* cc)  static pcre_uchar* bracketend(pcre_uchar* cc)
548  {  {
549  SLJIT_ASSERT((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND));  SLJIT_ASSERT((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND));
# Line 477  cc += 1 + LINK_SIZE; Line 553  cc += 1 + LINK_SIZE;
553  return cc;  return cc;
554  }  }
555    
556    static int no_alternatives(pcre_uchar* cc)
557    {
558    int count = 0;
559    SLJIT_ASSERT((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND));
560    do
561      {
562      cc += GET(cc, 1);
563      count++;
564      }
565    while (*cc == OP_ALT);
566    SLJIT_ASSERT(*cc >= OP_KET && *cc <= OP_KETRPOS);
567    return count;
568    }
569    
570    static int ones_in_half_byte[16] = {
571      /* 0 */ 0, 1, 1, 2, /* 4 */ 1, 2, 2, 3,
572      /* 8 */ 1, 2, 2, 3, /* 12 */ 2, 3, 3, 4
573    };
574    
575  /* Functions whose might need modification for all new supported opcodes:  /* Functions whose might need modification for all new supported opcodes:
576   next_opcode   next_opcode
577   get_private_data_length   check_opcode_types
578   set_private_data_ptrs   set_private_data_ptrs
579   get_framesize   get_framesize
580   init_frame   init_frame
581   get_private_data_length_for_copy   get_private_data_copy_length
582   copy_private_data   copy_private_data
583   compile_matchingpath   compile_matchingpath
584   compile_backtrackingpath   compile_backtrackingpath
# Line 507  switch(*cc) Line 602  switch(*cc)
602    case OP_WORDCHAR:    case OP_WORDCHAR:
603    case OP_ANY:    case OP_ANY:
604    case OP_ALLANY:    case OP_ALLANY:
605      case OP_NOTPROP:
606      case OP_PROP:
607    case OP_ANYNL:    case OP_ANYNL:
608    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
609    case OP_HSPACE:    case OP_HSPACE:
# Line 519  switch(*cc) Line 616  switch(*cc)
616    case OP_CIRCM:    case OP_CIRCM:
617    case OP_DOLL:    case OP_DOLL:
618    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:  
619    case OP_CRSTAR:    case OP_CRSTAR:
620    case OP_CRMINSTAR:    case OP_CRMINSTAR:
621    case OP_CRPLUS:    case OP_CRPLUS:
622    case OP_CRMINPLUS:    case OP_CRMINPLUS:
623    case OP_CRQUERY:    case OP_CRQUERY:
624    case OP_CRMINQUERY:    case OP_CRMINQUERY:
625      case OP_CRRANGE:
626      case OP_CRMINRANGE:
627      case OP_CRPOSSTAR:
628      case OP_CRPOSPLUS:
629      case OP_CRPOSQUERY:
630      case OP_CRPOSRANGE:
631      case OP_CLASS:
632      case OP_NCLASS:
633      case OP_REF:
634      case OP_REFI:
635      case OP_DNREF:
636      case OP_DNREFI:
637      case OP_RECURSE:
638      case OP_CALLOUT:
639      case OP_ALT:
640      case OP_KET:
641      case OP_KETRMAX:
642      case OP_KETRMIN:
643      case OP_KETRPOS:
644      case OP_REVERSE:
645      case OP_ASSERT:
646      case OP_ASSERT_NOT:
647      case OP_ASSERTBACK:
648      case OP_ASSERTBACK_NOT:
649      case OP_ONCE:
650      case OP_ONCE_NC:
651      case OP_BRA:
652      case OP_BRAPOS:
653      case OP_CBRA:
654      case OP_CBRAPOS:
655      case OP_COND:
656      case OP_SBRA:
657      case OP_SBRAPOS:
658      case OP_SCBRA:
659      case OP_SCBRAPOS:
660      case OP_SCOND:
661      case OP_CREF:
662      case OP_DNCREF:
663      case OP_RREF:
664      case OP_DNRREF:
665    case OP_DEF:    case OP_DEF:
666    case OP_BRAZERO:    case OP_BRAZERO:
667    case OP_BRAMINZERO:    case OP_BRAMINZERO:
668    case OP_BRAPOSZERO:    case OP_BRAPOSZERO:
669      case OP_PRUNE:
670      case OP_SKIP:
671      case OP_THEN:
672    case OP_COMMIT:    case OP_COMMIT:
673    case OP_FAIL:    case OP_FAIL:
674    case OP_ACCEPT:    case OP_ACCEPT:
675    case OP_ASSERT_ACCEPT:    case OP_ASSERT_ACCEPT:
676      case OP_CLOSE:
677    case OP_SKIPZERO:    case OP_SKIPZERO:
678    return cc + 1;    return cc + PRIV(OP_lengths)[*cc];
   
   case OP_ANYBYTE:  
 #ifdef SUPPORT_UTF  
   if (common->utf) return NULL;  
 #endif  
   return cc + 1;  
679    
680    case OP_CHAR:    case OP_CHAR:
681    case OP_CHARI:    case OP_CHARI:
# Line 561  switch(*cc) Line 687  switch(*cc)
687    case OP_MINPLUS:    case OP_MINPLUS:
688    case OP_QUERY:    case OP_QUERY:
689    case OP_MINQUERY:    case OP_MINQUERY:
690      case OP_UPTO:
691      case OP_MINUPTO:
692      case OP_EXACT:
693    case OP_POSSTAR:    case OP_POSSTAR:
694    case OP_POSPLUS:    case OP_POSPLUS:
695    case OP_POSQUERY:    case OP_POSQUERY:
696      case OP_POSUPTO:
697    case OP_STARI:    case OP_STARI:
698    case OP_MINSTARI:    case OP_MINSTARI:
699    case OP_PLUSI:    case OP_PLUSI:
700    case OP_MINPLUSI:    case OP_MINPLUSI:
701    case OP_QUERYI:    case OP_QUERYI:
702    case OP_MINQUERYI:    case OP_MINQUERYI:
703      case OP_UPTOI:
704      case OP_MINUPTOI:
705      case OP_EXACTI:
706    case OP_POSSTARI:    case OP_POSSTARI:
707    case OP_POSPLUSI:    case OP_POSPLUSI:
708    case OP_POSQUERYI:    case OP_POSQUERYI:
709      case OP_POSUPTOI:
710    case OP_NOTSTAR:    case OP_NOTSTAR:
711    case OP_NOTMINSTAR:    case OP_NOTMINSTAR:
712    case OP_NOTPLUS:    case OP_NOTPLUS:
713    case OP_NOTMINPLUS:    case OP_NOTMINPLUS:
714    case OP_NOTQUERY:    case OP_NOTQUERY:
715    case OP_NOTMINQUERY:    case OP_NOTMINQUERY:
716      case OP_NOTUPTO:
717      case OP_NOTMINUPTO:
718      case OP_NOTEXACT:
719    case OP_NOTPOSSTAR:    case OP_NOTPOSSTAR:
720    case OP_NOTPOSPLUS:    case OP_NOTPOSPLUS:
721    case OP_NOTPOSQUERY:    case OP_NOTPOSQUERY:
722      case OP_NOTPOSUPTO:
723    case OP_NOTSTARI:    case OP_NOTSTARI:
724    case OP_NOTMINSTARI:    case OP_NOTMINSTARI:
725    case OP_NOTPLUSI:    case OP_NOTPLUSI:
726    case OP_NOTMINPLUSI:    case OP_NOTMINPLUSI:
727    case OP_NOTQUERYI:    case OP_NOTQUERYI:
728    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:  
729    case OP_NOTUPTOI:    case OP_NOTUPTOI:
730    case OP_NOTMINUPTOI:    case OP_NOTMINUPTOI:
731    case OP_NOTEXACTI:    case OP_NOTEXACTI:
732      case OP_NOTPOSSTARI:
733      case OP_NOTPOSPLUSI:
734      case OP_NOTPOSQUERYI:
735    case OP_NOTPOSUPTOI:    case OP_NOTPOSUPTOI:
736    cc += 2 + IMM2_SIZE;    cc += PRIV(OP_lengths)[*cc];
737  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
738    if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);    if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
739  #endif  #endif
740    return cc;    return cc;
741    
742    case OP_NOTPROP:    /* Special cases. */
743    case OP_PROP:    case OP_TYPESTAR:
744    return cc + 1 + 2;    case OP_TYPEMINSTAR:
745      case OP_TYPEPLUS:
746      case OP_TYPEMINPLUS:
747      case OP_TYPEQUERY:
748      case OP_TYPEMINQUERY:
749    case OP_TYPEUPTO:    case OP_TYPEUPTO:
750    case OP_TYPEMINUPTO:    case OP_TYPEMINUPTO:
751    case OP_TYPEEXACT:    case OP_TYPEEXACT:
752      case OP_TYPEPOSSTAR:
753      case OP_TYPEPOSPLUS:
754      case OP_TYPEPOSQUERY:
755    case OP_TYPEPOSUPTO:    case OP_TYPEPOSUPTO:
756    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;  
757    
758    case OP_CRRANGE:    case OP_ANYBYTE:
759    case OP_CRMINRANGE:  #ifdef SUPPORT_UTF
760    return cc + 1 + 2 * IMM2_SIZE;    if (common->utf) return NULL;
761    #endif
762    case OP_CLASS:    return cc + 1;
   case OP_NCLASS:  
   return cc + 1 + 32 / sizeof(pcre_uchar);  
763    
764  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
765    case OP_XCLASS:    case OP_XCLASS:
766    return cc + GET(cc, 1);    return cc + GET(cc, 1);
767  #endif  #endif
768    
   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;  
   
769    case OP_MARK:    case OP_MARK:
770      case OP_PRUNE_ARG:
771      case OP_SKIP_ARG:
772      case OP_THEN_ARG:
773    return cc + 1 + 2 + cc[1];    return cc + 1 + 2 + cc[1];
774    
775    default:    default:
776      /* All opcodes are supported now! */
777      SLJIT_ASSERT_STOP();
778    return NULL;    return NULL;
779    }    }
780  }  }
781    
782  #define CASE_ITERATOR_PRIVATE_DATA_1 \  static BOOL check_opcode_types(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend)
783      case OP_MINSTAR: \  {
784      case OP_MINPLUS: \  int count;
785      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:  
786    
787  #define CASE_ITERATOR_PRIVATE_DATA_2A \  /* Calculate important variables (like stack size) and checks whether all opcodes are supported. */
788      case OP_STAR: \  while (cc < ccend)
789      case OP_PLUS: \    {
790      case OP_STARI: \    switch(*cc)
791      case OP_PLUSI: \      {
792      case OP_NOTSTAR: \      case OP_SET_SOM:
793      case OP_NOTPLUS: \      common->has_set_som = TRUE;
794      case OP_NOTSTARI: \      common->might_be_empty = TRUE;
795      case OP_NOTPLUSI:      cc += 1;
796        break;
797    
798  #define CASE_ITERATOR_PRIVATE_DATA_2B \      case OP_REF:
799      case OP_UPTO: \      case OP_REFI:
800      case OP_MINUPTO: \      common->optimized_cbracket[GET2(cc, 1)] = 0;
801      case OP_UPTOI: \      cc += 1 + IMM2_SIZE;
802      case OP_MINUPTOI: \      break;
     case OP_NOTUPTO: \  
     case OP_NOTMINUPTO: \  
     case OP_NOTUPTOI: \  
     case OP_NOTMINUPTOI:  
803    
804  #define CASE_ITERATOR_TYPE_PRIVATE_DATA_1 \      case OP_BRA:
805      case OP_TYPEMINSTAR: \      case OP_CBRA:
806      case OP_TYPEMINPLUS: \      case OP_SBRA:
807      case OP_TYPEQUERY: \      case OP_SCBRA:
808      case OP_TYPEMINQUERY:      count = no_alternatives(cc);
809        if (count > 4)
810          common->read_only_data_size += count * sizeof(sljit_uw);
811        cc += 1 + LINK_SIZE + (*cc == OP_CBRA || *cc == OP_SCBRA ? IMM2_SIZE : 0);
812        break;
813    
814  #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2A \      case OP_CBRAPOS:
815      case OP_TYPESTAR: \      case OP_SCBRAPOS:
816      case OP_TYPEPLUS:      common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] = 0;
817        cc += 1 + LINK_SIZE + IMM2_SIZE;
818        break;
819    
820  #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2B \      case OP_COND:
821      case OP_TYPEUPTO: \      case OP_SCOND:
822      case OP_TYPEMINUPTO:      /* Only AUTO_CALLOUT can insert this opcode. We do
823           not intend to support this case. */
824        if (cc[1 + LINK_SIZE] == OP_CALLOUT)
825          return FALSE;
826        cc += 1 + LINK_SIZE;
827        break;
828    
829  static int get_class_iterator_size(pcre_uchar *cc)      case OP_CREF:
830  {      common->optimized_cbracket[GET2(cc, 1)] = 0;
831        cc += 1 + IMM2_SIZE;
832        break;
833    
834        case OP_DNREF:
835        case OP_DNREFI:
836        case OP_DNCREF:
837        count = GET2(cc, 1 + IMM2_SIZE);
838        slot = common->name_table + GET2(cc, 1) * common->name_entry_size;
839        while (count-- > 0)
840          {
841          common->optimized_cbracket[GET2(slot, 0)] = 0;
842          slot += common->name_entry_size;
843          }
844        cc += 1 + 2 * IMM2_SIZE;
845        break;
846    
847        case OP_RECURSE:
848        /* Set its value only once. */
849        if (common->recursive_head_ptr == 0)
850          {
851          common->recursive_head_ptr = common->ovector_start;
852          common->ovector_start += sizeof(sljit_sw);
853          }
854        cc += 1 + LINK_SIZE;
855        break;
856    
857        case OP_CALLOUT:
858        if (common->capture_last_ptr == 0)
859          {
860          common->capture_last_ptr = common->ovector_start;
861          common->ovector_start += sizeof(sljit_sw);
862          }
863        cc += 2 + 2 * LINK_SIZE;
864        break;
865    
866        case OP_THEN_ARG:
867        common->has_then = TRUE;
868        common->control_head_ptr = 1;
869        /* Fall through. */
870    
871        case OP_PRUNE_ARG:
872        common->needs_start_ptr = TRUE;
873        /* Fall through. */
874    
875        case OP_MARK:
876        if (common->mark_ptr == 0)
877          {
878          common->mark_ptr = common->ovector_start;
879          common->ovector_start += sizeof(sljit_sw);
880          }
881        cc += 1 + 2 + cc[1];
882        break;
883    
884        case OP_THEN:
885        common->has_then = TRUE;
886        common->control_head_ptr = 1;
887        /* Fall through. */
888    
889        case OP_PRUNE:
890        case OP_SKIP:
891        common->needs_start_ptr = TRUE;
892        cc += 1;
893        break;
894    
895        case OP_SKIP_ARG:
896        common->control_head_ptr = 1;
897        common->has_skip_arg = TRUE;
898        cc += 1 + 2 + cc[1];
899        break;
900    
901        default:
902        cc = next_opcode(common, cc);
903        if (cc == NULL)
904          return FALSE;
905        break;
906        }
907      }
908    return TRUE;
909    }
910    
911    static int get_class_iterator_size(pcre_uchar *cc)
912    {
913  switch(*cc)  switch(*cc)
914    {    {
915    case OP_CRSTAR:    case OP_CRSTAR:
# Line 762  switch(*cc) Line 933  switch(*cc)
933    }    }
934  }  }
935    
936  static int get_private_data_length(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend)  static BOOL detect_repeat(compiler_common *common, pcre_uchar *begin)
937  {  {
938  int private_data_length = 0;  pcre_uchar *end = bracketend(begin);
939  pcre_uchar *alternative;  pcre_uchar *next;
940  pcre_uchar *name;  pcre_uchar *next_end;
941  pcre_uchar *end = NULL;  pcre_uchar *max_end;
942  int space, size, i;  pcre_uchar type;
943  pcre_uint32 bracketlen;  sljit_sw length = end - begin;
944    int min, max, i;
945    
946  /* Calculate important variables (like stack size) and checks whether all opcodes are supported. */  /* Detect fixed iterations first. */
947  while (cc < ccend)  if (end[-(1 + LINK_SIZE)] != OP_KET)
948      return FALSE;
949    
950    /* Already detected repeat. */
951    if (common->private_data_ptrs[end - common->start - LINK_SIZE] != 0)
952      return TRUE;
953    
954    next = end;
955    min = 1;
956    while (1)
957    {    {
958    space = 0;    if (*next != *begin)
   size = 0;  
   bracketlen = 0;  
   switch(*cc)  
     {  
     case OP_SET_SOM:  
     common->has_set_som = TRUE;  
     cc += 1;  
959      break;      break;
960      next_end = bracketend(next);
961      case OP_REF:    if (next_end - next != length || memcmp(begin, next, IN_UCHARS(length)) != 0)
     case OP_REFI:  
     common->optimized_cbracket[GET2(cc, 1)] = 0;  
     cc += 1 + IMM2_SIZE;  
962      break;      break;
963      next = next_end;
964      min++;
965      }
966    
967      case OP_ASSERT:  if (min == 2)
968      case OP_ASSERT_NOT:    return FALSE;
     case OP_ASSERTBACK:  
     case OP_ASSERTBACK_NOT:  
     case OP_ONCE:  
     case OP_ONCE_NC:  
     case OP_BRAPOS:  
     case OP_SBRA:  
     case OP_SBRAPOS:  
     private_data_length += sizeof(sljit_sw);  
     bracketlen = 1 + LINK_SIZE;  
     break;  
969    
970      case OP_CBRAPOS:  max = 0;
971      case OP_SCBRAPOS:  max_end = next;
972      private_data_length += sizeof(sljit_sw);  if (*next == OP_BRAZERO || *next == OP_BRAMINZERO)
973      common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] = 0;    {
974      bracketlen = 1 + LINK_SIZE + IMM2_SIZE;    type = *next;
975      break;    while (1)
976        {
977        if (next[0] != type || next[1] != OP_BRA || next[2 + LINK_SIZE] != *begin)
978          break;
979        next_end = bracketend(next + 2 + LINK_SIZE);
980        if (next_end - next != (length + 2 + LINK_SIZE) || memcmp(begin, next + 2 + LINK_SIZE, IN_UCHARS(length)) != 0)
981          break;
982        next = next_end;
983        max++;
984        }
985    
986      case OP_COND:    if (next[0] == type && next[1] == *begin && max >= 1)
987      case OP_SCOND:      {
988      bracketlen = cc[1 + LINK_SIZE];      next_end = bracketend(next + 1);
989      if (bracketlen == OP_CREF)      if (next_end - next == (length + 1) && memcmp(begin, next + 1, IN_UCHARS(length)) == 0)
       {  
       bracketlen = GET2(cc, 1 + LINK_SIZE + 1);  
       common->optimized_cbracket[bracketlen] = 0;  
       }  
     else if (bracketlen == OP_NCREF)  
990        {        {
991        bracketlen = GET2(cc, 1 + LINK_SIZE + 1);        for (i = 0; i < max; i++, next_end += 1 + LINK_SIZE)
992        name = (pcre_uchar *)common->name_table;          if (*next_end != OP_KET)
993        alternative = name;            break;
       for (i = 0; i < common->name_count; i++)  
         {  
         if (GET2(name, 0) == bracketlen) break;  
         name += common->name_entry_size;  
         }  
       SLJIT_ASSERT(i != common->name_count);  
994    
995        for (i = 0; i < common->name_count; i++)        if (i == max)
996          {          {
997          if (STRCMP_UC_UC(alternative + IMM2_SIZE, name + IMM2_SIZE) == 0)          common->private_data_ptrs[max_end - common->start - LINK_SIZE] = next_end - max_end;
998            common->optimized_cbracket[GET2(alternative, 0)] = 0;          common->private_data_ptrs[max_end - common->start - LINK_SIZE + 1] = (type == OP_BRAZERO) ? OP_UPTO : OP_MINUPTO;
999          alternative += common->name_entry_size;          /* +2 the original and the last. */
1000            common->private_data_ptrs[max_end - common->start - LINK_SIZE + 2] = max + 2;
1001            if (min == 1)
1002              return TRUE;
1003            min--;
1004            max_end -= (1 + LINK_SIZE) + GET(max_end, -LINK_SIZE);
1005          }          }
1006        }        }
1007        }
1008      }
1009    
1010      if (*cc == OP_COND)  if (min >= 3)
1011        {    {
1012        /* Might be a hidden SCOND. */    common->private_data_ptrs[end - common->start - LINK_SIZE] = max_end - end;
1013        alternative = cc + GET(cc, 1);    common->private_data_ptrs[end - common->start - LINK_SIZE + 1] = OP_EXACT;
1014        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)    common->private_data_ptrs[end - common->start - LINK_SIZE + 2] = min;
1015          private_data_length += sizeof(sljit_sw);    return TRUE;
1016        }    }
     else  
       private_data_length += sizeof(sljit_sw);  
     bracketlen = 1 + LINK_SIZE;  
     break;  
   
