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

Diff of /code/trunk/pcre_jit_compile.c

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

revision 1244 by zherczeg, Tue Feb 5 12:03:43 2013 UTC revision 1452 by zherczeg, Tue Jan 28 16:07:52 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      struct sljit_label *ff_newline_shortcut;
402    stub_list *stubs;    stub_list *stubs;
403      label_addr_list *label_addrs;
404    recurse_entry *entries;    recurse_entry *entries;
405    recurse_entry *currententry;    recurse_entry *currententry;
406    jump_list *partialmatch;    jump_list *partialmatch;
407    jump_list *quit;    jump_list *quit;
408      jump_list *positive_assert_quit;
409      jump_list *forced_quit;
410    jump_list *accept;    jump_list *accept;
411    jump_list *calllimit;    jump_list *calllimit;
412    jump_list *stackalloc;    jump_list *stackalloc;
# Line 338  typedef struct compiler_common { Line 417  typedef struct compiler_common {
417    jump_list *vspace;    jump_list *vspace;
418    jump_list *casefulcmp;    jump_list *casefulcmp;
419    jump_list *caselesscmp;    jump_list *caselesscmp;
420      jump_list *reset_match;
421    BOOL jscript_compat;    BOOL jscript_compat;
422  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
423    BOOL utf;    BOOL utf;
424  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
425    BOOL use_ucp;    BOOL use_ucp;
426  #endif  #endif
 #ifndef COMPILE_PCRE32  
   jump_list *utfreadchar;  
 #endif  
427  #ifdef COMPILE_PCRE8  #ifdef COMPILE_PCRE8
428      jump_list *utfreadchar;
429      jump_list *utfreadchar16;
430    jump_list *utfreadtype8;    jump_list *utfreadtype8;
431  #endif  #endif
432  #endif /* SUPPORT_UTF */  #endif /* SUPPORT_UTF */
# Line 390  typedef struct compare_context { Line 469  typedef struct compare_context {
469  #endif  #endif
470  } compare_context;  } compare_context;
471    
 enum {  
   frame_end = 0,  
   frame_setstrbegin = -1,  
   frame_setmark = -2  
 };  
   
472  /* Undefine sljit macros. */  /* Undefine sljit macros. */
473  #undef CMP  #undef CMP
474    
# Line 410  enum { Line 483  enum {
483  #define STACK_TOP     SLJIT_SCRATCH_REG2  #define STACK_TOP     SLJIT_SCRATCH_REG2
484  #define STACK_LIMIT   SLJIT_SAVED_REG3  #define STACK_LIMIT   SLJIT_SAVED_REG3
485  #define ARGUMENTS     SLJIT_SAVED_EREG1  #define ARGUMENTS     SLJIT_SAVED_EREG1
486  #define CALL_COUNT    SLJIT_SAVED_EREG2  #define COUNT_MATCH   SLJIT_SAVED_EREG2
487  #define RETURN_ADDR   SLJIT_TEMPORARY_EREG1  #define RETURN_ADDR   SLJIT_TEMPORARY_EREG1
488    
489  /* Local space layout. */  /* Local space layout. */
# Line 421  enum { Line 494  enum {
494  #define POSSESSIVE0      (2 * sizeof(sljit_sw))  #define POSSESSIVE0      (2 * sizeof(sljit_sw))
495  #define POSSESSIVE1      (3 * sizeof(sljit_sw))  #define POSSESSIVE1      (3 * sizeof(sljit_sw))
496  /* Max limit of recursions. */  /* Max limit of recursions. */
497  #define CALL_LIMIT       (4 * sizeof(sljit_sw))  #define LIMIT_MATCH      (4 * sizeof(sljit_sw))
498  /* The output vector is stored on the stack, and contains pointers  /* The output vector is stored on the stack, and contains pointers
499  to characters. The vector data is divided into two groups: the first  to characters. The vector data is divided into two groups: the first
500  group contains the start / end character pointers, and the second is  group contains the start / end character pointers, and the second is
501  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. */
502  #define OVECTOR_START    (common->ovector_start)  #define OVECTOR_START    (common->ovector_start)
503  #define OVECTOR(i)       (OVECTOR_START + (i) * sizeof(sljit_sw))  #define OVECTOR(i)       (OVECTOR_START + (i) * (sljit_sw)sizeof(sljit_sw))
504  #define OVECTOR_PRIV(i)  (common->cbraptr + (i) * sizeof(sljit_sw))  #define OVECTOR_PRIV(i)  (common->cbra_ptr + (i) * (sljit_sw)sizeof(sljit_sw))
505  #define PRIVATE_DATA(cc) (common->private_data_ptrs[(cc) - common->start])  #define PRIVATE_DATA(cc) (common->private_data_ptrs[(cc) - common->start])
506    
507  #if defined COMPILE_PCRE8  #if defined COMPILE_PCRE8
# Line 459  the start pointers when the end of the c Line 532  the start pointers when the end of the c
532    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))
533  #define JUMPHERE(jump) \  #define JUMPHERE(jump) \
534    sljit_set_label((jump), sljit_emit_label(compiler))    sljit_set_label((jump), sljit_emit_label(compiler))
535    #define SET_LABEL(jump, label) \
536      sljit_set_label((jump), (label))
537  #define CMP(type, src1, src1w, src2, src2w) \  #define CMP(type, src1, src1w, src2, src2w) \
538    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))
539  #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 543  the start pointers when the end of the c
543  #define GET_LOCAL_BASE(dst, dstw, offset) \  #define GET_LOCAL_BASE(dst, dstw, offset) \
544    sljit_get_local_base(compiler, (dst), (dstw), (offset))    sljit_get_local_base(compiler, (dst), (dstw), (offset))
545    
546    #define READ_CHAR_MAX 0x7fffffff
547    
548  static pcre_uchar* bracketend(pcre_uchar* cc)  static pcre_uchar* bracketend(pcre_uchar* cc)
549  {  {
550  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 554  cc += 1 + LINK_SIZE;
554  return cc;  return cc;
555  }  }
556    
557    static int no_alternatives(pcre_uchar* cc)
558    {
559    int count = 0;
560    SLJIT_ASSERT((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND));
561    do
562      {
563      cc += GET(cc, 1);
564      count++;
565      }
566    while (*cc == OP_ALT);
567    SLJIT_ASSERT(*cc >= OP_KET && *cc <= OP_KETRPOS);
568    return count;
569    }
570    
571    static int ones_in_half_byte[16] = {
572      /* 0 */ 0, 1, 1, 2, /* 4 */ 1, 2, 2, 3,
573      /* 8 */ 1, 2, 2, 3, /* 12 */ 2, 3, 3, 4
574    };
575    
576  /* Functions whose might need modification for all new supported opcodes:  /* Functions whose might need modification for all new supported opcodes:
577   next_opcode   next_opcode
578   get_private_data_length   check_opcode_types
579   set_private_data_ptrs   set_private_data_ptrs
580   get_framesize   get_framesize
581   init_frame   init_frame
582   get_private_data_length_for_copy   get_private_data_copy_length
583   copy_private_data   copy_private_data
584   compile_matchingpath   compile_matchingpath
585   compile_backtrackingpath   compile_backtrackingpath
# Line 507  switch(*cc) Line 603  switch(*cc)
603    case OP_WORDCHAR:    case OP_WORDCHAR:
604    case OP_ANY:    case OP_ANY:
605    case OP_ALLANY:    case OP_ALLANY:
606      case OP_NOTPROP:
607      case OP_PROP:
608    case OP_ANYNL:    case OP_ANYNL:
609    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
610    case OP_HSPACE:    case OP_HSPACE:
# Line 519  switch(*cc) Line 617  switch(*cc)
617    case OP_CIRCM:    case OP_CIRCM:
618    case OP_DOLL:    case OP_DOLL:
619    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:  
620    case OP_CRSTAR:    case OP_CRSTAR:
621    case OP_CRMINSTAR:    case OP_CRMINSTAR:
622    case OP_CRPLUS:    case OP_CRPLUS:
623    case OP_CRMINPLUS:    case OP_CRMINPLUS:
624    case OP_CRQUERY:    case OP_CRQUERY:
625    case OP_CRMINQUERY:    case OP_CRMINQUERY:
626      case OP_CRRANGE:
627      case OP_CRMINRANGE:
628      case OP_CRPOSSTAR:
629      case OP_CRPOSPLUS:
630      case OP_CRPOSQUERY:
631      case OP_CRPOSRANGE:
632      case OP_CLASS:
633      case OP_NCLASS:
634      case OP_REF:
635      case OP_REFI:
636      case OP_DNREF:
637      case OP_DNREFI:
638      case OP_RECURSE:
639      case OP_CALLOUT:
640      case OP_ALT:
641      case OP_KET:
642      case OP_KETRMAX:
643      case OP_KETRMIN:
644      case OP_KETRPOS:
645      case OP_REVERSE:
646      case OP_ASSERT:
647      case OP_ASSERT_NOT:
648      case OP_ASSERTBACK:
649      case OP_ASSERTBACK_NOT:
650      case OP_ONCE:
651      case OP_ONCE_NC:
652      case OP_BRA:
653      case OP_BRAPOS:
654      case OP_CBRA:
655      case OP_CBRAPOS:
656      case OP_COND:
657      case OP_SBRA:
658      case OP_SBRAPOS:
659      case OP_SCBRA:
660      case OP_SCBRAPOS:
661      case OP_SCOND:
662      case OP_CREF:
663      case OP_DNCREF:
664      case OP_RREF:
665      case OP_DNRREF:
666    case OP_DEF:    case OP_DEF:
667    case OP_BRAZERO:    case OP_BRAZERO:
668    case OP_BRAMINZERO:    case OP_BRAMINZERO:
669    case OP_BRAPOSZERO:    case OP_BRAPOSZERO:
670      case OP_PRUNE:
671      case OP_SKIP:
672      case OP_THEN:
673    case OP_COMMIT:    case OP_COMMIT:
674    case OP_FAIL:    case OP_FAIL:
675    case OP_ACCEPT:    case OP_ACCEPT:
676    case OP_ASSERT_ACCEPT:    case OP_ASSERT_ACCEPT:
677      case OP_CLOSE:
678    case OP_SKIPZERO:    case OP_SKIPZERO:
679    return cc + 1;    return cc + PRIV(OP_lengths)[*cc];
   
   case OP_ANYBYTE:  
 #ifdef SUPPORT_UTF  
   if (common->utf) return NULL;  
 #endif  
   return cc + 1;  
680    
681    case OP_CHAR:    case OP_CHAR:
682    case OP_CHARI:    case OP_CHARI:
# Line 561  switch(*cc) Line 688  switch(*cc)
688    case OP_MINPLUS:    case OP_MINPLUS:
689    case OP_QUERY:    case OP_QUERY:
690    case OP_MINQUERY:    case OP_MINQUERY:
691      case OP_UPTO:
692      case OP_MINUPTO:
693      case OP_EXACT:
694    case OP_POSSTAR:    case OP_POSSTAR:
695    case OP_POSPLUS:    case OP_POSPLUS:
696    case OP_POSQUERY:    case OP_POSQUERY:
697      case OP_POSUPTO:
698    case OP_STARI:    case OP_STARI:
699    case OP_MINSTARI:    case OP_MINSTARI:
700    case OP_PLUSI:    case OP_PLUSI:
701    case OP_MINPLUSI:    case OP_MINPLUSI:
702    case OP_QUERYI:    case OP_QUERYI:
703    case OP_MINQUERYI:    case OP_MINQUERYI:
704      case OP_UPTOI:
705      case OP_MINUPTOI:
706      case OP_EXACTI:
707    case OP_POSSTARI:    case OP_POSSTARI:
708    case OP_POSPLUSI:    case OP_POSPLUSI:
709    case OP_POSQUERYI:    case OP_POSQUERYI:
710      case OP_POSUPTOI:
711    case OP_NOTSTAR:    case OP_NOTSTAR:
712    case OP_NOTMINSTAR:    case OP_NOTMINSTAR:
713    case OP_NOTPLUS:    case OP_NOTPLUS:
714    case OP_NOTMINPLUS:    case OP_NOTMINPLUS:
715    case OP_NOTQUERY:    case OP_NOTQUERY:
716    case OP_NOTMINQUERY:    case OP_NOTMINQUERY:
717      case OP_NOTUPTO:
718      case OP_NOTMINUPTO:
719      case OP_NOTEXACT:
720    case OP_NOTPOSSTAR:    case OP_NOTPOSSTAR:
721    case OP_NOTPOSPLUS:    case OP_NOTPOSPLUS:
722    case OP_NOTPOSQUERY:    case OP_NOTPOSQUERY:
723      case OP_NOTPOSUPTO:
724    case OP_NOTSTARI:    case OP_NOTSTARI:
725    case OP_NOTMINSTARI:    case OP_NOTMINSTARI:
726    case OP_NOTPLUSI:    case OP_NOTPLUSI:
727    case OP_NOTMINPLUSI:    case OP_NOTMINPLUSI:
728    case OP_NOTQUERYI:    case OP_NOTQUERYI:
729    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:  
730    case OP_NOTUPTOI:    case OP_NOTUPTOI:
731    case OP_NOTMINUPTOI:    case OP_NOTMINUPTOI:
732    case OP_NOTEXACTI:    case OP_NOTEXACTI:
733      case OP_NOTPOSSTARI:
734      case OP_NOTPOSPLUSI:
735      case OP_NOTPOSQUERYI:
736    case OP_NOTPOSUPTOI:    case OP_NOTPOSUPTOI:
737    cc += 2 + IMM2_SIZE;    cc += PRIV(OP_lengths)[*cc];
738  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
739    if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);    if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
740  #endif  #endif
741    return cc;    return cc;
742    
743    case OP_NOTPROP:    /* Special cases. */
744    case OP_PROP:    case OP_TYPESTAR:
745    return cc + 1 + 2;    case OP_TYPEMINSTAR:
746      case OP_TYPEPLUS:
747      case OP_TYPEMINPLUS:
748      case OP_TYPEQUERY:
749      case OP_TYPEMINQUERY:
750    case OP_TYPEUPTO:    case OP_TYPEUPTO:
751    case OP_TYPEMINUPTO:    case OP_TYPEMINUPTO:
752    case OP_TYPEEXACT:    case OP_TYPEEXACT:
753      case OP_TYPEPOSSTAR:
754      case OP_TYPEPOSPLUS:
755      case OP_TYPEPOSQUERY:
756    case OP_TYPEPOSUPTO:    case OP_TYPEPOSUPTO:
757    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;  
758    
759    case OP_CRRANGE:    case OP_ANYBYTE:
760    case OP_CRMINRANGE:  #ifdef SUPPORT_UTF
761    return cc + 1 + 2 * IMM2_SIZE;    if (common->utf) return NULL;
762    #endif
763    case OP_CLASS:    return cc + 1;
   case OP_NCLASS:  
   return cc + 1 + 32 / sizeof(pcre_uchar);  
764    
765  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
766    case OP_XCLASS:    case OP_XCLASS:
767    return cc + GET(cc, 1);    return cc + GET(cc, 1);
768  #endif  #endif
769    
   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;  
   
770    case OP_MARK:    case OP_MARK:
771      case OP_PRUNE_ARG:
772      case OP_SKIP_ARG:
773      case OP_THEN_ARG:
774    return cc + 1 + 2 + cc[1];    return cc + 1 + 2 + cc[1];
775    
776    default:    default:
777      /* All opcodes are supported now! */
778      SLJIT_ASSERT_STOP();
779    return NULL;    return NULL;
780    }    }
781  }  }
782    
783  #define CASE_ITERATOR_PRIVATE_DATA_1 \  static BOOL check_opcode_types(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend)
784      case OP_MINSTAR: \  {
785      case OP_MINPLUS: \  int count;
786      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:  
787    
788  #define CASE_ITERATOR_PRIVATE_DATA_2A \  /* Calculate important variables (like stack size) and checks whether all opcodes are supported. */
789      case OP_STAR: \  while (cc < ccend)
790      case OP_PLUS: \    {
791      case OP_STARI: \    switch(*cc)
792      case OP_PLUSI: \      {
793      case OP_NOTSTAR: \      case OP_SET_SOM:
794      case OP_NOTPLUS: \      common->has_set_som = TRUE;
795      case OP_NOTSTARI: \      common->might_be_empty = TRUE;
796      case OP_NOTPLUSI:      cc += 1;
797        break;
798    
799  #define CASE_ITERATOR_PRIVATE_DATA_2B \      case OP_REF:
800      case OP_UPTO: \      case OP_REFI:
801      case OP_MINUPTO: \      common->optimized_cbracket[GET2(cc, 1)] = 0;
802      case OP_UPTOI: \      cc += 1 + IMM2_SIZE;
803      case OP_MINUPTOI: \      break;
     case OP_NOTUPTO: \  
     case OP_NOTMINUPTO: \  
     case OP_NOTUPTOI: \  
     case OP_NOTMINUPTOI:  
804    
805  #define CASE_ITERATOR_TYPE_PRIVATE_DATA_1 \      case OP_BRA:
806      case OP_TYPEMINSTAR: \      case OP_CBRA:
807      case OP_TYPEMINPLUS: \      case OP_SBRA:
808      case OP_TYPEQUERY: \      case OP_SCBRA:
809      case OP_TYPEMINQUERY:      count = no_alternatives(cc);
810        if (count > 4)
811          common->read_only_data_size += count * sizeof(sljit_uw);
812        cc += 1 + LINK_SIZE + (*cc == OP_CBRA || *cc == OP_SCBRA ? IMM2_SIZE : 0);
813        break;
814    
815  #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2A \      case OP_CBRAPOS:
816      case OP_TYPESTAR: \      case OP_SCBRAPOS:
817      case OP_TYPEPLUS:      common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] = 0;
818        cc += 1 + LINK_SIZE + IMM2_SIZE;
819        break;
820    
821  #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2B \      case OP_COND:
822      case OP_TYPEUPTO: \      case OP_SCOND:
823      case OP_TYPEMINUPTO:      /* Only AUTO_CALLOUT can insert this opcode. We do
824           not intend to support this case. */
825        if (cc[1 + LINK_SIZE] == OP_CALLOUT)
826          return FALSE;
827        cc += 1 + LINK_SIZE;
828        break;
829    
830  static int get_class_iterator_size(pcre_uchar *cc)      case OP_CREF:
831  {      common->optimized_cbracket[GET2(cc, 1)] = 0;
832        cc += 1 + IMM2_SIZE;
833        break;
834    
835        case OP_DNREF:
836        case OP_DNREFI:
837        case OP_DNCREF:
838        count = GET2(cc, 1 + IMM2_SIZE);
839        slot = common->name_table + GET2(cc, 1) * common->name_entry_size;
840        while (count-- > 0)
841          {
842          common->optimized_cbracket[GET2(slot, 0)] = 0;
843          slot += common->name_entry_size;
844          }
845        cc += 1 + 2 * IMM2_SIZE;
846        break;
847    
848        case OP_RECURSE:
849        /* Set its value only once. */
850        if (common->recursive_head_ptr == 0)
851          {
852          common->recursive_head_ptr = common->ovector_start;