     case OP_BRA:  
     bracketlen = 1 + LINK_SIZE;  
     break;  
   
     case OP_CBRA:  
     case OP_SCBRA:  
     bracketlen = 1 + LINK_SIZE + IMM2_SIZE;  
     break;  
   
     CASE_ITERATOR_PRIVATE_DATA_1  
     space = 1;  
     size = -2;  
     break;  
   
     CASE_ITERATOR_PRIVATE_DATA_2A  
     space = 2;  
     size = -2;  
     break;  
   
     CASE_ITERATOR_PRIVATE_DATA_2B  
     space = 2;  
     size = -(2 + IMM2_SIZE);  
     break;  
   
     CASE_ITERATOR_TYPE_PRIVATE_DATA_1  
     space = 1;  
     size = 1;  
     break;  
   
     CASE_ITERATOR_TYPE_PRIVATE_DATA_2A  
     if (cc[1] != OP_ANYNL && cc[1] != OP_EXTUNI)  
       space = 2;  
     size = 1;  
     break;  
   
     CASE_ITERATOR_TYPE_PRIVATE_DATA_2B  
     if (cc[1 + IMM2_SIZE] != OP_ANYNL && cc[1 + IMM2_SIZE] != OP_EXTUNI)  
       space = 2;  
     size = 1 + IMM2_SIZE;  
     break;  
   
     case OP_CLASS:  
     case OP_NCLASS:  
     size += 1 + 32 / sizeof(pcre_uchar);  
     space = get_class_iterator_size(cc + size);  
     break;  
1017    
1018  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  return FALSE;
1019      case OP_XCLASS:  }
     size = GET(cc, 1);  
     space = get_class_iterator_size(cc + size);  
     break;  
 #endif  
1020    
1021      case OP_RECURSE:  #define CASE_ITERATOR_PRIVATE_DATA_1 \
1022      /* Set its value only once. */      case OP_MINSTAR: \
1023      if (common->recursive_head == 0)      case OP_MINPLUS: \
1024        {      case OP_QUERY: \
1025        common->recursive_head = common->ovector_start;      case OP_MINQUERY: \
1026        common->ovector_start += sizeof(sljit_sw);      case OP_MINSTARI: \
1027        }      case OP_MINPLUSI: \
1028      cc += 1 + LINK_SIZE;      case OP_QUERYI: \
1029      break;      case OP_MINQUERYI: \
1030        case OP_NOTMINSTAR: \
1031        case OP_NOTMINPLUS: \
1032        case OP_NOTQUERY: \
1033        case OP_NOTMINQUERY: \
1034        case OP_NOTMINSTARI: \
1035        case OP_NOTMINPLUSI: \
1036        case OP_NOTQUERYI: \
1037        case OP_NOTMINQUERYI:
1038    
1039      case OP_MARK:  #define CASE_ITERATOR_PRIVATE_DATA_2A \
1040      if (common->mark_ptr == 0)      case OP_STAR: \
1041        {      case OP_PLUS: \
1042        common->mark_ptr = common->ovector_start;      case OP_STARI: \
1043        common->ovector_start += sizeof(sljit_sw);      case OP_PLUSI: \
1044        }      case OP_NOTSTAR: \
1045      cc += 1 + 2 + cc[1];      case OP_NOTPLUS: \
1046      break;      case OP_NOTSTARI: \
1047        case OP_NOTPLUSI:
1048    
1049      default:  #define CASE_ITERATOR_PRIVATE_DATA_2B \
1050      cc = next_opcode(common, cc);      case OP_UPTO: \
1051      if (cc == NULL)      case OP_MINUPTO: \
1052        return -1;      case OP_UPTOI: \
1053      break;      case OP_MINUPTOI: \
1054      }      case OP_NOTUPTO: \
1055        case OP_NOTMINUPTO: \
1056        case OP_NOTUPTOI: \
1057        case OP_NOTMINUPTOI:
1058    
1059    if (space > 0 && cc >= end)  #define CASE_ITERATOR_TYPE_PRIVATE_DATA_1 \
1060      private_data_length += sizeof(sljit_sw) * space;      case OP_TYPEMINSTAR: \
1061        case OP_TYPEMINPLUS: \
1062        case OP_TYPEQUERY: \
1063        case OP_TYPEMINQUERY:
1064    
1065    if (size != 0)  #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2A \
1066      {      case OP_TYPESTAR: \
1067      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;  
     }  
1068    
1069    if (bracketlen != 0)  #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2B \
1070      {      case OP_TYPEUPTO: \
1071      if (cc >= end)      case OP_TYPEMINUPTO:
       {  
       end = bracketend(cc);  
       if (end[-1 - LINK_SIZE] == OP_KET)  
         end = NULL;  
       }  
     cc += bracketlen;  
     }  
   }  
 return private_data_length;  
 }  
1072    
1073  static void set_private_data_ptrs(compiler_common *common, int private_data_ptr, pcre_uchar *ccend)  static void set_private_data_ptrs(compiler_common *common, int *private_data_start, pcre_uchar *ccend)
1074  {  {
1075  pcre_uchar *cc = common->start;  pcre_uchar *cc = common->start;
1076  pcre_uchar *alternative;  pcre_uchar *alternative;
1077  pcre_uchar *end = NULL;  pcre_uchar *end = NULL;
1078    int private_data_ptr = *private_data_start;
1079  int space, size, bracketlen;  int space, size, bracketlen;
1080    
1081  while (cc < ccend)  while (cc < ccend)
# Line 972  while (cc < ccend) Line 1083  while (cc < ccend)
1083    space = 0;    space = 0;
1084    size = 0;    size = 0;
1085    bracketlen = 0;    bracketlen = 0;
1086      if (private_data_ptr > SLJIT_MAX_LOCAL_SIZE)
1087        return;
1088    
1089      if (*cc == OP_ONCE || *cc == OP_ONCE_NC || *cc == OP_BRA || *cc == OP_CBRA || *cc == OP_COND)
1090        if (detect_repeat(common, cc))
1091          {
1092          /* These brackets are converted to repeats, so no global
1093          based single character repeat is allowed. */
1094          if (cc >= end)
1095            end = bracketend(cc);
1096          }
1097    
1098    switch(*cc)    switch(*cc)
1099      {      {
1100        case OP_KET:
1101        if (common->private_data_ptrs[cc + 1 - common->start] != 0)
1102          {
1103          common->private_data_ptrs[cc - common->start] = private_data_ptr;
1104          private_data_ptr += sizeof(sljit_sw);
1105          cc += common->private_data_ptrs[cc + 1 - common->start];
1106          }
1107        cc += 1 + LINK_SIZE;
1108        break;
1109    
1110      case OP_ASSERT:      case OP_ASSERT:
1111      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1112      case OP_ASSERTBACK:      case OP_ASSERTBACK:
# Line 1067  while (cc < ccend) Line 1200  while (cc < ccend)
1200      break;      break;
1201      }      }
1202    
1203      /* Character iterators, which are not inside a repeated bracket,
1204         gets a private slot instead of allocating it on the stack. */
1205    if (space > 0 && cc >= end)    if (space > 0 && cc >= end)
1206      {      {
1207      common->private_data_ptrs[cc - common->start] = private_data_ptr;      common->private_data_ptrs[cc - common->start] = private_data_ptr;
# Line 1097  while (cc < ccend) Line 1232  while (cc < ccend)
1232      cc += bracketlen;      cc += bracketlen;
1233      }      }
1234    }    }
1235    *private_data_start = private_data_ptr;
1236  }  }
1237    
1238  /* Returns with -1 if no need for frame. */  /* Returns with a frame_types (always < 0) if no need for frame. */
1239  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)
1240  {  {
 pcre_uchar *ccend = bracketend(cc);  
1241  int length = 0;  int length = 0;
1242  BOOL possessive = FALSE;  int possessive = 0;
1243    BOOL stack_restore = FALSE;
1244  BOOL setsom_found = recursive;  BOOL setsom_found = recursive;
1245  BOOL setmark_found = recursive;  BOOL setmark_found = recursive;
1246    /* The last capture is a local variable even for recursions. */
1247    BOOL capture_last_found = FALSE;
1248    
1249    #if defined DEBUG_FORCE_CONTROL_HEAD && DEBUG_FORCE_CONTROL_HEAD
1250    SLJIT_ASSERT(common->control_head_ptr != 0);
1251    *needs_control_head = TRUE;
1252    #else
1253    *needs_control_head = FALSE;
1254    #endif
1255    
1256  if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))  if (ccend == NULL)
1257    {    {
1258    length = 3;    ccend = bracketend(cc) - (1 + LINK_SIZE);
1259    possessive = TRUE;    if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))
1260        {
1261        possessive = length = (common->capture_last_ptr != 0) ? 5 : 3;
1262        /* This is correct regardless of common->capture_last_ptr. */
1263        capture_last_found = TRUE;
1264        }
1265      cc = next_opcode(common, cc);
1266    }    }
1267    
 cc = next_opcode(common, cc);  
1268  SLJIT_ASSERT(cc != NULL);  SLJIT_ASSERT(cc != NULL);
1269  while (cc < ccend)  while (cc < ccend)
1270    switch(*cc)    switch(*cc)
1271      {      {
1272      case OP_SET_SOM:      case OP_SET_SOM:
1273      SLJIT_ASSERT(common->has_set_som);      SLJIT_ASSERT(common->has_set_som);
1274        stack_restore = TRUE;
1275      if (!setsom_found)      if (!setsom_found)
1276        {        {
1277        length += 2;        length += 2;
# Line 1130  while (cc < ccend) Line 1281  while (cc < ccend)
1281      break;      break;
1282    
1283      case OP_MARK:      case OP_MARK:
1284        case OP_PRUNE_ARG:
1285        case OP_THEN_ARG:
1286      SLJIT_ASSERT(common->mark_ptr != 0);      SLJIT_ASSERT(common->mark_ptr != 0);
1287        stack_restore = TRUE;
1288      if (!setmark_found)      if (!setmark_found)
1289        {        {
1290        length += 2;        length += 2;
1291        setmark_found = TRUE;        setmark_found = TRUE;
1292        }        }
1293        if (common->control_head_ptr != 0)
1294          *needs_control_head = TRUE;
1295      cc += 1 + 2 + cc[1];      cc += 1 + 2 + cc[1];
1296      break;      break;
1297    
1298      case OP_RECURSE:      case OP_RECURSE:
1299        stack_restore = TRUE;
1300      if (common->has_set_som && !setsom_found)      if (common->has_set_som && !setsom_found)
1301        {        {
1302        length += 2;        length += 2;
# Line 1150  while (cc < ccend) Line 1307  while (cc < ccend)
1307        length += 2;        length += 2;
1308        setmark_found = TRUE;        setmark_found = TRUE;
1309        }        }
1310        if (common->capture_last_ptr != 0 && !capture_last_found)
1311          {
1312          length += 2;
1313          capture_last_found = TRUE;
1314          }
1315      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1316      break;      break;
1317    
# Line 1157  while (cc < ccend) Line 1319  while (cc < ccend)
1319      case OP_CBRAPOS:      case OP_CBRAPOS:
1320      case OP_SCBRA:      case OP_SCBRA:
1321      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1322        stack_restore = TRUE;
1323        if (common->capture_last_ptr != 0 && !capture_last_found)
1324          {
1325          length += 2;
1326          capture_last_found = TRUE;
1327          }
1328      length += 3;      length += 3;
1329      cc += 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1330      break;      break;
1331    
1332      default:      default:
1333        stack_restore = TRUE;
1334        /* Fall through. */
1335    
1336        case OP_NOT_WORD_BOUNDARY:
1337        case OP_WORD_BOUNDARY:
1338        case OP_NOT_DIGIT:
1339        case OP_DIGIT:
1340        case OP_NOT_WHITESPACE:
1341        case OP_WHITESPACE:
1342        case OP_NOT_WORDCHAR:
1343        case OP_WORDCHAR:
1344        case OP_ANY:
1345        case OP_ALLANY:
1346        case OP_ANYBYTE:
1347        case OP_NOTPROP:
1348        case OP_PROP:
1349        case OP_ANYNL:
1350        case OP_NOT_HSPACE:
1351        case OP_HSPACE:
1352        case OP_NOT_VSPACE:
1353        case OP_VSPACE:
1354        case OP_EXTUNI:
1355        case OP_EODN:
1356        case OP_EOD:
1357        case OP_CIRC:
1358        case OP_CIRCM:
1359        case OP_DOLL:
1360        case OP_DOLLM:
1361        case OP_CHAR:
1362        case OP_CHARI:
1363        case OP_NOT:
1364        case OP_NOTI:
1365    
1366        case OP_EXACT:
1367        case OP_POSSTAR:
1368        case OP_POSPLUS:
1369        case OP_POSQUERY:
1370        case OP_POSUPTO:
1371    
1372        case OP_EXACTI:
1373        case OP_POSSTARI:
1374        case OP_POSPLUSI:
1375        case OP_POSQUERYI:
1376        case OP_POSUPTOI:
1377    
1378        case OP_NOTEXACT:
1379        case OP_NOTPOSSTAR:
1380        case OP_NOTPOSPLUS:
1381        case OP_NOTPOSQUERY:
1382        case OP_NOTPOSUPTO:
1383    
1384        case OP_NOTEXACTI:
1385        case OP_NOTPOSSTARI:
1386        case OP_NOTPOSPLUSI:
1387        case OP_NOTPOSQUERYI:
1388        case OP_NOTPOSUPTOI:
1389    
1390        case OP_TYPEEXACT:
1391        case OP_TYPEPOSSTAR:
1392        case OP_TYPEPOSPLUS:
1393        case OP_TYPEPOSQUERY:
1394        case OP_TYPEPOSUPTO:
1395    
1396        case OP_CLASS:
1397        case OP_NCLASS:
1398        case OP_XCLASS:
1399    
1400      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
1401      SLJIT_ASSERT(cc != NULL);      SLJIT_ASSERT(cc != NULL);
1402      break;      break;
1403      }      }
1404    
1405  /* Possessive quantifiers can use a special case. */  /* Possessive quantifiers can use a special case. */
1406  if (SLJIT_UNLIKELY(possessive) && length == 3)  if (SLJIT_UNLIKELY(possessive == length))
1407    return -1;    return stack_restore ? no_frame : no_stack;
1408    
1409  if (length > 0)  if (length > 0)
1410    return length + 1;    return length + 1;
1411  return -1;  return stack_restore ? no_frame : no_stack;
1412  }  }
1413    
1414  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)
1415  {  {
1416  DEFINE_COMPILER;  DEFINE_COMPILER;
 pcre_uchar *ccend = bracketend(cc);  
1417  BOOL setsom_found = recursive;  BOOL setsom_found = recursive;
1418  BOOL setmark_found = recursive;  BOOL setmark_found = recursive;
1419    /* The last capture is a local variable even for recursions. */
1420    BOOL capture_last_found = FALSE;
1421  int offset;  int offset;
1422    
1423  /* >= 1 + shortest item size (2) */  /* >= 1 + shortest item size (2) */
# Line 1189  SLJIT_UNUSED_ARG(stacktop); Line 1425  SLJIT_UNUSED_ARG(stacktop);
1425  SLJIT_ASSERT(stackpos >= stacktop + 2);  SLJIT_ASSERT(stackpos >= stacktop + 2);
1426    
1427  stackpos = STACK(stackpos);  stackpos = STACK(stackpos);
1428  if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))  if (ccend == NULL)
1429    cc = next_opcode(common, cc);    {
1430      ccend = bracketend(cc) - (1 + LINK_SIZE);
1431      if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))
1432        cc = next_opcode(common, cc);
1433      }
1434    
1435  SLJIT_ASSERT(cc != NULL);  SLJIT_ASSERT(cc != NULL);
1436  while (cc < ccend)  while (cc < ccend)
1437    switch(*cc)    switch(*cc)
# Line 1200  while (cc < ccend) Line 1441  while (cc < ccend)
1441      if (!setsom_found)      if (!setsom_found)
1442        {        {
1443        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));
1444        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setstrbegin);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1445        stackpos += (int)sizeof(sljit_sw);        stackpos += (int)sizeof(sljit_sw);
1446        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1447        stackpos += (int)sizeof(sljit_sw);        stackpos += (int)sizeof(sljit_sw);
# Line 1210  while (cc < ccend) Line 1451  while (cc < ccend)
1451      break;      break;
1452    
1453      case OP_MARK:      case OP_MARK:
1454        case OP_PRUNE_ARG:
1455        case OP_THEN_ARG:
1456      SLJIT_ASSERT(common->mark_ptr != 0);      SLJIT_ASSERT(common->mark_ptr != 0);
1457      if (!setmark_found)      if (!setmark_found)
1458        {        {
1459        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);
1460        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);
1461        stackpos += (int)sizeof(sljit_sw);        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_sw);        stackpos += (int)sizeof(sljit_sw);
# Line 1227  while (cc < ccend) Line 1470  while (cc < ccend)
1470      if (common->has_set_som && !setsom_found)      if (common->has_set_som && !setsom_found)
1471        {        {
1472        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));
1473        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setstrbegin);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1474        stackpos += (int)sizeof(sljit_sw);        stackpos += (int)sizeof(sljit_sw);
1475        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1476        stackpos += (int)sizeof(sljit_sw);        stackpos += (int)sizeof(sljit_sw);
# Line 1236  while (cc < ccend) Line 1479  while (cc < ccend)
1479      if (common->mark_ptr != 0 && !setmark_found)      if (common->mark_ptr != 0 && !setmark_found)
1480        {        {
1481        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);
1482        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);
1483        stackpos += (int)sizeof(sljit_sw);        stackpos += (int)sizeof(sljit_sw);
1484        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1485        stackpos += (int)sizeof(sljit_sw);        stackpos += (int)sizeof(sljit_sw);
1486        setmark_found = TRUE;        setmark_found = TRUE;
1487        }        }
1488        if (common->capture_last_ptr != 0 && !capture_last_found)
1489          {
1490          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->capture_last_ptr);
1491          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1492          stackpos += (int)sizeof(sljit_sw);
1493          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1494          stackpos += (int)sizeof(sljit_sw);
1495          capture_last_found = TRUE;
1496          }
1497      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1498      break;      break;
1499    
# Line 1249  while (cc < ccend) Line 1501  while (cc < ccend)
1501      case OP_CBRAPOS:      case OP_CBRAPOS:
1502      case OP_SCBRA:      case OP_SCBRA:
1503      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1504        if (common->capture_last_ptr != 0 && !capture_last_found)
1505          {
1506          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->capture_last_ptr);
1507          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1508          stackpos += (int)sizeof(sljit_sw);
1509          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1510          stackpos += (int)sizeof(sljit_sw);
1511          capture_last_found = TRUE;
1512          }
1513      offset = (GET2(cc, 1 + LINK_SIZE)) << 1;      offset = (GET2(cc, 1 + LINK_SIZE)) << 1;
1514      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));
1515      stackpos += (int)sizeof(sljit_sw);      stackpos += (int)sizeof(sljit_sw);
# Line 1268  while (cc < ccend) Line 1529  while (cc < ccend)
1529      break;      break;
1530      }      }
1531    
1532  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_end);  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, 0);
1533  SLJIT_ASSERT(stackpos == STACK(stacktop));  SLJIT_ASSERT(stackpos == STACK(stacktop));
1534  }  }
1535    
1536  static SLJIT_INLINE int get_private_data_length_for_copy(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)
1537  {  {
1538  int private_data_length = 2;  int private_data_length = needs_control_head ? 3 : 2;
1539  int size;  int size;
1540  pcre_uchar *alternative;  pcre_uchar *alternative;
1541  /* Calculate the sum of the private machine words. */  /* Calculate the sum of the private machine words. */
# Line 1283  while (cc < ccend) Line 1544  while (cc < ccend)
1544    size = 0;    size = 0;
1545    switch(*cc)    switch(*cc)
1546      {      {
1547        case OP_KET:
1548        if (PRIVATE_DATA(cc) != 0)
1549          private_data_length++;
1550        cc += 1 + LINK_SIZE;
1551        break;
1552    
1553      case OP_ASSERT:      case OP_ASSERT:
1554      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1555      case OP_ASSERTBACK:      case OP_ASSERTBACK:
# Line 1387  return private_data_length; Line 1654  return private_data_length;
1654  }  }
1655    
1656  static void copy_private_data(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend,  static void copy_private_data(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend,
1657    BOOL save, int stackptr, int stacktop)    BOOL save, int stackptr, int stacktop, BOOL needs_control_head)
1658  {  {
1659  DEFINE_COMPILER;  DEFINE_COMPILER;
1660  int srcw[2];  int srcw[2];
# Line 1408  stacktop = STACK(stacktop - 1); Line 1675  stacktop = STACK(stacktop - 1);
1675    
1676  if (!save)  if (!save)
1677    {    {
1678    stackptr += sizeof(sljit_sw);    stackptr += (needs_control_head ? 2 : 1) * sizeof(sljit_sw);
1679    if (stackptr < stacktop)    if (stackptr < stacktop)
1680      {      {
1681      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
# Line 1424  if (!save) Line 1691  if (!save)
1691    /* The tmp1next must be TRUE in either way. */    /* The tmp1next must be TRUE in either way. */
1692    }    }
1693    
1694  while (status != end)  do
1695    {    {
1696    count = 0;    count = 0;
1697    switch(status)    switch(status)
1698      {      {
1699      case start:      case start:
1700      SLJIT_ASSERT(save && common->recursive_head != 0);      SLJIT_ASSERT(save && common->recursive_head_ptr != 0);
1701      count = 1;      count = 1;
1702      srcw[0] = common->recursive_head;      srcw[0] = common->recursive_head_ptr;
1703        if (needs_control_head)
1704          {
1705          SLJIT_ASSERT(common->control_head_ptr != 0);
1706          count = 2;
1707          srcw[1] = common->control_head_ptr;
1708          }
1709      status = loop;      status = loop;
1710      break;      break;
1711    
# Line 1445  while (status != end) Line 1718  while (status != end)
1718    
1719      switch(*cc)      switch(*cc)
1720        {        {
1721          case OP_KET:
1722          if (PRIVATE_DATA(cc) != 0)
1723            {
1724            count = 1;
1725            srcw[0] = PRIVATE_DATA(cc);
1726            }
1727          cc += 1 + LINK_SIZE;
1728          break;
1729    
1730        case OP_ASSERT:        case OP_ASSERT:
1731        case OP_ASSERT_NOT:        case OP_ASSERT_NOT:
1732        case OP_ASSERTBACK:        case OP_ASSERTBACK:
# Line 1657  while (status != end) Line 1939  while (status != end)
1939        }        }
1940      }      }
1941    }    }
1942    while (status != end);
1943    
1944  if (save)  if (save)
1945    {    {
# Line 1690  if (save) Line 1973  if (save)
1973  SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));  SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));
1974  }  }
1975    
1976    static SLJIT_INLINE pcre_uchar *set_then_offsets(compiler_common *common, pcre_uchar *cc, pcre_uint8 *current_offset)
1977    {
1978    pcre_uchar *end = bracketend(cc);
1979    BOOL has_alternatives = cc[GET(cc, 1)] == OP_ALT;
1980    
1981    /* Assert captures then. */
1982    if (*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT)
1983      current_offset = NULL;
1984    /* Conditional block does not. */
1985    if (*cc == OP_COND || *cc == OP_SCOND)
1986      has_alternatives = FALSE;
1987    
1988    cc = next_opcode(common, cc);
1989    if (has_alternatives)
1990      current_offset = common->then_offsets + (cc - common->start);
1991    
1992    while (cc < end)
1993      {
1994      if ((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND))
1995        cc = set_then_offsets(common, cc, current_offset);
1996      else
1997        {
1998        if (*cc == OP_ALT && has_alternatives)
1999          current_offset = common->then_offsets + (cc + 1 + LINK_SIZE - common->start);
2000        if (*cc >= OP_THEN && *cc <= OP_THEN_ARG && current_offset != NULL)
2001          *current_offset = 1;
2002        cc = next_opcode(common, cc);
2003        }
2004      }
2005    
2006    return end;
2007    }
2008    
2009  #undef CASE_ITERATOR_PRIVATE_DATA_1  #undef CASE_ITERATOR_PRIVATE_DATA_1
2010  #undef CASE_ITERATOR_PRIVATE_DATA_2A  #undef CASE_ITERATOR_PRIVATE_DATA_2A
2011  #undef CASE_ITERATOR_PRIVATE_DATA_2B  #undef CASE_ITERATOR_PRIVATE_DATA_2B
# Line 1708  while (list) Line 2024  while (list)
2024    {    {
2025    /* sljit_set_label is clever enough to do nothing    /* sljit_set_label is clever enough to do nothing
2026    if either the jump or the label is NULL. */    if either the jump or the label is NULL. */
2027    sljit_set_label(list->jump, label);    SET_LABEL(list->jump, label);
2028    list = list->next;    list = list->next;
2029    }    }
2030  }  }
# Line 1724  if (list_item) Line 2040  if (list_item)
2040    }    }
2041  }  }
2042    
2043  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)
2044  {  {
2045  DEFINE_COMPILER;  DEFINE_COMPILER;
2046  stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));  stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));
2047    
2048  if (list_item)  if (list_item)
2049    {    {
   list_item->type = type;  
   list_item->data = data;  
2050    list_item->start = start;    list_item->start = start;
2051    list_item->quit = LABEL();    list_item->quit = LABEL();
2052    list_item->next = common->stubs;    list_item->next = common->stubs;
# Line 1748  stub_list* list_item = common->stubs; Line 2062  stub_list* list_item = common->stubs;
2062  while (list_item)  while (list_item)
2063    {    {
2064    JUMPHERE(list_item->start);    JUMPHERE(list_item->start);
2065    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;  
     }  
2066    JUMPTO(SLJIT_JUMP, list_item->quit);    JUMPTO(SLJIT_JUMP, list_item->quit);
2067    list_item = list_item->next;    list_item = list_item->next;
2068    }    }
2069  common->stubs = NULL;  common->stubs = NULL;
2070  }  }
2071    
2072  static SLJIT_INLINE void decrease_call_count(compiler_common *common)  static void add_label_addr(compiler_common *common)
2073  {  {
2074  DEFINE_COMPILER;  DEFINE_COMPILER;
2075    label_addr_list *label_addr;
2076    
2077  OP2(SLJIT_SUB | SLJIT_SET_E, CALL_COUNT, 0, CALL_COUNT, 0, SLJIT_IMM, 1);  label_addr = sljit_alloc_memory(compiler, sizeof(label_addr_list));
2078    if (label_addr == NULL)
2079      return;
2080    label_addr->label = LABEL();
2081    label_addr->addr = common->read_only_data_ptr;
2082    label_addr->next = common->label_addrs;
2083    common->label_addrs = label_addr;
2084    common->read_only_data_ptr++;
2085    }
2086    
2087    static SLJIT_INLINE void count_match(compiler_common *common)
2088    {
2089    DEFINE_COMPILER;
2090    
2091    OP2(SLJIT_SUB | SLJIT_SET_E, COUNT_MATCH, 0, COUNT_MATCH, 0, SLJIT_IMM, 1);
2092  add_jump(compiler, &common->calllimit, JUMP(SLJIT_C_ZERO));  add_jump(compiler, &common->calllimit, JUMP(SLJIT_C_ZERO));
2093  }  }
2094    
# Line 1781  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0); Line 2105  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);
2105  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);
2106  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);
2107  #endif  #endif
2108  add_stub(common, stack_alloc, 0, CMP(SLJIT_C_GREATER, STACK_TOP, 0, STACK_LIMIT, 0));  add_stub(common, CMP(SLJIT_C_GREATER, STACK_TOP, 0, STACK_LIMIT, 0));
2109  }  }
2110    
2111  static SLJIT_INLINE void free_stack(compiler_common *common, int size)  static SLJIT_INLINE void free_stack(compiler_common *common, int size)
# Line 1795  static SLJIT_INLINE void reset_ovector(c Line 2119  static SLJIT_INLINE void reset_ovector(c
2119  DEFINE_COMPILER;  DEFINE_COMPILER;
2120  struct sljit_label *loop;  struct sljit_label *loop;
2121  int i;  int i;
2122    
2123  /* At this point we can freely use all temporary registers. */  /* At this point we can freely use all temporary registers. */
2124    SLJIT_ASSERT(length > 1);
2125  /* TMP1 returns with begin - 1. */  /* TMP1 returns with begin - 1. */
2126  OP2(SLJIT_SUB, SLJIT_SCRATCH_REG1, 0, SLJIT_MEM1(SLJIT_SAVED_REG1), SLJIT_OFFSETOF(jit_arguments, begin), SLJIT_IMM, IN_UCHARS(1));  OP2(SLJIT_SUB, SLJIT_SCRATCH_REG1, 0, SLJIT_MEM1(SLJIT_SAVED_REG1), SLJIT_OFFSETOF(jit_arguments, begin), SLJIT_IMM, IN_UCHARS(1));
2127  if (length < 8)  if (length < 8)
2128    {    {
2129    for (i = 0; i < length; i++)    for (i = 1; i < length; i++)
2130      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), SLJIT_SCRATCH_REG1, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), SLJIT_SCRATCH_REG1, 0);
2131    }    }
2132  else  else
2133    {    {
2134    GET_LOCAL_BASE(SLJIT_SCRATCH_REG2, 0, OVECTOR_START - sizeof(sljit_sw));    GET_LOCAL_BASE(SLJIT_SCRATCH_REG2, 0, OVECTOR_START);
2135    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, length);    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, length - 1);
2136    loop = LABEL();    loop = LABEL();
2137    OP1(SLJIT_MOVU, SLJIT_MEM1(SLJIT_SCRATCH_REG2), sizeof(sljit_sw), SLJIT_SCRATCH_REG1, 0);    OP1(SLJIT_MOVU, SLJIT_MEM1(SLJIT_SCRATCH_REG2), sizeof(sljit_sw), SLJIT_SCRATCH_REG1, 0);
2138    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 1);    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 1);
# Line 1814  else Line 2140  else
2140    }    }
2141  }  }
2142    
2143    static SLJIT_INLINE void do_reset_match(compiler_common *common, int length)
2144    {
2145    DEFINE_COMPILER;
2146    struct sljit_label *loop;
2147    int i;
2148    
2149    SLJIT_ASSERT(length > 1);
2150    /* OVECTOR(1) contains the "string begin - 1" constant. */
2151    if (length > 2)
2152      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));
2153    if (length < 8)
2154      {
2155      for (i = 2; i < length; i++)
2156        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), TMP1, 0);
2157      }
2158    else
2159      {
2160      GET_LOCAL_BASE(TMP2, 0, OVECTOR_START + sizeof(sljit_sw));
2161      OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_IMM, length - 2);
2162      loop = LABEL();
2163      OP1(SLJIT_MOVU, SLJIT_MEM1(TMP2), sizeof(sljit_sw), TMP1, 0);
2164      OP2(SLJIT_SUB | SLJIT_SET_E, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, 1);
2165      JUMPTO(SLJIT_C_NOT_ZERO, loop);
2166      }
2167    
2168    OP1(SLJIT_MOV, STACK_TOP, 0, ARGUMENTS, 0);
2169    if (common->mark_ptr != 0)
2170      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr, SLJIT_IMM, 0);
2171    if (common->control_head_ptr != 0)
2172      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->control_head_ptr, SLJIT_IMM, 0);
2173    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(jit_arguments, stack));
2174    OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_ptr);
2175    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(struct sljit_stack, base));
2176    }
2177    
2178    static sljit_sw SLJIT_CALL do_search_mark(sljit_sw *current, const pcre_uchar *skip_arg)
2179    {
2180    while (current != NULL)
2181      {
2182      switch (current[-2])
2183        {
2184        case type_then_trap:
2185        break;
2186    
2187        case type_mark:
2188        if (STRCMP_UC_UC(skip_arg, (pcre_uchar *)current[-3]) == 0)
2189          return current[-4];
2190        break;
2191    
2192        default:
2193        SLJIT_ASSERT_STOP();
2194        break;
2195        }
2196      current = (sljit_sw*)current[-1];
2197      }
2198    return -1;
2199    }
2200    
2201  static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)  static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)
2202  {  {
2203  DEFINE_COMPILER;  DEFINE_COMPILER;
2204  struct sljit_label *loop;  struct sljit_label *loop;
2205  struct sljit_jump *earlyexit;  struct sljit_jump *early_quit;
2206    
2207  /* At this point we can freely use all registers. */  /* At this point we can freely use all registers. */
2208  OP1(SLJIT_MOV, SLJIT_SAVED_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));  OP1(SLJIT_MOV, SLJIT_SAVED_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));
# Line 1827  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_R Line 2211  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_R
2211  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG1, 0, ARGUMENTS, 0);  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG1, 0, ARGUMENTS, 0);
2212  if (common->mark_ptr != 0)  if (common->mark_ptr != 0)
2213    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);
2214  OP1(SLJIT_MOV_SI, SLJIT_SCRATCH_REG2, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, offsetcount));  OP1(SLJIT_MOV_SI, SLJIT_SCRATCH_REG2, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, offset_count));
2215  if (common->mark_ptr != 0)  if (common->mark_ptr != 0)
2216    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, mark_ptr), SLJIT_SCRATCH_REG3, 0);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, mark_ptr), SLJIT_SCRATCH_REG3, 0);
2217  OP2(SLJIT_SUB, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));  OP2(SLJIT_SUB, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));
2218  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG1, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, begin));  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG1, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, begin));