853          common->ovector_start += sizeof(sljit_sw);
854          }
855        cc += 1 + LINK_SIZE;
856        break;
857    
858        case OP_CALLOUT:
859        if (common->capture_last_ptr == 0)
860          {
861          common->capture_last_ptr = common->ovector_start;
862          common->ovector_start += sizeof(sljit_sw);
863          }
864        cc += 2 + 2 * LINK_SIZE;
865        break;
866    
867        case OP_THEN_ARG:
868        common->has_then = TRUE;
869        common->control_head_ptr = 1;
870        /* Fall through. */
871    
872        case OP_PRUNE_ARG:
873        common->needs_start_ptr = TRUE;
874        /* Fall through. */
875    
876        case OP_MARK:
877        if (common->mark_ptr == 0)
878          {
879          common->mark_ptr = common->ovector_start;
880          common->ovector_start += sizeof(sljit_sw);
881          }
882        cc += 1 + 2 + cc[1];
883        break;
884    
885        case OP_THEN:
886        common->has_then = TRUE;
887        common->control_head_ptr = 1;
888        /* Fall through. */
889    
890        case OP_PRUNE:
891        case OP_SKIP:
892        common->needs_start_ptr = TRUE;
893        cc += 1;
894        break;
895    
896        case OP_SKIP_ARG:
897        common->control_head_ptr = 1;
898        common->has_skip_arg = TRUE;
899        cc += 1 + 2 + cc[1];
900        break;
901    
902        default:
903        cc = next_opcode(common, cc);
904        if (cc == NULL)
905          return FALSE;
906        break;
907        }
908      }
909    return TRUE;
910    }
911    
912    static int get_class_iterator_size(pcre_uchar *cc)
913    {
914  switch(*cc)  switch(*cc)
915    {    {
916    case OP_CRSTAR:    case OP_CRSTAR:
# Line 762  switch(*cc) Line 934  switch(*cc)
934    }    }
935  }  }
936    
937  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)
938  {  {
939  int private_data_length = 0;  pcre_uchar *end = bracketend(begin);
940  pcre_uchar *alternative;  pcre_uchar *next;
941  pcre_uchar *name;  pcre_uchar *next_end;
942  pcre_uchar *end = NULL;  pcre_uchar *max_end;
943  int space, size, i;  pcre_uchar type;
944  pcre_uint32 bracketlen;  sljit_sw length = end - begin;
945    int min, max, i;
946    
947  /* Calculate important variables (like stack size) and checks whether all opcodes are supported. */  /* Detect fixed iterations first. */
948  while (cc < ccend)  if (end[-(1 + LINK_SIZE)] != OP_KET)
949      return FALSE;
950    
951    /* Already detected repeat. */
952    if (common->private_data_ptrs[end - common->start - LINK_SIZE] != 0)
953      return TRUE;
954    
955    next = end;
956    min = 1;
957    while (1)
958    {    {
959    space = 0;    if (*next != *begin)
   size = 0;  
   bracketlen = 0;  
   switch(*cc)  
     {  
     case OP_SET_SOM:  
     common->has_set_som = TRUE;  
     cc += 1;  
960      break;      break;
961      next_end = bracketend(next);
962      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;  
963      break;      break;
964      next = next_end;
965      min++;
966      }
967    
968      case OP_ASSERT:  if (min == 2)
969      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;  
970    
971      case OP_CBRAPOS:  max = 0;
972      case OP_SCBRAPOS:  max_end = next;
973      private_data_length += sizeof(sljit_sw);  if (*next == OP_BRAZERO || *next == OP_BRAMINZERO)
974      common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] = 0;    {
975      bracketlen = 1 + LINK_SIZE + IMM2_SIZE;    type = *next;
976      break;    while (1)
977        {
978        if (next[0] != type || next[1] != OP_BRA || next[2 + LINK_SIZE] != *begin)
979          break;
980        next_end = bracketend(next + 2 + LINK_SIZE);
981        if (next_end - next != (length + 2 + LINK_SIZE) || memcmp(begin, next + 2 + LINK_SIZE, IN_UCHARS(length)) != 0)
982          break;
983        next = next_end;
984        max++;
985        }
986    
987      case OP_COND:    if (next[0] == type && next[1] == *begin && max >= 1)
988      case OP_SCOND:      {
989      bracketlen = cc[1 + LINK_SIZE];      next_end = bracketend(next + 1);
990      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)  
991        {        {
992        bracketlen = GET2(cc, 1 + LINK_SIZE + 1);        for (i = 0; i < max; i++, next_end += 1 + LINK_SIZE)
993        name = (pcre_uchar *)common->name_table;          if (*next_end != OP_KET)
994        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);  
995    
996        for (i = 0; i < common->name_count; i++)        if (i == max)
997          {          {
998          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;
999            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;
1000          alternative += common->name_entry_size;          /* +2 the original and the last. */
1001            common->private_data_ptrs[max_end - common->start - LINK_SIZE + 2] = max + 2;
1002            if (min == 1)
1003              return TRUE;
1004            min--;
1005            max_end -= (1 + LINK_SIZE) + GET(max_end, -LINK_SIZE);
1006          }          }
1007        }        }
1008        }
1009      }
1010    
1011      if (*cc == OP_COND)  if (min >= 3)
1012        {    {
1013        /* Might be a hidden SCOND. */    common->private_data_ptrs[end - common->start - LINK_SIZE] = max_end - end;
1014        alternative = cc + GET(cc, 1);    common->private_data_ptrs[end - common->start - LINK_SIZE + 1] = OP_EXACT;