2219  GET_LOCAL_BASE(SLJIT_SAVED_REG1, 0, OVECTOR_START);  GET_LOCAL_BASE(SLJIT_SAVED_REG1, 0, OVECTOR_START);
2220  /* Unlikely, but possible */  /* Unlikely, but possible */
2221  earlyexit = CMP(SLJIT_C_EQUAL, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, 0);  early_quit = CMP(SLJIT_C_EQUAL, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, 0);
2222  loop = LABEL();  loop = LABEL();
2223  OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, SLJIT_MEM1(SLJIT_SAVED_REG1), 0, SLJIT_SCRATCH_REG1, 0);  OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, SLJIT_MEM1(SLJIT_SAVED_REG1), 0, SLJIT_SCRATCH_REG1, 0);
2224  OP2(SLJIT_ADD, SLJIT_SAVED_REG1, 0, SLJIT_SAVED_REG1, 0, SLJIT_IMM, sizeof(sljit_sw));  OP2(SLJIT_ADD, SLJIT_SAVED_REG1, 0, SLJIT_SAVED_REG1, 0, SLJIT_IMM, sizeof(sljit_sw));
# Line 1845  OP2(SLJIT_ASHR, SLJIT_SAVED_REG2, 0, SLJ Line 2229  OP2(SLJIT_ASHR, SLJIT_SAVED_REG2, 0, SLJ
2229  OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG3), sizeof(int), SLJIT_SAVED_REG2, 0);  OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG3), sizeof(int), SLJIT_SAVED_REG2, 0);
2230  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_SCRATCH_REG2, 0, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, 1);  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_SCRATCH_REG2, 0, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, 1);
2231  JUMPTO(SLJIT_C_NOT_ZERO, loop);  JUMPTO(SLJIT_C_NOT_ZERO, loop);
2232  JUMPHERE(earlyexit);  JUMPHERE(early_quit);
2233    
2234  /* Calculate the return value, which is the maximum ovector value. */  /* Calculate the return value, which is the maximum ovector value. */
2235  if (topbracket > 1)  if (topbracket > 1)
# Line 1867  else Line 2251  else
2251  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)
2252  {  {
2253  DEFINE_COMPILER;  DEFINE_COMPILER;
2254    struct sljit_jump *jump;
2255    
2256  SLJIT_COMPILE_ASSERT(STR_END == SLJIT_SAVED_REG2, str_end_must_be_saved_reg2);  SLJIT_COMPILE_ASSERT(STR_END == SLJIT_SAVED_REG2, str_end_must_be_saved_reg2);
2257  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
2258      && (common->mode == JIT_PARTIAL_SOFT_COMPILE ? common->hit_start != 0 : common->hit_start == 0));
2259    
2260  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, ARGUMENTS, 0);  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, ARGUMENTS, 0);
2261  OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, PCRE_ERROR_PARTIAL);  OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, PCRE_ERROR_PARTIAL);
2262  OP1(SLJIT_MOV_SI, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, offsetcount));  OP1(SLJIT_MOV_SI, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, real_offset_count));
2263  CMPTO(SLJIT_C_LESS, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 2, quit);  CMPTO(SLJIT_C_SIG_LESS, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 2, quit);
2264    
2265  /* Store match begin and end. */  /* Store match begin and end. */
2266  OP1(SLJIT_MOV, SLJIT_SAVED_REG1, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, begin));  OP1(SLJIT_MOV, SLJIT_SAVED_REG1, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, begin));
2267  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, offsets));  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, offsets));
2268    
2269    jump = CMP(SLJIT_C_SIG_LESS, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 3);
2270    OP2(SLJIT_SUB, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mode == JIT_PARTIAL_HARD_COMPILE ? common->start_ptr : (common->hit_start + (int)sizeof(sljit_sw)), SLJIT_SAVED_REG1, 0);
2271    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2272    OP2(SLJIT_ASHR, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, UCHAR_SHIFT);
2273    #endif
2274    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG2), 2 * sizeof(int), SLJIT_SCRATCH_REG3, 0);
2275    JUMPHERE(jump);
2276    
2277  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mode == JIT_PARTIAL_HARD_COMPILE ? common->start_used_ptr : common->hit_start);  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mode == JIT_PARTIAL_HARD_COMPILE ? common->start_used_ptr : common->hit_start);
2278  OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, STR_END, 0, SLJIT_SAVED_REG1, 0);  OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, STR_END, 0, SLJIT_SAVED_REG1, 0);
2279  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
# Line 2040  return (bit < 256) ? ((0 << 8) | bit) : Line 2435  return (bit < 256) ? ((0 << 8) | bit) :
2435    
2436  static void check_partial(compiler_common *common, BOOL force)  static void check_partial(compiler_common *common, BOOL force)
2437  {  {
2438  /* Checks whether a partial matching is occured. Does not modify registers. */  /* Checks whether a partial matching is occurred. Does not modify registers. */
2439  DEFINE_COMPILER;  DEFINE_COMPILER;
2440  struct sljit_jump *jump = NULL;  struct sljit_jump *jump = NULL;
2441    
# Line 2055  else if (common->mode == JIT_PARTIAL_SOF Line 2450  else if (common->mode == JIT_PARTIAL_SOF
2450    jump = CMP(SLJIT_C_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, SLJIT_IMM, -1);    jump = CMP(SLJIT_C_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, SLJIT_IMM, -1);
2451    
2452  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2453    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, -1);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2454  else  else
2455    {    {
2456    if (common->partialmatchlabel != NULL)    if (common->partialmatchlabel != NULL)
# Line 2068  if (jump != NULL) Line 2463  if (jump != NULL)
2463    JUMPHERE(jump);    JUMPHERE(jump);
2464  }  }
2465    
2466  static struct sljit_jump *check_str_end(compiler_common *common)  static void check_str_end(compiler_common *common, jump_list **end_reached)
2467  {  {
2468  /* Does not affect registers. Usually used in a tight spot. */  /* Does not affect registers. Usually used in a tight spot. */
2469  DEFINE_COMPILER;  DEFINE_COMPILER;
2470  struct sljit_jump *jump;  struct sljit_jump *jump;
 struct sljit_jump *nohit;  
 struct sljit_jump *return_value;  
2471    
2472  if (common->mode == JIT_COMPILE)  if (common->mode == JIT_COMPILE)
2473    return CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);    {
2474      add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2475      return;
2476      }
2477    
2478  jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);  jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);
2479  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2480    {    {
2481    nohit = CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);    add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));
2482    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, -1);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2483    JUMPHERE(nohit);    add_jump(compiler, end_reached, JUMP(SLJIT_JUMP));
   return_value = JUMP(SLJIT_JUMP);  
2484    }    }
2485  else  else
2486    {    {
2487    return_value = CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);    add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));
2488    if (common->partialmatchlabel != NULL)    if (common->partialmatchlabel != NULL)
2489      JUMPTO(SLJIT_JUMP, common->partialmatchlabel);      JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2490    else    else
2491      add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));      add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2492    }    }
2493  JUMPHERE(jump);  JUMPHERE(jump);
 return return_value;  
2494  }  }
2495    
2496  static void detect_partial_match(compiler_common *common, jump_list **backtracks)  static void detect_partial_match(compiler_common *common, jump_list **backtracks)
# Line 2115  jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR Line 2509  jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR
2509  add_jump(compiler, backtracks, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));  add_jump(compiler, backtracks, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));
2510  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2511    {    {
2512    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, -1);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2513    add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));    add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));
2514    }    }
2515  else  else
# Line 2128  else Line 2522  else
2522  JUMPHERE(jump);  JUMPHERE(jump);
2523  }  }
2524    
2525  static void read_char(compiler_common *common)  static void peek_char(compiler_common *common, pcre_uint32 max)
2526  {  {
2527  /* Reads the character into TMP1, updates STR_PTR.  /* Reads the character into TMP1, keeps STR_PTR.
2528  Does not check STR_END. TMP2 Destroyed. */  Does not check STR_END. TMP2 Destroyed. */
2529  DEFINE_COMPILER;  DEFINE_COMPILER;
2530  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2531  struct sljit_jump *jump;  struct sljit_jump *jump;
2532  #endif  #endif
2533    
2534    SLJIT_UNUSED_ARG(max);
2535    
2536  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2537  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2538  if (common->utf)  if (common->utf)
2539    {    {
2540  #if defined COMPILE_PCRE8    if (max < 128) return;
2541    
2542    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2543  #elif defined 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 /* COMPILE_PCRE[8|16] */  
2544    add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2545      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2546    JUMPHERE(jump);    JUMPHERE(jump);
2547    }    }
2548  #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */  #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2549    
2550    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2551    if (common->utf)
2552      {
2553      if (max < 0xd800) return;
2554    
2555      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2556      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2557      /* TMP2 contains the high surrogate. */
2558      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2559      OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2560      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2561      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2562      OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2563      JUMPHERE(jump);
2564      }
2565    #endif
2566    }
2567    
2568    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2569    
2570    static BOOL is_char7_bitset(const pcre_uint8 *bitset, BOOL nclass)
2571    {
2572    /* Tells whether the character codes below 128 are enough
2573    to determine a match. */
2574    const pcre_uint8 value = nclass ? 0xff : 0;
2575    const pcre_uint8* end = bitset + 32;
2576    
2577    bitset += 16;
2578    do
2579      {
2580      if (*bitset++ != value)
2581        return FALSE;
2582      }
2583    while (bitset < end);
2584    return TRUE;
2585    }
2586    
2587    static void read_char7_type(compiler_common *common, BOOL full_read)
2588    {
2589    /* Reads the precise character type of a character into TMP1, if the character
2590    is less than 128. Otherwise it returns with zero. Does not check STR_END. The
2591    full_read argument tells whether characters above max are accepted or not. */
2592    DEFINE_COMPILER;
2593    struct sljit_jump *jump;
2594    
2595    SLJIT_ASSERT(common->utf);
2596    
2597    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2598  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));
2599    
2600    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2601    
2602    if (full_read)
2603      {
2604      jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2605      OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2606      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2607      JUMPHERE(jump);
2608      }
2609  }  }
2610    
2611  static void peek_char(compiler_common *common)  #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2612    
2613    static void read_char_range(compiler_common *common, pcre_uint32 min, pcre_uint32 max, BOOL update_str_ptr)
2614  {  {
2615  /* Reads the character into TMP1, keeps STR_PTR.  /* Reads the precise value of a character into TMP1, if the character is
2616  Does not check STR_END. TMP2 Destroyed. */  between min and max (c >= min && c <= max). Otherwise it returns with a value
2617    outside the range. Does not check STR_END. */
2618  DEFINE_COMPILER;  DEFINE_COMPILER;
2619  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2620  struct sljit_jump *jump;  struct sljit_jump *jump;
2621  #endif  #endif
2622    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2623    struct sljit_jump *jump2;
2624    #endif
2625    
2626  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  SLJIT_UNUSED_ARG(update_str_ptr);
2627  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  SLJIT_UNUSED_ARG(min);
2628    SLJIT_UNUSED_ARG(max);
2629    SLJIT_ASSERT(min <= max);
2630    
2631    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2632    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2633    
2634    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2635  if (common->utf)  if (common->utf)
2636    {    {
2637  #if defined COMPILE_PCRE8    if (max < 128 && !update_str_ptr) return;
2638    
2639    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2640  #elif defined COMPILE_PCRE16    if (min >= 0x10000)
2641    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);      {
2642  #endif /* COMPILE_PCRE[8|16] */      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xf0);
2643    add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));      if (update_str_ptr)
2644    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);        OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2645        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2646        jump2 = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0x7);
2647        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2648        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2649        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2650        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2651        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2652        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2653        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2654        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2655        if (!update_str_ptr)
2656          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2657        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2658        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2659        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2660        JUMPHERE(jump2);
2661        if (update_str_ptr)
2662          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2663        }
2664      else if (min >= 0x800 && max <= 0xffff)
2665        {
2666        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xe0);
2667        if (update_str_ptr)
2668          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2669        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2670        jump2 = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xf);
2671        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2672        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2673        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2674        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2675        if (!update_str_ptr)
2676          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2677        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2678        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2679        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2680        JUMPHERE(jump2);
2681        if (update_str_ptr)
2682          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2683        }
2684      else if (max >= 0x800)
2685        add_jump(compiler, (max < 0x10000) ? &common->utfreadchar16 : &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2686      else if (max < 128)
2687        {
2688        OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2689        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2690        }
2691      else
2692        {
2693        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2694        if (!update_str_ptr)
2695          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2696        else
2697          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2698        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2699        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2700        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2701        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2702        if (update_str_ptr)
2703          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2704        }
2705    JUMPHERE(jump);    JUMPHERE(jump);
2706    }    }
2707  #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */  #endif
2708    
2709    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2710    if (common->utf)
2711      {
2712      if (max >= 0x10000)
2713        {
2714        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2715        jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2716        /* TMP2 contains the high surrogate. */
2717        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2718        OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2719        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2720        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2721        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2722        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2723        JUMPHERE(jump);
2724        return;
2725        }
2726    
2727      if (max < 0xd800 && !update_str_ptr) return;
2728    
2729      /* Skip low surrogate if necessary. */
2730      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2731      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2732      if (update_str_ptr)
2733        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2734      if (max >= 0xd800)
2735        OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0x10000);
2736      JUMPHERE(jump);
2737      }
2738    #endif
2739  }  }
2740    
2741  static void read_char8_type(compiler_common *common)  static SLJIT_INLINE void read_char(compiler_common *common)
2742    {
2743    read_char_range(common, 0, READ_CHAR_MAX, TRUE);
2744    }
2745    
2746    static void read_char8_type(compiler_common *common, BOOL update_str_ptr)
2747  {  {
2748  /* 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. */
2749  DEFINE_COMPILER;  DEFINE_COMPILER;
2750  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2751  struct sljit_jump *jump;  struct sljit_jump *jump;
2752  #endif  #endif
2753    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2754    struct sljit_jump *jump2;
2755    #endif
2756    
2757  #ifdef SUPPORT_UTF  SLJIT_UNUSED_ARG(update_str_ptr);
2758    
2759    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2760    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2761    
2762    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2763  if (common->utf)  if (common->utf)
2764    {    {
   OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);  
   OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  
 #if defined COMPILE_PCRE8  
2765    /* This can be an extra read in some situations, but hopefully    /* This can be an extra read in some situations, but hopefully
2766    it is needed in most cases. */    it is needed in most cases. */
2767    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2768    jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);    jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2769    add_jump(compiler, &common->utfreadtype8, JUMP(SLJIT_FAST_CALL));    if (!update_str_ptr)
2770    JUMPHERE(jump);      {
2771  #elif defined COMPILE_PCRE16      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2772    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2773    jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);      OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2774    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2775    JUMPHERE(jump);      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2776    /* Skip low surrogate if necessary. */      OP2(SLJIT_OR, TMP2, 0, TMP2, 0, TMP1, 0);
2777    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0xfc00);      OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2778    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0xd800);      jump2 = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);
2779    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2780    OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 1);      JUMPHERE(jump2);
2781    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);      }
2782  #elif defined COMPILE_PCRE32    else
2783    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);      add_jump(compiler, &common->utfreadtype8, JUMP(SLJIT_FAST_CALL));
   jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);  
   OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);  
2784    JUMPHERE(jump);    JUMPHERE(jump);
 #endif /* COMPILE_PCRE[8|16|32] */  
2785    return;    return;
2786    }    }
2787  #endif /* SUPPORT_UTF */  #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2788  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);  
2789  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  #if !defined COMPILE_PCRE8
 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32  
2790  /* The ctypes array contains only 256 values. */  /* The ctypes array contains only 256 values. */
2791  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2792  jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);  jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);
2793  #endif  #endif
2794  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2795  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32  #if !defined COMPILE_PCRE8
2796  JUMPHERE(jump);  JUMPHERE(jump);
2797  #endif  #endif
2798    
2799    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2800    if (common->utf && update_str_ptr)
2801      {
2802      /* Skip low surrogate if necessary. */
2803      OP2(SLJIT_SUB, TMP2, 0, TMP2, 0, SLJIT_IMM, 0xd800);
2804      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2805      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2806      JUMPHERE(jump);
2807      }
2808    #endif /* SUPPORT_UTF && COMPILE_PCRE16 */
2809  }  }
2810    
2811  static void skip_char_back(compiler_common *common)  static void skip_char_back(compiler_common *common)
# Line 2266  if (common->utf) Line 2843  if (common->utf)
2843  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));
2844  }  }
2845    
2846  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)
2847  {  {
2848  /* 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. */
2849  DEFINE_COMPILER;  DEFINE_COMPILER;
2850    struct sljit_jump *jump;
2851    
2852  if (nltype == NLTYPE_ANY)  if (nltype == NLTYPE_ANY)
2853    {    {
2854    add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));
2855    add_jump(compiler, backtracks, JUMP(jumpiftrue ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));    add_jump(compiler, backtracks, JUMP(jumpifmatch ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));
2856    }    }
2857  else if (nltype == NLTYPE_ANYCRLF)  else if (nltype == NLTYPE_ANYCRLF)
2858    {    {
2859    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_CR);    if (jumpifmatch)
2860    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);      {
2861    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_NL);      add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR));
2862    OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);      add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL));
2863    add_jump(compiler, backtracks, JUMP(jumpiftrue ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));      }
2864      else
2865        {
2866        jump = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR);
2867        add_jump(compiler, backtracks, CMP(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL));
2868        JUMPHERE(jump);
2869        }
2870    }    }
2871  else  else
2872    {    {
2873    SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline < 256);    SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline < 256);
2874    add_jump(compiler, backtracks, CMP(jumpiftrue ? SLJIT_C_EQUAL : SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, common->newline));    add_jump(compiler, backtracks, CMP(jumpifmatch ? SLJIT_C_EQUAL : SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, common->newline));
2875    }    }
2876  }  }
2877    
# Line 2297  else Line 2881  else
2881  static void do_utfreadchar(compiler_common *common)  static void do_utfreadchar(compiler_common *common)
2882  {  {
2883  /* Fast decoding a UTF-8 character. TMP1 contains the first byte  /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2884  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. */
2885  DEFINE_COMPILER;  DEFINE_COMPILER;
2886  struct sljit_jump *jump;  struct sljit_jump *jump;
2887    
2888  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2889    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2890    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2891    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2892    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2893    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2894    
2895  /* Searching for the first zero. */  /* Searching for the first zero. */
2896  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x20);  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
2897  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_C_NOT_ZERO);
2898  /* Two byte sequence. */  /* Two byte sequence. */
 OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  
2899  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));
2900  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x1f);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(2));
2901    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2902    
2903    JUMPHERE(jump);
2904    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2905    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
2906  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2907  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2908  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
 OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(1));  
 sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  
 JUMPHERE(jump);  
2909    
2910  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x10);  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2911  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_C_NOT_ZERO);
2912  /* Three byte sequence. */  /* Three byte sequence. */
 OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  
 OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x0f);  
 OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 12);  
 OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  
 OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
 OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));  
2913  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));
2914  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));  
2915  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
 JUMPHERE(jump);  
2916    
2917  /* Four byte sequence. */  /* Four byte sequence. */
2918  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  JUMPHERE(jump);
2919  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x07);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2920  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 18);  OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2921    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2922    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2923  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);  
2924  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2925  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(4));
2926    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2927    }
2928    
2929    static void do_utfreadchar16(compiler_common *common)
2930    {
2931    /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2932    of the character (>= 0xc0). Return value in TMP1. */
2933    DEFINE_COMPILER;
2934    struct sljit_jump *jump;
2935    
2936    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2937    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2938    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2939    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2940  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);  
2941  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2942  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(3));  
2943  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));  /* Searching for the first zero. */
2944    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
2945    jump = JUMP(SLJIT_C_NOT_ZERO);
2946    /* Two byte sequence. */
2947    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2948    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2949    
2950    JUMPHERE(jump);
2951    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x400);
2952    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_NOT_ZERO);
2953    /* This code runs only in 8 bit mode. No need to shift the value. */
2954    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2955    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2956    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
2957    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2958  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2959  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2960  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(3));  /* Three byte sequence. */
2961    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2962  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2963  }  }
2964    
# Line 2368  jump = JUMP(SLJIT_C_NOT_ZERO); Line 2978  jump = JUMP(SLJIT_C_NOT_ZERO);
2978  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2979  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));
2980  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);
2981    /* The upper 5 bits are known at this point. */
2982    compare = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0x3);
2983  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2984  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2985  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);  
2986  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2987  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2988    
2989  JUMPHERE(compare);  JUMPHERE(compare);
2990  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2991  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
 JUMPHERE(jump);  
2992    
2993  /* 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_sw)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);  
 }  
   
 #elif defined 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);  
   
2994  JUMPHERE(jump);  JUMPHERE(jump);
2995  /* Combine two 16 bit characters. */  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2996  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2997  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);  
2998  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2999  }  }
3000    
3001  #endif /* COMPILE_PCRE[8|16] */  #endif /* COMPILE_PCRE8 */
3002    
3003  #endif /* SUPPORT_UTF */  #endif /* SUPPORT_UTF */
3004    
# Line 2489  if (firstline) Line 3073  if (firstline)
3073      mainloop = LABEL();      mainloop = LABEL();
3074      /* Continual stores does not cause data dependency. */      /* Continual stores does not cause data dependency. */
3075      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);
3076      read_char(common);      read_char_range(common, common->nlmin, common->nlmax, TRUE);
3077      check_newlinechar(common, common->nltype, &newline, TRUE);      check_newlinechar(common, common->nltype, &newline, TRUE);
3078      CMPTO(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0, mainloop);      CMPTO(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0, mainloop);
3079      JUMPHERE(end);      JUMPHERE(end);
# Line 2565  if (newlinecheck) Line 3149  if (newlinecheck)
3149  return mainloop;  return mainloop;
3150  }  }
3151    
3152  #define MAX_N_CHARS 3  static int scan_prefix(compiler_common *common, pcre_uchar *cc, pcre_uint32 *chars, int max_chars)
   
 static SLJIT_INLINE BOOL fast_forward_first_n_chars(compiler_common *common, BOOL firstline)  
3153  {  {
3154  DEFINE_COMPILER;  /* Recursive function, which scans prefix literals. */
3155  struct sljit_label *start;  int len, repeat, len_save, consumed = 0;
3156  struct sljit_jump *quit;  pcre_uint32 caseless, chr, mask;
3157  pcre_uint32 chars[MAX_N_CHARS * 2];  pcre_uchar *alternative, *cc_save;
3158  pcre_uchar *cc = common->start + 1 + IMM2_SIZE;  BOOL last, any;
 int location = 0;  
 pcre_int32 len, c, bit, caseless;  
 int must_stop;  
   
 /* We do not support alternatives now. */  
 if (*(common->start + GET(common->start, 1)) == OP_ALT)  
   return FALSE;  
3159    
3160    repeat = 1;
3161  while (TRUE)  while (TRUE)
3162    {    {
3163      last = TRUE;
3164      any = FALSE;
3165    caseless = 0;    caseless = 0;
3166    must_stop = 1;    switch (*cc)
   switch(*cc)  
3167      {      {
     case OP_CHAR:  
     must_stop = 0;  
     cc++;  
     break;  
   
3168      case OP_CHARI:      case OP_CHARI:
3169      caseless = 1;      caseless = 1;
3170      must_stop = 0;      case OP_CHAR:
3171        last = FALSE;
3172      cc++;      cc++;
3173      break;      break;
3174    
# Line 2614  while (TRUE) Line 3187  while (TRUE)
3187      cc++;      cc++;
3188      continue;      continue;
3189    
3190      case OP_PLUS:      case OP_ASSERT:
3191      case OP_MINPLUS:      case OP_ASSERT_NOT:
3192      case OP_POSPLUS:      case OP_ASSERTBACK:
3193        case OP_ASSERTBACK_NOT:
3194        cc = bracketend(cc);
3195        continue;
3196    
3197        case OP_PLUS:
3198        case OP_MINPLUS:
3199        case OP_POSPLUS:
3200        cc++;
3201        break;
3202    
3203        case OP_EXACTI:
3204        caseless = 1;
3205        case OP_EXACT:
3206        repeat = GET2(cc, 1);
3207        last = FALSE;
3208        cc += 1 + IMM2_SIZE;
3209        break;
3210    
3211        case OP_PLUSI:
3212        case OP_MINPLUSI:
3213        case OP_POSPLUSI:
3214        caseless = 1;
3215        cc++;
3216        break;
3217    
3218        case OP_KET:
3219        cc += 1 + LINK_SIZE;
3220        continue;
3221    
3222        case OP_ALT:
3223        cc += GET(cc, 1);
3224        continue;
3225    
3226        case OP_ONCE:
3227        case OP_ONCE_NC:
3228        case OP_BRA:
3229        case OP_BRAPOS:
3230        case OP_CBRA:
3231        case OP_CBRAPOS:
3232        alternative = cc + GET(cc, 1);
3233        while (*alternative == OP_ALT)
3234          {
3235          max_chars = scan_prefix(common, alternative + 1 + LINK_SIZE, chars, max_chars);
3236          if (max_chars == 0)
3237            return consumed;
3238          alternative += GET(alternative, 1);
3239          }
3240    
3241        if (*cc == OP_CBRA || *cc == OP_CBRAPOS)
3242          cc += IMM2_SIZE;
3243        cc += 1 + LINK_SIZE;
3244        continue;
3245    
3246        case OP_CLASS:
3247    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3248        if (common->utf && !is_char7_bitset((const pcre_uint8 *)(cc + 1), FALSE)) return consumed;
3249    #endif
3250        any = TRUE;
3251        cc += 1 + 32 / sizeof(pcre_uchar);
3252        break;
3253    
3254        case OP_NCLASS:
3255    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3256        if (common->utf) return consumed;
3257    #endif
3258        any = TRUE;
3259        cc += 1 + 32 / sizeof(pcre_uchar);
3260        break;
3261    
3262    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3263        case OP_XCLASS:
3264    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3265        if (common->utf) return consumed;
3266    #endif
3267        any = TRUE;
3268        cc += GET(cc, 1);
3269        break;
3270    #endif
3271    
3272        case OP_DIGIT:
3273    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3274        if (common->utf && !is_char7_bitset((const pcre_uint8 *)common->ctypes - cbit_length + cbit_digit, FALSE))
3275          return consumed;
3276    #endif
3277        any = TRUE;
3278        cc++;
3279        break;
3280    
3281        case OP_WHITESPACE:
3282    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3283        if (common->utf && !is_char7_bitset((const pcre_uint8 *)common->ctypes - cbit_length + cbit_space, FALSE))
3284          return consumed;
3285    #endif
3286        any = TRUE;
3287      cc++;      cc++;
3288      break;      break;
3289    
3290      case OP_EXACT:      case OP_WORDCHAR:
3291      cc += 1 + IMM2_SIZE;  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3292        if (common->utf && !is_char7_bitset((const pcre_uint8 *)common->ctypes - cbit_length + cbit_word, FALSE))
3293          return consumed;
3294    #endif
3295        any = TRUE;
3296        cc++;
3297      break;      break;
3298    
3299      case OP_PLUSI:      case OP_NOT_DIGIT:
3300      case OP_MINPLUSI:      case OP_NOT_WHITESPACE:
3301      case OP_POSPLUSI:      case OP_NOT_WORDCHAR:
3302      caseless = 1;      case OP_ANY:
3303        case OP_ALLANY:
3304    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3305        if (common->utf) return consumed;
3306    #endif
3307        any = TRUE;
3308      cc++;      cc++;
3309      break;      break;
3310    
3311      case OP_EXACTI:  #ifdef SUPPORT_UCP
3312      caseless = 1;      case OP_NOTPROP:
3313      cc += 1 + IMM2_SIZE;      case OP_PROP:
3314    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3315        if (common->utf) return consumed;
3316    #endif
3317        any = TRUE;
3318        cc += 1 + 2;
3319      break;      break;
3320    #endif
3321    
3322        case OP_TYPEEXACT:
3323        repeat = GET2(cc, 1);
3324        cc += 1 + IMM2_SIZE;
3325        continue;
3326    
3327      default:      default:
3328      must_stop = 2;      return consumed;
     break;  
3329      }      }
3330    
3331    if (must_stop == 2)    if (any)
3332        break;      {
3333    #if defined COMPILE_PCRE8
3334        mask = 0xff;
3335    #elif defined COMPILE_PCRE16
3336        mask = 0xffff;
3337    #elif defined COMPILE_PCRE32
3338        mask = 0xffffffff;
3339    #else
3340        SLJIT_ASSERT_STOP();
3341    #endif
3342    
3343        do
3344          {
3345          chars[0] = mask;
3346          chars[1] = mask;
3347    
3348          if (--max_chars == 0)
3349            return consumed;
3350          consumed++;
3351          chars += 2;
3352          }
3353        while (--repeat > 0);
3354    
3355        repeat = 1;
3356        continue;
3357        }
3358    
3359    len = 1;    len = 1;
3360  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
3361    if (common->utf && HAS_EXTRALEN(cc[0])) len += GET_EXTRALEN(cc[0]);    if (common->utf && HAS_EXTRALEN(*cc)) len += GET_EXTRALEN(*cc);
3362  #endif  #endif
3363    
3364    if (caseless && char_has_othercase(common, cc))    if (caseless != 0 && char_has_othercase(common, cc))
3365      {      {
3366      caseless = char_get_othercase_bit(common, cc);      caseless = char_get_othercase_bit(common, cc);
3367      if (caseless == 0)      if (caseless == 0)
3368        return FALSE;        return consumed;
3369  #ifdef COMPILE_PCRE8  #ifdef COMPILE_PCRE8
3370      caseless = ((caseless & 0xff) << 8) | (len - (caseless >> 8));      caseless = ((caseless & 0xff) << 8) | (len - (caseless >> 8));
3371  #else  #else
# Line 2666  while (TRUE) Line 3378  while (TRUE)
3378    else    else
3379      caseless = 0;      caseless = 0;
3380    
3381    while (len > 0 && location < MAX_N_CHARS * 2)    len_save = len;
3382      {    cc_save = cc;
3383      c = *cc;    while (TRUE)
3384      bit = 0;      {
3385      if (len == (caseless & 0xff))      do
3386        {        {
3387        bit = caseless >> 8;        chr = *cc;
3388        c |= bit;  #ifdef COMPILE_PCRE32
3389          if (SLJIT_UNLIKELY(chr == NOTACHAR))
3390            return consumed;
3391    #endif
3392          mask = 0;
3393          if ((pcre_uint32)len == (caseless & 0xff))
3394            {
3395            mask = caseless >> 8;
3396            chr |= mask;
3397            }
3398    
3399          if (chars[0] == NOTACHAR)
3400            {
3401            chars[0] = chr;
3402            chars[1] = mask;
3403            }
3404          else
3405            {
3406            mask |= chars[0] ^ chr;
3407            chr |= mask;
3408            chars[0] = chr;
3409            chars[1] |= mask;
3410            }
3411    
3412          len--;
3413          if (--max_chars == 0)
3414            return consumed;
3415          consumed++;
3416          chars += 2;
3417          cc++;
3418        }        }
3419        while (len > 0);
3420    
3421        if (--repeat == 0)
3422          break;
3423    
3424      chars[location] = c;      len = len_save;
3425      chars[location + 1] = bit;      cc = cc_save;
3426        }
3427    
3428      len--;    repeat = 1;
3429      location += 2;    if (last)
3430      cc++;      return consumed;
3431      }
3432    }
3433    
3434    #define MAX_N_CHARS 16
3435    
3436    static SLJIT_INLINE BOOL fast_forward_first_n_chars(compiler_common *common, BOOL firstline)
3437    {
3438    DEFINE_COMPILER;
3439    struct sljit_label *start;
3440    struct sljit_jump *quit;
3441    pcre_uint32 chars[MAX_N_CHARS * 2];
3442    pcre_uint8 ones[MAX_N_CHARS];
3443    pcre_uint32 mask;
3444    int i, max;
3445    int offsets[3];
3446    
3447    for (i = 0; i < MAX_N_CHARS; i++)
3448      {
3449      chars[i << 1] = NOTACHAR;
3450      chars[(i << 1) + 1] = 0;
3451      }
3452    
3453    max = scan_prefix(common, common->start, chars, MAX_N_CHARS);
3454    
3455    if (max <= 1)
3456      return FALSE;
3457    
3458    for (i = 0; i < max; i++)
3459      {
3460      mask = chars[(i << 1) + 1];
3461      ones[i] = ones_in_half_byte[mask & 0xf];
3462      mask >>= 4;
3463      while (mask != 0)
3464        {
3465        ones[i] += ones_in_half_byte[mask & 0xf];
3466        mask >>= 4;
3467      }      }
3468      }
3469    
3470    if (location >= MAX_N_CHARS * 2 || must_stop != 0)  offsets[0] = -1;
3471    /* Scan forward. */
3472    for (i = 0; i < max; i++)
3473      if (ones[i] <= 2) {
3474        offsets[0] = i;
3475      break;      break;
3476    }    }
3477    
3478  /* At least two characters are required. */  if (offsets[0] == -1)
3479  if (location < 2 * 2)    return FALSE;
3480      return FALSE;  
3481    /* Scan backward. */
3482    offsets[1] = -1;
3483    for (i = max - 1; i > offsets[0]; i--)
3484      if (ones[i] <= 2) {
3485        offsets[1] = i;
3486        break;
3487      }
3488    
3489    offsets[2] = -1;
3490    if (offsets[1] >= 0)
3491      {
3492      /* Scan from middle. */
3493      for (i = (offsets[0] + offsets[1]) / 2 + 1; i < offsets[1]; i++)
3494        if (ones[i] <= 2)
3495          {
3496          offsets[2] = i;
3497          break;
3498          }
3499    
3500      if (offsets[2] == -1)
3501        {
3502        for (i = (offsets[0] + offsets[1]) / 2; i > offsets[0]; i--)
3503          if (ones[i] <= 2)
3504            {
3505            offsets[2] = i;
3506            break;
3507            }
3508        }
3509      }
3510    
3511    SLJIT_ASSERT(offsets[1] == -1 || (offsets[0] < offsets[1]));
3512    SLJIT_ASSERT(offsets[2] == -1 || (offsets[0] < offsets[2] && offsets[1] > offsets[2]));
3513    
3514    chars[0] = chars[offsets[0] << 1];
3515    chars[1] = chars[(offsets[0] << 1) + 1];
3516    if (offsets[2] >= 0)
3517      {
3518      chars[2] = chars[offsets[2] << 1];
3519      chars[3] = chars[(offsets[2] << 1) + 1];
3520      }
3521    if (offsets[1] >= 0)
3522      {
3523      chars[4] = chars[offsets[1] << 1];
3524      chars[5] = chars[(offsets[1] << 1) + 1];
3525      }
3526    
3527    max -= 1;
3528  if (firstline)  if (firstline)
3529    {    {
3530    SLJIT_ASSERT(common->first_line_end != 0);    SLJIT_ASSERT(common->first_line_end != 0);
3531    OP1(SLJIT_MOV, TMP3, 0, STR_END, 0);    OP1(SLJIT_MOV, TMP3, 0, STR_END, 0);
3532    OP2(SLJIT_SUB, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, SLJIT_IMM, (location >> 1) - 1);    OP2(SLJIT_SUB, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, SLJIT_IMM, IN_UCHARS(max));
3533    }    }
3534  else  else
3535    OP2(SLJIT_SUB, STR_END, 0, STR_END, 0, SLJIT_IMM, (location >> 1) - 1);    OP2(SLJIT_SUB, STR_END, 0, STR_END, 0, SLJIT_IMM, IN_UCHARS(max));
3536    
3537  start = LABEL();  start = LABEL();
3538  quit = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);  quit = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3539    
3540  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(offsets[0]));
3541  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  if (offsets[1] >= 0)
3542      OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(offsets[1]));
3543  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));
3544    
3545  if (chars[1] != 0)  if (chars[1] != 0)
3546    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, chars[1]);    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, chars[1]);
3547  CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, chars[0], start);  CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, chars[0], start);
3548  if (location > 2 * 2)  if (offsets[2] >= 0)
3549    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(offsets[2] - 1));
3550  if (chars[3] != 0)  
3551    OP2(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_IMM, chars[3]);  if (offsets[1] >= 0)
 CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, chars[2], start);  
 if (location > 2 * 2)  
3552    {    {
3553    if (chars[5] != 0)    if (chars[5] != 0)
3554      OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, chars[5]);      OP2(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_IMM, chars[5]);
3555    CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, chars[4], start);    CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, chars[4], start);
3556      }
3557    
3558    if (offsets[2] >= 0)
3559      {
3560      if (chars[3] != 0)
3561        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, chars[3]);
3562      CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, chars[2], start);
3563    }    }
3564  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));
3565    
# Line 2728  JUMPHERE(quit); Line 3568  JUMPHERE(quit);
3568  if (firstline)  if (firstline)
3569    OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);    OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);
3570  else  else
3571    OP2(SLJIT_ADD, STR_END, 0, STR_END, 0, SLJIT_IMM, (location >> 1) - 1);    OP2(SLJIT_ADD, STR_END, 0, STR_END, 0, SLJIT_IMM, IN_UCHARS(max));
3572  return TRUE;  return TRUE;
3573  }  }
3574    
# Line 2848  firstchar = CMP(SLJIT_C_LESS_EQUAL, STR_ Line 3688  firstchar = CMP(SLJIT_C_LESS_EQUAL, STR_
3688  skip_char_back(common);  skip_char_back(common);
3689    
3690  loop = LABEL();  loop = LABEL();
3691  read_char(common);  read_char_range(common, common->nlmin, common->nlmax, TRUE);
3692  lastchar = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);  lastchar = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3693  if (common->nltype == NLTYPE_ANY || common->nltype == NLTYPE_ANYCRLF)  if (common->nltype == NLTYPE_ANY || common->nltype == NLTYPE_ANYCRLF)
3694    foundcr = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR);    foundcr = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR);
# Line 2877  if (firstline) Line 3717  if (firstline)
3717    OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);    OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);
3718  }  }
3719    
3720  static SLJIT_INLINE void fast_forward_start_bits(compiler_common *common, sljit_uw start_bits, BOOL firstline)  static BOOL check_class_ranges(compiler_common *common, const pcre_uint8 *bits, BOOL nclass, BOOL invert, jump_list **backtracks);
3721    
3722    static SLJIT_INLINE void fast_forward_start_bits(compiler_common *common, pcre_uint8 *start_bits, BOOL firstline)
3723  {  {
3724  DEFINE_COMPILER;  DEFINE_COMPILER;
3725  struct sljit_label *start;  struct sljit_label *start;
3726  struct sljit_jump *quit;  struct sljit_jump *quit;
3727  struct sljit_jump *found;  struct sljit_jump *found = NULL;
3728    jump_list *matches = NULL;
3729  #ifndef COMPILE_PCRE8  #ifndef COMPILE_PCRE8
3730  struct sljit_jump *jump;  struct sljit_jump *jump;
3731  #endif  #endif
# Line 2901  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_P Line 3744  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_P
3744  if (common->utf)  if (common->utf)
3745    OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);    OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);
3746  #endif  #endif
3747    
3748    if (!check_class_ranges(common, start_bits, (start_bits[31] & 0x80) != 0, TRUE, &matches))
3749      {
3750  #ifndef COMPILE_PCRE8  #ifndef COMPILE_PCRE8
3751  jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 255);    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 255);
3752  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 255);    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 255);
3753  JUMPHERE(jump);    JUMPHERE(jump);
3754  #endif  #endif
3755  OP2(SLJIT_AND, TMP2, 0, TMP1, 0, SLJIT_IMM, 0x7);    OP2(SLJIT_AND, TMP2, 0, TMP1, 0, SLJIT_IMM, 0x7);
3756  OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 3);    OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 3);
3757  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), start_bits);    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)start_bits);
3758  OP2(SLJIT_SHL, TMP2, 0, SLJIT_IMM, 1, TMP2, 0);    OP2(SLJIT_SHL, TMP2, 0, SLJIT_IMM, 1, TMP2, 0);
3759  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);
3760  found = JUMP(SLJIT_C_NOT_ZERO);    found = JUMP(SLJIT_C_NOT_ZERO);
3761      }
3762    
3763  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
3764  if (common->utf)  if (common->utf)
# Line 2939  if (common->utf) Line 3786  if (common->utf)
3786  #endif /* COMPILE_PCRE[8|16] */  #endif /* COMPILE_PCRE[8|16] */
3787  #endif /* SUPPORT_UTF */  #endif /* SUPPORT_UTF */
3788  JUMPTO(SLJIT_JUMP, start);  JUMPTO(SLJIT_JUMP, start);
3789  JUMPHERE(found);  if (found != NULL)
3790      JUMPHERE(found);
3791    if (matches != NULL)
3792      set_jumps(matches, LABEL());
3793  JUMPHERE(quit);  JUMPHERE(quit);
3794    
3795  if (firstline)  if (firstline)
# Line 3022  GET_LOCAL_BASE(TMP3, 0, 0); Line 3872  GET_LOCAL_BASE(TMP3, 0, 0);
3872  /* Drop frames until we reach STACK_TOP. */  /* Drop frames until we reach STACK_TOP. */
3873  mainloop = LABEL();  mainloop = LABEL();
3874  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(TMP1), 0);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(TMP1), 0);
3875  jump = CMP(SLJIT_C_SIG_LESS_EQUAL, TMP2, 0, SLJIT_IMM, frame_end);  OP2(SLJIT_SUB | SLJIT_SET_S, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0);
3876    jump = JUMP(SLJIT_C_SIG_LESS_EQUAL);
3877    
3878  OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, TMP3, 0);  OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, TMP3, 0);
3879  OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), 0, SLJIT_MEM1(TMP1), sizeof(sljit_sw));  OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), 0, SLJIT_MEM1(TMP1), sizeof(sljit_sw));
3880  OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), sizeof(sljit_sw), SLJIT_MEM1(TMP1), 2 * sizeof(sljit_sw));  OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), sizeof(sljit_sw), SLJIT_MEM1(TMP1), 2 * sizeof(sljit_sw));
# Line 3030  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_I Line 3882  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_I
3882  JUMPTO(SLJIT_JUMP, mainloop);  JUMPTO(SLJIT_JUMP, mainloop);
3883    
3884  JUMPHERE(jump);  JUMPHERE(jump);
3885  jump = CMP(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, frame_end);  jump = JUMP(SLJIT_C_SIG_LESS);
3886  /* End of dropping frames. */  /* End of dropping frames. */
3887  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3888    
3889  JUMPHERE(jump);  JUMPHERE(jump);
3890  jump = CMP(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, frame_setstrbegin);  OP1(SLJIT_NEG, TMP2, 0, TMP2, 0);
3891  /* Set string begin. */  OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, TMP3, 0);
3892  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(TMP1), sizeof(sljit_sw));  OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), 0, SLJIT_MEM1(TMP1), sizeof(sljit_sw));
 OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 2 * sizeof(sljit_sw));  
 OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0), TMP2, 0);  
 JUMPTO(SLJIT_JUMP, mainloop);  
   