1015        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)    common->private_data_ptrs[end - common->start - LINK_SIZE + 2] = min;
1016          private_data_length += sizeof(sljit_sw);    return TRUE;
1017        }    }
     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;  
1018    
1019  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  return FALSE;
1020      case OP_XCLASS:  }
     size = GET(cc, 1);  
     space = get_class_iterator_size(cc + size);  
     break;  
 #endif  
1021    
1022      case OP_RECURSE:  #define CASE_ITERATOR_PRIVATE_DATA_1 \
1023      /* Set its value only once. */      case OP_MINSTAR: \
1024      if (common->recursive_head == 0)      case OP_MINPLUS: \
1025        {      case OP_QUERY: \
1026        common->recursive_head = common->ovector_start;      case OP_MINQUERY: \
1027        common->ovector_start += sizeof(sljit_sw);      case OP_MINSTARI: \
1028        }      case OP_MINPLUSI: \
1029      cc += 1 + LINK_SIZE;      case OP_QUERYI: \
1030      break;      case OP_MINQUERYI: \
1031        case OP_NOTMINSTAR: \
1032        case OP_NOTMINPLUS: \
1033        case OP_NOTQUERY: \
1034        case OP_NOTMINQUERY: \
1035        case OP_NOTMINSTARI: \
1036        case OP_NOTMINPLUSI: \
1037        case OP_NOTQUERYI: \
1038        case OP_NOTMINQUERYI:
1039    
1040      case OP_MARK:  #define CASE_ITERATOR_PRIVATE_DATA_2A \
1041      if (common->mark_ptr == 0)      case OP_STAR: \
1042        {      case OP_PLUS: \
1043        common->mark_ptr = common->ovector_start;      case OP_STARI: \
1044        common->ovector_start += sizeof(sljit_sw);      case OP_PLUSI: \
1045        }      case OP_NOTSTAR: \
1046      cc += 1 + 2 + cc[1];      case OP_NOTPLUS: \
1047      break;      case OP_NOTSTARI: \
1048        case OP_NOTPLUSI:
1049    
1050      default:  #define CASE_ITERATOR_PRIVATE_DATA_2B \
1051      cc = next_opcode(common, cc);      case OP_UPTO: \
1052      if (cc == NULL)      case OP_MINUPTO: \
1053        return -1;      case OP_UPTOI: \
1054      break;      case OP_MINUPTOI: \
1055      }      case OP_NOTUPTO: \
1056        case OP_NOTMINUPTO: \
1057        case OP_NOTUPTOI: \
1058        case OP_NOTMINUPTOI:
1059    
1060    if (space > 0 && cc >= end)  #define CASE_ITERATOR_TYPE_PRIVATE_DATA_1 \
1061      private_data_length += sizeof(sljit_sw) * space;      case OP_TYPEMINSTAR: \
1062        case OP_TYPEMINPLUS: \
1063        case OP_TYPEQUERY: \
1064        case OP_TYPEMINQUERY:
1065    
1066    if (size != 0)  #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2A \
1067      {      case OP_TYPESTAR: \
1068      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;  
     }  
1069    
1070    if (bracketlen != 0)  #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2B \
1071      {      case OP_TYPEUPTO: \
1072      if (cc >= end)      case OP_TYPEMINUPTO:
       {  
       end = bracketend(cc);  
       if (end[-1 - LINK_SIZE] == OP_KET)  
         end = NULL;  
       }  
     cc += bracketlen;  
     }  
   }  
 return private_data_length;  
 }  
1073    
1074  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)
1075  {  {
1076  pcre_uchar *cc = common->start;  pcre_uchar *cc = common->start;
1077  pcre_uchar *alternative;  pcre_uchar *alternative;
1078  pcre_uchar *end = NULL;  pcre_uchar *end = NULL;
1079    int private_data_ptr = *private_data_start;
1080  int space, size, bracketlen;  int space, size, bracketlen;
1081    
1082  while (cc < ccend)  while (cc < ccend)
# Line 972  while (cc < ccend) Line 1084  while (cc < ccend)
1084    space = 0;    space = 0;
1085    size = 0;    size = 0;
1086    bracketlen = 0;    bracketlen = 0;
1087      if (private_data_ptr > SLJIT_MAX_LOCAL_SIZE)
1088        return;
1089    
1090      if (*cc == OP_ONCE || *cc == OP_ONCE_NC || *cc == OP_BRA || *cc == OP_CBRA || *cc == OP_COND)
1091        if (detect_repeat(common, cc))
1092          {
1093          /* These brackets are converted to repeats, so no global
1094          based single character repeat is allowed. */
1095          if (cc >= end)
1096            end = bracketend(cc);
1097          }
1098    
1099    switch(*cc)    switch(*cc)
1100      {      {
1101        case OP_KET:
1102        if (common->private_data_ptrs[cc + 1 - common->start] != 0)
1103          {
1104          common->private_data_ptrs[cc - common->start] = private_data_ptr;
1105          private_data_ptr += sizeof(sljit_sw);
1106          cc += common->private_data_ptrs[cc + 1 - common->start];
1107          }
1108        cc += 1 + LINK_SIZE;
1109        break;
1110    
1111      case OP_ASSERT:      case OP_ASSERT:
1112      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1113      case OP_ASSERTBACK:      case OP_ASSERTBACK:
# Line 1067  while (cc < ccend) Line 1201  while (cc < ccend)
1201      break;      break;
1202      }      }
1203    
1204      /* Character iterators, which are not inside a repeated bracket,
1205         gets a private slot instead of allocating it on the stack. */
1206    if (space > 0 && cc >= end)    if (space > 0 && cc >= end)
1207      {      {
1208      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 1233  while (cc < ccend)
1233      cc += bracketlen;      cc += bracketlen;
1234      }      }
1235    }    }
1236    *private_data_start = private_data_ptr;