 JUMPHERE(jump);  
 if (common->mark_ptr != 0)  
   {  
   jump = CMP(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, frame_setmark);  
   OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(TMP1), sizeof(sljit_sw));  
   OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 2 * sizeof(sljit_sw));  
   OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr, TMP2, 0);  
   JUMPTO(SLJIT_JUMP, mainloop);  
   
   JUMPHERE(jump);  
   }  
   
 /* Unknown command. */  
3893  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 2 * sizeof(sljit_sw));  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 2 * sizeof(sljit_sw));
3894  JUMPTO(SLJIT_JUMP, mainloop);  JUMPTO(SLJIT_JUMP, mainloop);
3895  }  }
# Line 3063  static void check_wordboundary(compiler_ Line 3898  static void check_wordboundary(compiler_
3898  {  {
3899  DEFINE_COMPILER;  DEFINE_COMPILER;
3900  struct sljit_jump *skipread;  struct sljit_jump *skipread;
3901    jump_list *skipread_list = NULL;
3902  #if !(defined COMPILE_PCRE8) || defined SUPPORT_UTF  #if !(defined COMPILE_PCRE8) || defined SUPPORT_UTF
3903  struct sljit_jump *jump;  struct sljit_jump *jump;
3904  #endif  #endif
# Line 3120  else Line 3956  else
3956  JUMPHERE(skipread);  JUMPHERE(skipread);
3957    
3958  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 0);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 0);
3959  skipread = check_str_end(common);  check_str_end(common, &skipread_list);
3960  peek_char(common);  peek_char(common, READ_CHAR_MAX);
3961    
3962  /* Testing char type. This is a code duplication. */  /* Testing char type. This is a code duplication. */
3963  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 3161  else Line 3997  else
3997      JUMPHERE(jump);      JUMPHERE(jump);
3998  #endif /* COMPILE_PCRE8 */  #endif /* COMPILE_PCRE8 */
3999    }    }
4000  JUMPHERE(skipread);  set_jumps(skipread_list, LABEL());
4001    
4002  OP2(SLJIT_XOR | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1);  OP2(SLJIT_XOR | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1);
4003  sljit_emit_fast_return(compiler, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0);  sljit_emit_fast_return(compiler, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0);
4004  }  }
4005    
4006  /*  static BOOL check_class_ranges(compiler_common *common, const pcre_uint8 *bits, BOOL nclass, BOOL invert, jump_list **backtracks)
   range format:  
   
   ranges[0] = length of the range (max MAX_RANGE_SIZE, -1 means invalid range).  
   ranges[1] = first bit (0 or 1)  
   ranges[2-length] = position of the bit change (when the current bit is not equal to the previous)  
 */  
   
 static BOOL check_ranges(compiler_common *common, int *ranges, jump_list **backtracks, BOOL readch)  
4007  {  {
4008  DEFINE_COMPILER;  DEFINE_COMPILER;
4009  struct sljit_jump *jump;  int ranges[MAX_RANGE_SIZE];
   
 if (ranges[0] < 0)  
   return FALSE;  
   
 switch(ranges[0])  
   {  
   case 1:  
   if (readch)  
     read_char(common);  
   add_jump(compiler, backtracks, CMP(ranges[1] == 0 ? SLJIT_C_LESS : SLJIT_C_GREATER_EQUAL, TMP1, 0, SLJIT_IMM, ranges[2]));  
   return TRUE;  
   
   case 2:  
   if (readch)  
     read_char(common);  
   OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[2]);  
   add_jump(compiler, backtracks, CMP(ranges[1] != 0 ? SLJIT_C_LESS : SLJIT_C_GREATER_EQUAL, TMP1, 0, SLJIT_IMM, ranges[3] - ranges[2]));  
   return TRUE;  
   
   case 4:  
   if (ranges[2] + 1 == ranges[3] && ranges[4] + 1 == ranges[5])  
     {  
     if (readch)  
       read_char(common);  
     if (ranges[1] != 0)  
       {  
       add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, ranges[2]));  
       add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, ranges[4]));  
       }  
     else  
       {  
       jump = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, ranges[2]);  
       add_jump(compiler, backtracks, CMP(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, ranges[4]));  
       JUMPHERE(jump);  
       }  
     return TRUE;  
     }  
   if ((ranges[3] - ranges[2]) == (ranges[5] - ranges[4]) && is_powerof2(ranges[4] - ranges[2]))  
     {  
     if (readch)  
       read_char(common);  
     OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[4] - ranges[2]);  
     OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[4]);  
     add_jump(compiler, backtracks, CMP(ranges[1] != 0 ? SLJIT_C_LESS : SLJIT_C_GREATER_EQUAL, TMP1, 0, SLJIT_IMM, ranges[5] - ranges[4]));  
     return TRUE;  
     }  
   return FALSE;  
   
   default:  
   return FALSE;  
   }  
 }  
   
 static void get_ctype_ranges(compiler_common *common, int flag, int *ranges)  
 {  
 int i, bit, length;  
 const pcre_uint8 *ctypes = (const pcre_uint8*)common->ctypes;  
   
 bit = ctypes[0] & flag;  
 ranges[0] = -1;  
 ranges[1] = bit != 0 ? 1 : 0;  
 length = 0;  
   
 for (i = 1; i < 256; i++)  
   if ((ctypes[i] & flag) != bit)  
     {  
     if (length >= MAX_RANGE_SIZE)  
       return;  
     ranges[2 + length] = i;  
     length++;  
     bit ^= flag;  
     }  
   
 if (bit != 0)  
   {  
   if (length >= MAX_RANGE_SIZE)  
     return;  
   ranges[2 + length] = 256;  
   length++;  
   }  
 ranges[0] = length;  
 }  
   
 static BOOL check_class_ranges(compiler_common *common, const pcre_uint8 *bits, BOOL nclass, jump_list **backtracks)  
 {  
 int ranges[2 + MAX_RANGE_SIZE];  
4010  pcre_uint8 bit, cbit, all;  pcre_uint8 bit, cbit, all;
4011  int i, byte, length = 0;  int i, byte, length = 0;
4012    
4013  bit = bits[0] & 0x1;  bit = bits[0] & 0x1;
4014  ranges[1] = bit;  /* All bits will be zero or one (since bit is zero or one). */
 /* Can be 0 or 255. */  
4015  all = -bit;  all = -bit;
4016    
4017  for (i = 0; i < 256; )  for (i = 0; i < 256; )
# Line 3285  for (i = 0; i < 256; ) Line 4026  for (i = 0; i < 256; )
4026        {        {
4027        if (length >= MAX_RANGE_SIZE)        if (length >= MAX_RANGE_SIZE)
4028          return FALSE;          return FALSE;
4029        ranges[2 + length] = i;        ranges[length] = i;
4030        length++;        length++;
4031        bit = cbit;        bit = cbit;
4032        all = -cbit;        all = -cbit;
# Line 3298  if (((bit == 0) && nclass) || ((bit == 1 Line 4039  if (((bit == 0) && nclass) || ((bit == 1
4039    {    {
4040    if (length >= MAX_RANGE_SIZE)    if (length >= MAX_RANGE_SIZE)
4041      return FALSE;      return FALSE;
4042    ranges[2 + length] = 256;    ranges[length] = 256;
4043    length++;    length++;
4044    }    }
4045  ranges[0] = length;  
4046    if (length < 0 || length > 4)
4047  return check_ranges(common, ranges, backtracks, FALSE);    return FALSE;
4048    
4049    bit = bits[0] & 0x1;
4050    if (invert) bit ^= 0x1;
4051    
4052    /* No character is accepted. */
4053    if (length == 0 && bit == 0)
4054      add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));
4055    
4056    switch(length)
4057      {
4058      case 0:
4059      /* When bit != 0, all characters are accepted. */
4060      return TRUE;
4061    
4062      case 1:
4063      add_jump(compiler, backtracks, CMP(bit == 0 ? SLJIT_C_LESS : SLJIT_C_GREATER_EQUAL, TMP1, 0, SLJIT_IMM, ranges[0]));
4064      return TRUE;
4065    
4066      case 2:
4067      if (ranges[0] + 1 != ranges[1])
4068        {
4069        OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[0]);
4070        add_jump(compiler, backtracks, CMP(bit != 0 ? SLJIT_C_LESS : SLJIT_C_GREATER_EQUAL, TMP1, 0, SLJIT_IMM, ranges[1] - ranges[0]));
4071        }
4072      else
4073        add_jump(compiler, backtracks, CMP(bit != 0 ? SLJIT_C_EQUAL : SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, ranges[0]));
4074      return TRUE;
4075    
4076      case 3:
4077      if (bit != 0)
4078        {
4079        add_jump(compiler, backtracks, CMP(SLJIT_C_GREATER_EQUAL, TMP1, 0, SLJIT_IMM, ranges[2]));
4080        if (ranges[0] + 1 != ranges[1])
4081          {
4082          OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[0]);
4083          add_jump(compiler, backtracks, CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, ranges[1] - ranges[0]));
4084          }
4085        else
4086          add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, ranges[0]));
4087        return TRUE;
4088        }
4089    
4090      add_jump(compiler, backtracks, CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, ranges[0]));
4091      if (ranges[1] + 1 != ranges[2])
4092        {
4093        OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[1]);
4094        add_jump(compiler, backtracks, CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, ranges[2] - ranges[1]));
4095        }
4096      else
4097        add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, ranges[1]));
4098      return TRUE;
4099    
4100      case 4:
4101      if ((ranges[1] - ranges[0]) == (ranges[3] - ranges[2])
4102          && (ranges[0] | (ranges[2] - ranges[0])) == ranges[2]
4103          && is_powerof2(ranges[2] - ranges[0]))
4104        {
4105        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[2] - ranges[0]);
4106        if (ranges[2] + 1 != ranges[3])
4107          {
4108          OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[2]);
4109          add_jump(compiler, backtracks, CMP(bit != 0 ? SLJIT_C_LESS : SLJIT_C_GREATER_EQUAL, TMP1, 0, SLJIT_IMM, ranges[3] - ranges[2]));
4110          }
4111        else
4112          add_jump(compiler, backtracks, CMP(bit != 0 ? SLJIT_C_EQUAL : SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, ranges[2]));
4113        return TRUE;
4114        }
4115    
4116      if (bit != 0)
4117        {
4118        i = 0;
4119        if (ranges[0] + 1 != ranges[1])
4120          {
4121          OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[0]);
4122          add_jump(compiler, backtracks, CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, ranges[1] - ranges[0]));
4123          i = ranges[0];
4124          }
4125        else
4126          add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, ranges[0]));
4127    
4128        if (ranges[2] + 1 != ranges[3])
4129          {
4130          OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[2] - i);
4131          add_jump(compiler, backtracks, CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, ranges[3] - ranges[2]));
4132          }
4133        else
4134          add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, ranges[2] - i));
4135        return TRUE;
4136        }
4137    
4138      OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[0]);
4139      add_jump(compiler, backtracks, CMP(SLJIT_C_GREATER_EQUAL, TMP1, 0, SLJIT_IMM, ranges[3] - ranges[0]));
4140      if (ranges[1] + 1 != ranges[2])
4141        {
4142        OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[1] - ranges[0]);
4143        add_jump(compiler, backtracks, CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, ranges[2] - ranges[1]));
4144        }
4145      else
4146        add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, ranges[1] - ranges[0]));
4147      return TRUE;
4148    
4149      default:
4150      SLJIT_ASSERT_STOP();
4151      return FALSE;
4152      }
4153  }  }
4154    
4155  static void check_anynewline(compiler_common *common)  static void check_anynewline(compiler_common *common)
# Line 3481  sljit_emit_fast_return(compiler, RETURN_ Line 4327  sljit_emit_fast_return(compiler, RETURN_
4327    
4328  #if defined SUPPORT_UTF && defined SUPPORT_UCP  #if defined SUPPORT_UTF && defined SUPPORT_UCP
4329    
4330  static const pcre_uchar *SLJIT_CALL do_utf_caselesscmp(pcre_uchar *src1, jit_arguments *args, pcre_uchar *end1)  static const pcre_uchar * SLJIT_CALL do_utf_caselesscmp(pcre_uchar *src1, jit_arguments *args, pcre_uchar *end1)
4331  {  {
4332  /* This function would be ineffective to do in JIT level. */  /* This function would be ineffective to do in JIT level. */
4333  pcre_uint32 c1, c2;  pcre_uint32 c1, c2;
# Line 3577  do Line 4423  do
4423  #endif  #endif
4424    
4425    context->length -= IN_UCHARS(1);    context->length -= IN_UCHARS(1);
4426  #if defined SLJIT_UNALIGNED && SLJIT_UNALIGNED  #if (defined SLJIT_UNALIGNED && SLJIT_UNALIGNED) && (defined COMPILE_PCRE8 || defined COMPILE_PCRE16)
4427    
4428    /* Unaligned read is supported. */    /* Unaligned read is supported. */
4429    if (othercasebit != 0 && othercasechar == cc)    if (othercasebit != 0 && othercasechar == cc)
# Line 3594  do Line 4440  do
4440    
4441  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
4442    if (context->ucharptr >= 4 || context->length == 0 || (context->ucharptr == 2 && context->length == 1))    if (context->ucharptr >= 4 || context->length == 0 || (context->ucharptr == 2 && context->length == 1))
4443  #elif defined COMPILE_PCRE16  #else
4444    if (context->ucharptr >= 2 || context->length == 0)    if (context->ucharptr >= 2 || context->length == 0)
 #elif defined COMPILE_PCRE32  
   if (1 /* context->ucharptr >= 1 || context->length == 0 */)  
4445  #endif  #endif
4446      {      {
 #if defined COMPILE_PCRE8 || defined COMPILE_PCRE16  
4447      if (context->length >= 4)      if (context->length >= 4)
4448        OP1(SLJIT_MOV_SI, context->sourcereg, 0, SLJIT_MEM1(STR_PTR), -context->length);        OP1(SLJIT_MOV_SI, context->sourcereg, 0, SLJIT_MEM1(STR_PTR), -context->length);
 #if defined COMPILE_PCRE8  
4449      else if (context->length >= 2)      else if (context->length >= 2)
4450        OP1(SLJIT_MOV_UH, context->sourcereg, 0, SLJIT_MEM1(STR_PTR), -context->length);        OP1(SLJIT_MOV_UH, context->sourcereg, 0, SLJIT_MEM1(STR_PTR), -context->length);
4451    #if defined COMPILE_PCRE8
4452      else if (context->length >= 1)      else if (context->length >= 1)
4453        OP1(SLJIT_MOV_UB, context->sourcereg, 0, SLJIT_MEM1(STR_PTR), -context->length);        OP1(SLJIT_MOV_UB, context->sourcereg, 0, SLJIT_MEM1(STR_PTR), -context->length);
4454  #elif defined COMPILE_PCRE16  #endif /* COMPILE_PCRE8 */
     else if (context->length >= 2)  
       OP1(SLJIT_MOV_UH, context->sourcereg, 0, SLJIT_MEM1(STR_PTR), -context->length);  
 #endif /* COMPILE_PCRE[8|16] */  
 #elif defined COMPILE_PCRE32  
     OP1(MOV_UCHAR, context->sourcereg, 0, SLJIT_MEM1(STR_PTR), -context->length);  
 #endif /* COMPILE_PCRE[8|16|32] */  
4455      context->sourcereg = context->sourcereg == TMP1 ? TMP2 : TMP1;      context->sourcereg = context->sourcereg == TMP1 ? TMP2 : TMP1;
4456    
4457      switch(context->ucharptr)      switch(context->ucharptr)
# Line 3625  do Line 4462  do
4462        add_jump(compiler, backtracks, CMP(SLJIT_C_NOT_EQUAL, context->sourcereg, 0, SLJIT_IMM, context->c.asint | context->oc.asint));        add_jump(compiler, backtracks, CMP(SLJIT_C_NOT_EQUAL, context->sourcereg, 0, SLJIT_IMM, context->c.asint | context->oc.asint));
4463        break;        break;
4464    
 #if defined COMPILE_PCRE8 || defined COMPILE_PCRE16  
4465        case 2 / sizeof(pcre_uchar):        case 2 / sizeof(pcre_uchar):
4466        if (context->oc.asushort != 0)        if (context->oc.asushort != 0)
4467          OP2(SLJIT_OR, context->sourcereg, 0, context->sourcereg, 0, SLJIT_IMM, context->oc.asushort);          OP2(SLJIT_OR, context->sourcereg, 0, context->sourcereg, 0, SLJIT_IMM, context->oc.asushort);
# Line 3640  do Line 4476  do
4476        break;        break;
4477  #endif  #endif
4478    
 #endif /* COMPILE_PCRE[8|16] */  
   
4479        default:        default:
4480        SLJIT_ASSERT_STOP();        SLJIT_ASSERT_STOP();
4481        break;        break;
# Line 3651  do Line 4485  do
4485    
4486  #else  #else
4487    
4488    /* Unaligned read is unsupported. */    /* Unaligned read is unsupported or in 32 bit mode. */
4489    if (context->length > 0)    if (context->length >= 1)
4490      OP1(MOV_UCHAR, context->sourcereg, 0, SLJIT_MEM1(STR_PTR), -context->length);      OP1(MOV_UCHAR, context->sourcereg, 0, SLJIT_MEM1(STR_PTR), -context->length);
4491    
4492    context->sourcereg = context->sourcereg == TMP1 ? TMP2 : TMP1;    context->sourcereg = context->sourcereg == TMP1 ? TMP2 : TMP1;
# Line 3682  return cc; Line 4516  return cc;
4516  #define SET_TYPE_OFFSET(value) \  #define SET_TYPE_OFFSET(value) \
4517    if ((value) != typeoffset) \    if ((value) != typeoffset) \
4518      { \      { \
4519      if ((value) > typeoffset) \      if ((value) < typeoffset) \
       OP2(SLJIT_SUB, typereg, 0, typereg, 0, SLJIT_IMM, (value) - typeoffset); \  
     else \  
4520        OP2(SLJIT_ADD, typereg, 0, typereg, 0, SLJIT_IMM, typeoffset - (value)); \        OP2(SLJIT_ADD, typereg, 0, typereg, 0, SLJIT_IMM, typeoffset - (value)); \
4521        else \
4522          OP2(SLJIT_SUB, typereg, 0, typereg, 0, SLJIT_IMM, (value) - typeoffset); \
4523      } \      } \
4524    typeoffset = (value);    typeoffset = (value);
4525    
4526  #define SET_CHAR_OFFSET(value) \  #define SET_CHAR_OFFSET(value) \
4527    if ((value) != charoffset) \    if ((value) != charoffset) \
4528      { \      { \
4529      if ((value) > charoffset) \      if ((value) < charoffset) \
4530        OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, (value) - charoffset); \        OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(charoffset - (value))); \
4531      else \      else \
4532        OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, charoffset - (value)); \        OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)((value) - charoffset)); \
4533      } \      } \
4534    charoffset = (value);    charoffset = (value);
4535    
# Line 3703  static void compile_xclass_matchingpath( Line 4537  static void compile_xclass_matchingpath(
4537  {  {
4538  DEFINE_COMPILER;  DEFINE_COMPILER;
4539  jump_list *found = NULL;  jump_list *found = NULL;
4540  jump_list **list = (*cc & XCL_NOT) == 0 ? &found : backtracks;  jump_list **list = (cc[0] & XCL_NOT) == 0 ? &found : backtracks;
4541  pcre_int32 c, charoffset;  sljit_uw c, charoffset, max = 256, min = READ_CHAR_MAX;
 const pcre_uint32 *other_cases;  
4542  struct sljit_jump *jump = NULL;  struct sljit_jump *jump = NULL;
4543  pcre_uchar *ccbegin;  pcre_uchar *ccbegin;
4544  int compares, invertcmp, numberofcmps;  int compares, invertcmp, numberofcmps;
4545    #if defined SUPPORT_UTF && (defined COMPILE_PCRE8 || defined COMPILE_PCRE16)
4546    BOOL utf = common->utf;
4547    #endif
4548    
4549  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4550  BOOL needstype = FALSE, needsscript = FALSE, needschar = FALSE;  BOOL needstype = FALSE, needsscript = FALSE, needschar = FALSE;
4551  BOOL charsaved = FALSE;  BOOL charsaved = FALSE;
4552  int typereg = TMP1, scriptreg = TMP1;  int typereg = TMP1, scriptreg = TMP1;
4553  pcre_int32 typeoffset;  const pcre_uint32 *other_cases;
4554    sljit_uw typeoffset;
4555  #endif  #endif
4556    
4557  /* Although SUPPORT_UTF must be defined, we are  /* Scanning the necessary info. */
4558     not necessary in utf mode even in 8 bit mode. */  cc++;
4559  detect_partial_match(common, backtracks);  ccbegin = cc;
4560  read_char(common);  compares = 0;
4561    if (cc[-1] & XCL_MAP)
 if ((*cc++ & XCL_MAP) != 0)  
4562    {    {
4563    OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);    min = 0;
 #ifndef COMPILE_PCRE8  
   jump = CMP(SLJIT_C_GREATER, TMP1, 0, SLJIT_IMM, 255);  
 #elif defined SUPPORT_UTF  
   if (common->utf)  
     jump = CMP(SLJIT_C_GREATER, TMP1, 0, SLJIT_IMM, 255);  
 #endif  
   
   if (!check_class_ranges(common, (const pcre_uint8 *)cc, TRUE, list))  
     {  
     OP2(SLJIT_AND, TMP2, 0, TMP1, 0, SLJIT_IMM, 0x7);  
     OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 3);  
     OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)cc);  
     OP2(SLJIT_SHL, TMP2, 0, SLJIT_IMM, 1, TMP2, 0);  
     OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);  
     add_jump(compiler, list, JUMP(SLJIT_C_NOT_ZERO));  
     }  
   