1237  }  }
1238    
1239  /* Returns with -1 if no need for frame. */  /* Returns with a frame_types (always < 0) if no need for frame. */
1240  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)
1241  {  {
 pcre_uchar *ccend = bracketend(cc);  
1242  int length = 0;  int length = 0;
1243  BOOL possessive = FALSE;  int possessive = 0;
1244    BOOL stack_restore = FALSE;
1245  BOOL setsom_found = recursive;  BOOL setsom_found = recursive;
1246  BOOL setmark_found = recursive;  BOOL setmark_found = recursive;
1247    /* The last capture is a local variable even for recursions. */
1248    BOOL capture_last_found = FALSE;
1249    
1250    #if defined DEBUG_FORCE_CONTROL_HEAD && DEBUG_FORCE_CONTROL_HEAD
1251    SLJIT_ASSERT(common->control_head_ptr != 0);
1252    *needs_control_head = TRUE;
1253    #else
1254    *needs_control_head = FALSE;
1255    #endif
1256    
1257  if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))  if (ccend == NULL)
1258    {    {
1259    length = 3;    ccend = bracketend(cc) - (1 + LINK_SIZE);
1260    possessive = TRUE;    if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))
1261        {
1262        possessive = length = (common->capture_last_ptr != 0) ? 5 : 3;
1263        /* This is correct regardless of common->capture_last_ptr. */
1264        capture_last_found = TRUE;
1265        }
1266      cc = next_opcode(common, cc);
1267    }    }
1268    
 cc = next_opcode(common, cc);  
1269  SLJIT_ASSERT(cc != NULL);  SLJIT_ASSERT(cc != NULL);
1270  while (cc < ccend)  while (cc < ccend)
1271    switch(*cc)    switch(*cc)
1272      {      {
1273      case OP_SET_SOM:      case OP_SET_SOM:
1274      SLJIT_ASSERT(common->has_set_som);      SLJIT_ASSERT(common->has_set_som);
1275        stack_restore = TRUE;
1276      if (!setsom_found)      if (!setsom_found)
1277        {        {
1278        length += 2;        length += 2;
# Line 1130  while (cc < ccend) Line 1282  while (cc < ccend)
1282      break;      break;
1283    
1284      case OP_MARK:      case OP_MARK:
1285        case OP_PRUNE_ARG:
1286        case OP_THEN_ARG:
1287      SLJIT_ASSERT(common->mark_ptr != 0);      SLJIT_ASSERT(common->mark_ptr != 0);
1288        stack_restore = TRUE;
1289      if (!setmark_found)      if (!setmark_found)
1290        {        {
1291        length += 2;        length += 2;
1292        setmark_found = TRUE;        setmark_found = TRUE;
1293        }        }
1294        if (common->control_head_ptr != 0)
1295          *needs_control_head = TRUE;
1296      cc += 1 + 2 + cc[1];      cc += 1 + 2 + cc[1];
1297      break;      break;
1298    
1299      case OP_RECURSE:      case OP_RECURSE:
1300        stack_restore = TRUE;
1301      if (common->has_set_som && !setsom_found)      if (common->has_set_som && !setsom_found)
1302        {        {
1303        length += 2;        length += 2;
# Line 1150  while (cc < ccend) Line 1308  while (cc < ccend)
1308        length += 2;        length += 2;
1309        setmark_found = TRUE;        setmark_found = TRUE;
1310        }        }
1311        if (common->capture_last_ptr != 0 && !capture_last_found)
1312          {
1313          length += 2;
1314          capture_last_found = TRUE;
1315          }
1316      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1317      break;      break;
1318    
# Line 1157  while (cc < ccend) Line 1320  while (cc < ccend)
1320      case OP_CBRAPOS:      case OP_CBRAPOS:
1321      case OP_SCBRA:      case OP_SCBRA:
1322      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1323        stack_restore = TRUE;
1324        if (common->capture_last_ptr != 0 && !capture_last_found)
1325          {
1326          length += 2;
1327          capture_last_found = TRUE;
1328          }
1329      length += 3;      length += 3;
1330      cc += 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1331      break;      break;
1332    
1333      default:      default:
1334        stack_restore = TRUE;
1335        /* Fall through. */
1336    
1337        case OP_NOT_WORD_BOUNDARY:
1338        case OP_WORD_BOUNDARY:
1339        case OP_NOT_DIGIT:
1340        case OP_DIGIT:
1341        case OP_NOT_WHITESPACE:
1342        case OP_WHITESPACE:
1343        case OP_NOT_WORDCHAR:
1344        case OP_WORDCHAR:
1345        case OP_ANY:
1346        case OP_ALLANY:
1347        case OP_ANYBYTE:
1348        case OP_NOTPROP:
1349        case OP_PROP:
1350        case OP_ANYNL:
1351        case OP_NOT_HSPACE:
1352        case OP_HSPACE:
1353        case OP_NOT_VSPACE:
1354        case OP_VSPACE:
1355        case OP_EXTUNI:
1356        case OP_EODN:
1357        case OP_EOD:
1358        case OP_CIRC:
1359        case OP_CIRCM:
1360        case OP_DOLL:
1361        case OP_DOLLM:
1362        case OP_CHAR:
1363        case OP_CHARI:
1364        case OP_NOT:
1365        case OP_NOTI:
1366    
1367        case OP_EXACT:
1368        case OP_POSSTAR:
1369        case OP_POSPLUS:
1370        case OP_POSQUERY:
1371        case OP_POSUPTO:
1372    
1373        case OP_EXACTI:
1374        case OP_POSSTARI:
1375        case OP_POSPLUSI:
1376        case OP_POSQUERYI:
1377        case OP_POSUPTOI:
1378    
1379        case OP_NOTEXACT:
1380        case OP_NOTPOSSTAR:
1381        case OP_NOTPOSPLUS:
1382        case OP_NOTPOSQUERY:
1383        case OP_NOTPOSUPTO:
1384    
1385        case OP_NOTEXACTI:
1386        case OP_NOTPOSSTARI:
1387        case OP_NOTPOSPLUSI:
1388        case OP_NOTPOSQUERYI:
1389        case OP_NOTPOSUPTOI:
1390    
1391        case OP_TYPEEXACT:
1392        case OP_TYPEPOSSTAR:
1393        case OP_TYPEPOSPLUS:
1394        case OP_TYPEPOSQUERY:
1395        case OP_TYPEPOSUPTO:
1396    
1397        case OP_CLASS:
1398        case OP_NCLASS:
1399        case OP_XCLASS:
1400    
1401      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
1402      SLJIT_ASSERT(cc != NULL);      SLJIT_ASSERT(cc != NULL);
1403      break;      break;
1404      }      }
1405    
1406  /* Possessive quantifiers can use a special case. */  /* Possessive quantifiers can use a special case. */
1407  if (SLJIT_UNLIKELY(possessive) && length == 3)  if (SLJIT_UNLIKELY(possessive == length))
1408    return -1;    return stack_restore ? no_frame : no_stack;
1409    
1410  if (length > 0)  if (length > 0)
1411    return length + 1;    return length + 1;
1412  return -1;  return stack_restore ? no_frame : no_stack;
1413  }  }
1414    
1415  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)
1416  {  {
1417  DEFINE_COMPILER;  DEFINE_COMPILER;
 pcre_uchar *ccend = bracketend(cc);  
1418  BOOL setsom_found = recursive;  BOOL setsom_found = recursive;
1419  BOOL setmark_found = recursive;  BOOL setmark_found = recursive;
1420    /* The last capture is a local variable even for recursions. */
1421    BOOL capture_last_found = FALSE;
1422  int offset;  int offset;
1423    
1424  /* >= 1 + shortest item size (2) */  /* >= 1 + shortest item size (2) */
# Line 1189  SLJIT_UNUSED_ARG(stacktop); Line 1426  SLJIT_UNUSED_ARG(stacktop);
1426  SLJIT_ASSERT(stackpos >= stacktop + 2);  SLJIT_ASSERT(stackpos >= stacktop + 2);
1427    
1428  stackpos = STACK(stackpos);  stackpos = STACK(stackpos);
1429  if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))  if (ccend == NULL)
1430    cc = next_opcode(common, cc);    {
1431      ccend = bracketend(cc) - (1 + LINK_SIZE);
1432      if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))
1433        cc = next_opcode(common, cc);
1434      }
1435    
1436  SLJIT_ASSERT(cc != NULL);  SLJIT_ASSERT(cc != NULL);
1437  while (cc < ccend)  while (cc < ccend)
1438    switch(*cc)    switch(*cc)
# Line 1200  while (cc < ccend) Line 1442  while (cc < ccend)
1442      if (!setsom_found)      if (!setsom_found)
1443        {        {
1444        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));
1445        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setstrbegin);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1446        stackpos += (int)sizeof(sljit_sw);        stackpos += (int)sizeof(sljit_sw);
1447        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1448        stackpos += (int)sizeof(sljit_sw);        stackpos += (int)sizeof(sljit_sw);
# Line 1210  while (cc < ccend) Line 1452  while (cc < ccend)
1452      break;      break;
1453    
1454      case OP_MARK:      case OP_MARK:
1455        case OP_PRUNE_ARG:
1456        case OP_THEN_ARG:
1457      SLJIT_ASSERT(common->mark_ptr != 0);      SLJIT_ASSERT(common->mark_ptr != 0);
1458      if (!setmark_found)      if (!setmark_found)
1459        {        {
1460        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);
1461        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);
1462        stackpos += (int)sizeof(sljit_sw);        stackpos += (int)sizeof(sljit_sw);
1463        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1464        stackpos += (int)sizeof(sljit_sw);        stackpos += (int)sizeof(sljit_sw);
# Line 1227  while (cc < ccend) Line 1471  while (cc < ccend)
1471      if (common->has_set_som && !setsom_found)      if (common->has_set_som && !setsom_found)
1472        {        {
1473        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));
1474        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setstrbegin);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1475        stackpos += (int)sizeof(sljit_sw);        stackpos += (int)sizeof(sljit_sw);
1476        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1477        stackpos += (int)sizeof(sljit_sw);        stackpos += (int)sizeof(sljit_sw);
# Line 1236  while (cc < ccend) Line 1480  while (cc < ccend)
1480      if (common->mark_ptr != 0 && !setmark_found)      if (common->mark_ptr != 0 && !setmark_found)
1481        {        {
1482        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);
1483        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);
1484        stackpos += (int)sizeof(sljit_sw);        stackpos += (int)sizeof(sljit_sw);
1485        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1486        stackpos += (int)sizeof(sljit_sw);        stackpos += (int)sizeof(sljit_sw);
1487        setmark_found = TRUE;        setmark_found = TRUE;
1488        }        }
1489        if (common->capture_last_ptr != 0 && !capture_last_found)
1490          {
1491          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->capture_last_ptr);
1492          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1493          stackpos += (int)sizeof(sljit_sw);
1494          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1495          stackpos += (int)sizeof(sljit_sw);
1496          capture_last_found = TRUE;
1497          }
1498      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1499      break;      break;
1500    