 #ifndef COMPILE_PCRE8  
   JUMPHERE(jump);  
 #elif defined SUPPORT_UTF  
   if (common->utf)  
     JUMPHERE(jump);  
 #endif  
   OP1(SLJIT_MOV, TMP1, 0, TMP3, 0);  
 #ifdef SUPPORT_UCP  
   charsaved = TRUE;  
 #endif  
4564    cc += 32 / sizeof(pcre_uchar);    cc += 32 / sizeof(pcre_uchar);
4565    }    }
4566    
 /* Scanning the necessary info. */  
 ccbegin = cc;  
 compares = 0;  
4567  while (*cc != XCL_END)  while (*cc != XCL_END)
4568    {    {
4569    compares++;    compares++;
4570    if (*cc == XCL_SINGLE)    if (*cc == XCL_SINGLE)
4571      {      {
4572      cc += 2;      cc ++;
4573  #ifdef SUPPORT_UTF      GETCHARINCTEST(c, cc);
4574      if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (c > max) max = c;
4575  #endif      if (c < min) min = c;
4576  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4577      needschar = TRUE;      needschar = TRUE;
4578  #endif  #endif
4579      }      }
4580    else if (*cc == XCL_RANGE)    else if (*cc == XCL_RANGE)
4581      {      {
4582      cc += 2;      cc ++;
4583  #ifdef SUPPORT_UTF      GETCHARINCTEST(c, cc);
4584      if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (c < min) min = c;
4585  #endif      GETCHARINCTEST(c, cc);
4586      cc++;      if (c > max) max = c;
 #ifdef SUPPORT_UTF  
     if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);  
 #endif  
4587  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4588      needschar = TRUE;      needschar = TRUE;
4589  #endif  #endif
# Line 3789  while (*cc != XCL_END) Line 4593  while (*cc != XCL_END)
4593      {      {
4594      SLJIT_ASSERT(*cc == XCL_PROP || *cc == XCL_NOTPROP);      SLJIT_ASSERT(*cc == XCL_PROP || *cc == XCL_NOTPROP);
4595      cc++;      cc++;
4596        if (*cc == PT_CLIST)
4597          {
4598          other_cases = PRIV(ucd_caseless_sets) + cc[1];
4599          while (*other_cases != NOTACHAR)
4600            {
4601            if (*other_cases > max) max = *other_cases;
4602            if (*other_cases < min) min = *other_cases;
4603            other_cases++;
4604            }
4605          }
4606        else
4607          {
4608          max = READ_CHAR_MAX;
4609          min = 0;
4610          }
4611    
4612      switch(*cc)      switch(*cc)
4613        {        {
4614        case PT_ANY:        case PT_ANY:
# Line 3808  while (*cc != XCL_END) Line 4628  while (*cc != XCL_END)
4628        case PT_SPACE:        case PT_SPACE:
4629        case PT_PXSPACE:        case PT_PXSPACE:
4630        case PT_WORD:        case PT_WORD:
4631          case PT_PXGRAPH:
4632          case PT_PXPRINT:
4633          case PT_PXPUNCT:
4634        needstype = TRUE;        needstype = TRUE;
4635        needschar = TRUE;        needschar = TRUE;
4636        break;        break;
4637    
4638        case PT_CLIST:        case PT_CLIST:
4639          case PT_UCNC:
4640        needschar = TRUE;        needschar = TRUE;
4641        break;        break;
4642    
# Line 3825  while (*cc != XCL_END) Line 4649  while (*cc != XCL_END)
4649  #endif  #endif
4650    }    }
4651    
4652    /* We are not necessary in utf mode even in 8 bit mode. */
4653    cc = ccbegin;
4654    detect_partial_match(common, backtracks);
4655    read_char_range(common, min, max, (cc[-1] & XCL_NOT) != 0);
4656    
4657    if ((cc[-1] & XCL_HASPROP) == 0)
4658      {
4659      if ((cc[-1] & XCL_MAP) != 0)
4660        {
4661        jump = CMP(SLJIT_C_GREATER, TMP1, 0, SLJIT_IMM, 255);
4662        if (!check_class_ranges(common, (const pcre_uint8 *)cc, (((const pcre_uint8 *)cc)[31] & 0x80) != 0, TRUE, &found))
4663          {
4664          OP2(SLJIT_AND, TMP2, 0, TMP1, 0, SLJIT_IMM, 0x7);
4665          OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 3);
4666          OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)cc);
4667          OP2(SLJIT_SHL, TMP2, 0, SLJIT_IMM, 1, TMP2, 0);
4668          OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);
4669          add_jump(compiler, &found, JUMP(SLJIT_C_NOT_ZERO));
4670          }
4671    
4672        add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));
4673        JUMPHERE(jump);
4674    
4675        cc += 32 / sizeof(pcre_uchar);
4676        }
4677      else
4678        {
4679        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, min);
4680        add_jump(compiler, (cc[-1] & XCL_NOT) == 0 ? backtracks : &found, CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, max - min));
4681        }
4682      }
4683    else if ((cc[-1] & XCL_MAP) != 0)
4684      {
4685      OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);
4686    #ifdef SUPPORT_UCP
4687      charsaved = TRUE;
4688    #endif
4689      if (!check_class_ranges(common, (const pcre_uint8 *)cc, FALSE, TRUE, list))
4690        {
4691    #ifdef COMPILE_PCRE8
4692        SLJIT_ASSERT(common->utf);
4693    #endif
4694        jump = CMP(SLJIT_C_GREATER, TMP1, 0, SLJIT_IMM, 255);
4695    
4696        OP2(SLJIT_AND, TMP2, 0, TMP1, 0, SLJIT_IMM, 0x7);
4697        OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 3);
4698        OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)cc);
4699        OP2(SLJIT_SHL, TMP2, 0, SLJIT_IMM, 1, TMP2, 0);
4700        OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);
4701        add_jump(compiler, list, JUMP(SLJIT_C_NOT_ZERO));
4702    
4703        JUMPHERE(jump);
4704        }
4705    
4706      OP1(SLJIT_MOV, TMP1, 0, TMP3, 0);
4707      cc += 32 / sizeof(pcre_uchar);
4708      }
4709    
4710  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4711  /* Simple register allocation. TMP1 is preferred if possible. */  /* Simple register allocation. TMP1 is preferred if possible. */
4712  if (needstype || needsscript)  if (needstype || needsscript)
# Line 3866  if (needstype || needsscript) Line 4748  if (needstype || needsscript)
4748  #endif  #endif
4749    
4750  /* Generating code. */  /* Generating code. */
 cc = ccbegin;  
4751  charoffset = 0;  charoffset = 0;
4752  numberofcmps = 0;  numberofcmps = 0;
4753  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 3882  while (*cc != XCL_END) Line 4763  while (*cc != XCL_END)
4763    if (*cc == XCL_SINGLE)    if (*cc == XCL_SINGLE)
4764      {      {
4765      cc ++;      cc ++;
4766  #ifdef SUPPORT_UTF      GETCHARINCTEST(c, cc);
     if (common->utf)  
       {  
       GETCHARINC(c, cc);  
       }  
     else  
 #endif  
       c = *cc++;  
4767    
4768      if (numberofcmps < 3 && (*cc == XCL_SINGLE || *cc == XCL_RANGE))      if (numberofcmps < 3 && (*cc == XCL_SINGLE || *cc == XCL_RANGE))
4769        {        {
4770        OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, c - charoffset);        OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(c - charoffset));
4771        OP_FLAGS(numberofcmps == 0 ? SLJIT_MOV : SLJIT_OR, TMP2, 0, numberofcmps == 0 ? SLJIT_UNUSED : TMP2, 0, SLJIT_C_EQUAL);        OP_FLAGS(numberofcmps == 0 ? SLJIT_MOV : SLJIT_OR, TMP2, 0, numberofcmps == 0 ? SLJIT_UNUSED : TMP2, 0, SLJIT_C_EQUAL);
4772        numberofcmps++;        numberofcmps++;
4773        }        }
4774      else if (numberofcmps > 0)      else if (numberofcmps > 0)
4775        {        {
4776        OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, c - charoffset);        OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(c - charoffset));
4777        OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);        OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);
4778        jump = JUMP(SLJIT_C_NOT_ZERO ^ invertcmp);        jump = JUMP(SLJIT_C_NOT_ZERO ^ invertcmp);
4779        numberofcmps = 0;        numberofcmps = 0;
4780        }        }
4781      else      else
4782        {        {
4783        jump = CMP(SLJIT_C_EQUAL ^ invertcmp, TMP1, 0, SLJIT_IMM, c - charoffset);        jump = CMP(SLJIT_C_EQUAL ^ invertcmp, TMP1, 0, SLJIT_IMM, (sljit_sw)(c - charoffset));
4784        numberofcmps = 0;        numberofcmps = 0;
4785        }        }
4786      }      }
4787    else if (*cc == XCL_RANGE)    else if (*cc == XCL_RANGE)
4788      {      {
4789      cc ++;      cc ++;
4790  #ifdef SUPPORT_UTF      GETCHARINCTEST(c, cc);
     if (common->utf)  
       {  
       GETCHARINC(c, cc);  
       }  
     else  
 #endif  
       c = *cc++;  
4791      SET_CHAR_OFFSET(c);      SET_CHAR_OFFSET(c);
4792  #ifdef SUPPORT_UTF      GETCHARINCTEST(c, cc);
4793      if (common->utf)  
       {  
       GETCHARINC(c, cc);  
       }  
     else  
 #endif  
       c = *cc++;  
4794      if (numberofcmps < 3 && (*cc == XCL_SINGLE || *cc == XCL_RANGE))      if (numberofcmps < 3 && (*cc == XCL_SINGLE || *cc == XCL_RANGE))
4795        {        {
4796        OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, c - charoffset);        OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(c - charoffset));
4797        OP_FLAGS(numberofcmps == 0 ? SLJIT_MOV : SLJIT_OR, TMP2, 0, numberofcmps == 0 ? SLJIT_UNUSED : TMP2, 0, SLJIT_C_LESS_EQUAL);        OP_FLAGS(numberofcmps == 0 ? SLJIT_MOV : SLJIT_OR, TMP2, 0, numberofcmps == 0 ? SLJIT_UNUSED : TMP2, 0, SLJIT_C_LESS_EQUAL);
4798        numberofcmps++;        numberofcmps++;
4799        }        }
4800      else if (numberofcmps > 0)      else if (numberofcmps > 0)
4801        {        {
4802        OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, c - charoffset);        OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(c - charoffset));
4803        OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_C_LESS_EQUAL);        OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_C_LESS_EQUAL);
4804        jump = JUMP(SLJIT_C_NOT_ZERO ^ invertcmp);        jump = JUMP(SLJIT_C_NOT_ZERO ^ invertcmp);
4805        numberofcmps = 0;        numberofcmps = 0;
4806        }        }
4807      else      else
4808        {        {
4809        jump = CMP(SLJIT_C_LESS_EQUAL ^ invertcmp, TMP1, 0, SLJIT_IMM, c - charoffset);        jump = CMP(SLJIT_C_LESS_EQUAL ^ invertcmp, TMP1, 0, SLJIT_IMM, (sljit_sw)(c - charoffset));
4810        numberofcmps = 0;        numberofcmps = 0;
4811        }        }
4812      }      }
# Line 3994  while (*cc != XCL_END) Line 4855  while (*cc != XCL_END)
4855    
4856        case PT_SPACE:        case PT_SPACE:
4857        case PT_PXSPACE:        case PT_PXSPACE:
       if (*cc == PT_SPACE)  
         {  
         OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 0);  
         jump = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, 11 - charoffset);  
         }  
4858        SET_CHAR_OFFSET(9);        SET_CHAR_OFFSET(9);
4859        OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 13 - 9);        OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd - 0x9);
4860        OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_LESS_EQUAL);        OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_LESS_EQUAL);
4861        if (*cc == PT_SPACE)  
4862          JUMPHERE(jump);        OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x85 - 0x9);
4863          OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);
4864    
4865          OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x180e - 0x9);
4866          OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);
4867    
4868        SET_TYPE_OFFSET(ucp_Zl);        SET_TYPE_OFFSET(ucp_Zl);
4869        OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_Zs - ucp_Zl);        OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_Zs - ucp_Zl);
# Line 4012  while (*cc != XCL_END) Line 4872  while (*cc != XCL_END)
4872        break;        break;
4873    
4874        case PT_WORD:        case PT_WORD:
4875        OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_UNDERSCORE - charoffset);        OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(CHAR_UNDERSCORE - charoffset));
4876        OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);        OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
4877        /* ... fall through */        /* Fall through. */
4878    
4879        case PT_ALNUM:        case PT_ALNUM:
4880        SET_TYPE_OFFSET(ucp_Ll);        SET_TYPE_OFFSET(ucp_Ll);
# Line 4060  while (*cc != XCL_END) Line 4920  while (*cc != XCL_END)
4920          OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, other_cases[2]);          OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, other_cases[2]);
4921          OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);          OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
4922    
4923          OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, other_cases[0] - charoffset);          OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(other_cases[0] - charoffset));
4924          OP_FLAGS(SLJIT_OR | ((other_cases[3] == NOTACHAR) ? SLJIT_SET_E : 0), TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);          OP_FLAGS(SLJIT_OR | ((other_cases[3] == NOTACHAR) ? SLJIT_SET_E : 0), TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);
4925    
4926          other_cases += 3;          other_cases += 3;
4927          }          }
4928        else        else
4929          {          {
4930          OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, *other_cases++ - charoffset);          OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(*other_cases++ - charoffset));
4931          OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);          OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
4932          }          }
4933    
4934        while (*other_cases != NOTACHAR)        while (*other_cases != NOTACHAR)
4935          {          {
4936          OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, *other_cases++ - charoffset);          OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(*other_cases++ - charoffset));
4937          OP_FLAGS(SLJIT_OR | ((*other_cases == NOTACHAR) ? SLJIT_SET_E : 0), TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);          OP_FLAGS(SLJIT_OR | ((*other_cases == NOTACHAR) ? SLJIT_SET_E : 0), TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);
4938          }          }
4939        jump = JUMP(SLJIT_C_NOT_ZERO ^ invertcmp);        jump = JUMP(SLJIT_C_NOT_ZERO ^ invertcmp);
4940        break;        break;
4941    
4942          case PT_UCNC:
4943          OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(CHAR_DOLLAR_SIGN - charoffset));
4944          OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
4945          OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(CHAR_COMMERCIAL_AT - charoffset));
4946          OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);
4947          OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(CHAR_GRAVE_ACCENT - charoffset));
4948          OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);
4949    
4950          SET_CHAR_OFFSET(0xa0);
4951          OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(0xd7ff - charoffset));
4952          OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_C_LESS_EQUAL);
4953          SET_CHAR_OFFSET(0);
4954          OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xe000 - 0);
4955          OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_C_GREATER_EQUAL);
4956          jump = JUMP(SLJIT_C_NOT_ZERO ^ invertcmp);
4957          break;
4958    
4959          case PT_PXGRAPH:
4960          /* C and Z groups are the farthest two groups. */
4961          SET_TYPE_OFFSET(ucp_Ll);
4962          OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_So - ucp_Ll);
4963          OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_GREATER);
4964    
4965          jump = CMP(SLJIT_C_NOT_EQUAL, typereg, 0, SLJIT_IMM, ucp_Cf - ucp_Ll);
4966    
4967          /* In case of ucp_Cf, we overwrite the result. */
4968          SET_CHAR_OFFSET(0x2066);
4969          OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x2069 - 0x2066);
4970          OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_LESS_EQUAL);
4971    
4972          OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x061c - 0x2066);
4973          OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);
4974    
4975          OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x180e - 0x2066);
4976          OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);
4977    
4978          JUMPHERE(jump);
4979          jump = CMP(SLJIT_C_ZERO ^ invertcmp, TMP2, 0, SLJIT_IMM, 0);
4980          break;
4981    
4982          case PT_PXPRINT:
4983          /* C and Z groups are the farthest two groups. */
4984          SET_TYPE_OFFSET(ucp_Ll);
4985          OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_So - ucp_Ll);
4986          OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_GREATER);
4987    
4988          OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_Zs - ucp_Ll);
4989          OP_FLAGS(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_C_NOT_EQUAL);
4990    
4991          jump = CMP(SLJIT_C_NOT_EQUAL, typereg, 0, SLJIT_IMM, ucp_Cf - ucp_Ll);
4992    
4993          /* In case of ucp_Cf, we overwrite the result. */
4994          SET_CHAR_OFFSET(0x2066);
4995          OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x2069 - 0x2066);
4996          OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_LESS_EQUAL);
4997    
4998          OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x061c - 0x2066);
4999          OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);
5000    
5001          JUMPHERE(jump);
5002          jump = CMP(SLJIT_C_ZERO ^ invertcmp, TMP2, 0, SLJIT_IMM, 0);
5003          break;
5004    
5005          case PT_PXPUNCT:
5006          SET_TYPE_OFFSET(ucp_Sc);
5007          OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_So - ucp_Sc);
5008          OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_LESS_EQUAL);
5009    
5010          SET_CHAR_OFFSET(0);
5011          OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xff);
5012          OP_FLAGS(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_C_LESS_EQUAL);
5013    
5014          SET_TYPE_OFFSET(ucp_Pc);
5015          OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_Ps - ucp_Pc);
5016          OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_C_LESS_EQUAL);
5017          jump = JUMP(SLJIT_C_NOT_ZERO ^ invertcmp);
5018          break;
5019        }        }
5020      cc += 2;      cc += 2;
5021      }      }
# Line 4103  int length; Line 5041  int length;
5041  unsigned int c, oc, bit;  unsigned int c, oc, bit;
5042  compare_context context;  compare_context context;
5043  struct sljit_jump *jump[4];  struct sljit_jump *jump[4];
5044    jump_list *end_list;
5045  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
5046  struct sljit_label *label;  struct sljit_label *label;
5047  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
5048  pcre_uchar propdata[5];  pcre_uchar propdata[5];
5049  #endif  #endif
5050  #endif  #endif /* SUPPORT_UTF */
5051    
5052  switch(type)  switch(type)
5053    {    {
# Line 4133  switch(type) Line 5072  switch(type)
5072    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
5073    case OP_DIGIT:    case OP_DIGIT:
5074    /* Digits are usually 0-9, so it is worth to optimize them. */    /* Digits are usually 0-9, so it is worth to optimize them. */
   if (common->digits[0] == -2)  
     get_ctype_ranges(common, ctype_digit, common->digits);  
5075    detect_partial_match(common, backtracks);    detect_partial_match(common, backtracks);
5076    /* Flip the starting bit in the negative case. */  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
5077    if (type == OP_NOT_DIGIT)    if (common->utf && is_char7_bitset((const pcre_uint8*)common->ctypes - cbit_length + cbit_digit, FALSE))
5078      common->digits[1] ^= 1;      read_char7_type(common, type == OP_NOT_DIGIT);
5079    if (!check_ranges(common, common->digits, backtracks, TRUE))    else
5080      {  #endif
5081      read_char8_type(common);      read_char8_type(common, type == OP_NOT_DIGIT);
5082      OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ctype_digit);      /* Flip the starting bit in the negative case. */
5083      add_jump(compiler, backtracks, JUMP(type == OP_DIGIT ? SLJIT_C_ZERO : SLJIT_C_NOT_ZERO));    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ctype_digit);
5084      }    add_jump(compiler, backtracks, JUMP(type == OP_DIGIT ? SLJIT_C_ZERO : SLJIT_C_NOT_ZERO));
   if (type == OP_NOT_DIGIT)  
     common->digits[1] ^= 1;  
5085    return cc;    return cc;
5086    
5087    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
5088    case OP_WHITESPACE:    case OP_WHITESPACE:
5089    detect_partial_match(common, backtracks);    detect_partial_match(common, backtracks);
5090    read_char8_type(common);  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
5091      if (common->utf && is_char7_bitset((const pcre_uint8*)common->ctypes - cbit_length + cbit_space, FALSE))
5092        read_char7_type(common, type == OP_NOT_WHITESPACE);
5093      else
5094    #endif
5095        read_char8_type(common, type == OP_NOT_WHITESPACE);
5096    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ctype_space);    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ctype_space);
5097    add_jump(compiler, backtracks, JUMP(type == OP_WHITESPACE ? SLJIT_C_ZERO : SLJIT_C_NOT_ZERO));    add_jump(compiler, backtracks, JUMP(type == OP_WHITESPACE ? SLJIT_C_ZERO : SLJIT_C_NOT_ZERO));
5098    return cc;    return cc;
# Line 4160  switch(type) Line 5100  switch(type)
5100    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
5101    case OP_WORDCHAR:    case OP_WORDCHAR:
5102    detect_partial_match(common, backtracks);    detect_partial_match(common, backtracks);
5103    read_char8_type(common);  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
5104      if (common->utf && is_char7_bitset((const pcre_uint8*)common->ctypes - cbit_length + cbit_word, FALSE))
5105        read_char7_type(common, type == OP_NOT_WORDCHAR);
5106      else
5107    #endif
5108        read_char8_type(common, type == OP_NOT_WORDCHAR);
5109    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ctype_word);    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ctype_word);
5110    add_jump(compiler, backtracks, JUMP(type == OP_WORDCHAR ? SLJIT_C_ZERO : SLJIT_C_NOT_ZERO));    add_jump(compiler, backtracks, JUMP(type == OP_WORDCHAR ? SLJIT_C_ZERO : SLJIT_C_NOT_ZERO));
5111    return cc;    return cc;
5112    
5113    case OP_ANY:    case OP_ANY:
5114    detect_partial_match(common, backtracks);    detect_partial_match(common, backtracks);
5115    read_char(common);    read_char_range(common, common->nlmin, common->nlmax, TRUE);
5116    if (common->nltype == NLTYPE_FIXED && common->newline > 255)    if (common->nltype == NLTYPE_FIXED && common->newline > 255)
5117      {      {
5118      jump[0] = CMP(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff);      jump[0] = CMP(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff);
5119        end_list = NULL;
5120      if (common->mode != JIT_PARTIAL_HARD_COMPILE)      if (common->mode != JIT_PARTIAL_HARD_COMPILE)
5121        jump[1] = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);        add_jump(compiler, &end_list, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
5122      else      else
5123        jump[1] = check_str_end(common);        check_str_end(common, &end_list);
5124    
5125      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
5126      add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, common->newline & 0xff));      add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, common->newline & 0xff));
5127      if (jump[1] != NULL)      set_jumps(end_list, LABEL());
       JUMPHERE(jump[1]);  
5128      JUMPHERE(jump[0]);      JUMPHERE(jump[0]);
5129      }      }
5130    else    else
# Line 4223  switch(type) Line 5168  switch(type)
5168  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
5169    case OP_NOTPROP:    case OP_NOTPROP:
5170    case OP_PROP:    case OP_PROP:
5171    propdata[0] = 0;    propdata[0] = XCL_HASPROP;
5172    propdata[1] = type == OP_NOTPROP ? XCL_NOTPROP : XCL_PROP;    propdata[1] = type == OP_NOTPROP ? XCL_NOTPROP : XCL_PROP;
5173    propdata[2] = cc[0];    propdata[2] = cc[0];
5174    propdata[3] = cc[1];    propdata[3] = cc[1];
# Line 4235  switch(type) Line 5180  switch(type)
5180    
5181    case OP_ANYNL:    case OP_ANYNL:
5182    detect_partial_match(common, backtracks);    detect_partial_match(common, backtracks);
5183    read_char(common);    read_char_range(common, common->bsr_nlmin, common->bsr_nlmax, FALSE);
5184    jump[0] = CMP(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR);    jump[0] = CMP(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR);
5185    /* We don't need to handle soft partial matching case. */    /* We don't need to handle soft partial matching case. */
5186      end_list = NULL;
5187    if (common->mode != JIT_PARTIAL_HARD_COMPILE)    if (common->mode != JIT_PARTIAL_HARD_COMPILE)
5188      jump[1] = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);      add_jump(compiler, &end_list, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
5189    else    else
5190      jump[1] = check_str_end(common);      check_str_end(common, &end_list);
5191    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
5192    jump[2] = CMP(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL);    jump[1] = CMP(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL);
5193    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));
5194    jump[3] = JUMP(SLJIT_JUMP);    jump[2] = JUMP(SLJIT_JUMP);
5195    JUMPHERE(jump[0]);    JUMPHERE(jump[0]);
5196    check_newlinechar(common, common->bsr_nltype, backtracks, FALSE);    check_newlinechar(common, common->bsr_nltype, backtracks, FALSE);
5197      set_jumps(end_list, LABEL());
5198    JUMPHERE(jump[1]);    JUMPHERE(jump[1]);
5199    JUMPHERE(jump[2]);    JUMPHERE(jump[2]);
   JUMPHERE(jump[3]);  
5200    return cc;    return cc;