# Line 1249  while (cc < ccend) Line 1502  while (cc < ccend)
1502      case OP_CBRAPOS:      case OP_CBRAPOS:
1503      case OP_SCBRA:      case OP_SCBRA:
1504      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1505        if (common->capture_last_ptr != 0 && !capture_last_found)
1506          {
1507          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->capture_last_ptr);
1508          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1509          stackpos += (int)sizeof(sljit_sw);
1510          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1511          stackpos += (int)sizeof(sljit_sw);
1512          capture_last_found = TRUE;
1513          }
1514      offset = (GET2(cc, 1 + LINK_SIZE)) << 1;      offset = (GET2(cc, 1 + LINK_SIZE)) << 1;
1515      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));
1516      stackpos += (int)sizeof(sljit_sw);      stackpos += (int)sizeof(sljit_sw);
# Line 1268  while (cc < ccend) Line 1530  while (cc < ccend)
1530      break;      break;
1531      }      }
1532    
1533  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_end);  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, 0);
1534  SLJIT_ASSERT(stackpos == STACK(stacktop));  SLJIT_ASSERT(stackpos == STACK(stacktop));
1535  }  }
1536    
1537  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)
1538  {  {
1539  int private_data_length = 2;  int private_data_length = needs_control_head ? 3 : 2;
1540  int size;  int size;
1541  pcre_uchar *alternative;  pcre_uchar *alternative;
1542  /* Calculate the sum of the private machine words. */  /* Calculate the sum of the private machine words. */
# Line 1283  while (cc < ccend) Line 1545  while (cc < ccend)
1545    size = 0;    size = 0;
1546    switch(*cc)    switch(*cc)
1547      {      {
1548        case OP_KET:
1549        if (PRIVATE_DATA(cc) != 0)
1550          private_data_length++;
1551        cc += 1 + LINK_SIZE;
1552        break;
1553    
1554      case OP_ASSERT:      case OP_ASSERT:
1555      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1556      case OP_ASSERTBACK:      case OP_ASSERTBACK:
# Line 1387  return private_data_length; Line 1655  return private_data_length;
1655  }  }
1656    
1657  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,
1658    BOOL save, int stackptr, int stacktop)    BOOL save, int stackptr, int stacktop, BOOL needs_control_head)
1659  {  {
1660  DEFINE_COMPILER;  DEFINE_COMPILER;
1661  int srcw[2];  int srcw[2];
# Line 1408  stacktop = STACK(stacktop - 1); Line 1676  stacktop = STACK(stacktop - 1);
1676    
1677  if (!save)  if (!save)
1678    {    {
1679    stackptr += sizeof(sljit_sw);    stackptr += (needs_control_head ? 2 : 1) * sizeof(sljit_sw);
1680    if (stackptr < stacktop)    if (stackptr < stacktop)
1681      {      {
1682      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 1692  if (!save)
1692    /* The tmp1next must be TRUE in either way. */    /* The tmp1next must be TRUE in either way. */
1693    }    }
1694    
1695  while (status != end)  do
1696    {    {
1697    count = 0;    count = 0;
1698    switch(status)    switch(status)
1699      {      {
1700      case start:      case start:
1701      SLJIT_ASSERT(save && common->recursive_head != 0);      SLJIT_ASSERT(save && common->recursive_head_ptr != 0);
1702      count = 1;      count = 1;
1703      srcw[0] = common->recursive_head;      srcw[0] = common->recursive_head_ptr;
1704        if (needs_control_head)
1705          {
1706          SLJIT_ASSERT(common->control_head_ptr != 0);
1707          count = 2;
1708          srcw[1] = common->control_head_ptr;
1709          }
1710      status = loop;      status = loop;
1711      break;      break;
1712    
# Line 1445  while (status != end) Line 1719  while (status != end)
1719    
1720      switch(*cc)      switch(*cc)
1721        {        {
1722          case OP_KET:
1723          if (PRIVATE_DATA(cc) != 0)
1724            {
1725            count = 1;
1726            srcw[0] = PRIVATE_DATA(cc);
1727            }
1728          cc += 1 + LINK_SIZE;
1729          break;
1730    
1731        case OP_ASSERT:        case OP_ASSERT:
1732        case OP_ASSERT_NOT:        case OP_ASSERT_NOT:
1733        case OP_ASSERTBACK:        case OP_ASSERTBACK:
# Line 1657  while (status != end) Line 1940  while (status != end)
1940        }        }
1941      }      }
1942    }    }
1943    while (status != end);
1944    
1945  if (save)  if (save)
1946    {    {
# Line 1687  if (save) Line 1971  if (save)
1971        }        }
1972      }      }
1973    }    }
1974  SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));  SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));
1975    }
1976    
1977    static SLJIT_INLINE pcre_uchar *set_then_offsets(compiler_common *common, pcre_uchar *cc, pcre_uint8 *current_offset)
1978    {
1979    pcre_uchar *end = bracketend(cc);
1980    BOOL has_alternatives = cc[GET(cc, 1)] == OP_ALT;
1981    
1982    /* Assert captures then. */
1983    if (*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT)
1984      current_offset = NULL;
1985    /* Conditional block does not. */
1986    if (*cc == OP_COND || *cc == OP_SCOND)
1987      has_alternatives = FALSE;
1988    
1989    cc = next_opcode(common, cc);
1990    if (has_alternatives)
1991      current_offset = common->then_offsets + (cc - common->start);
1992    
1993    while (cc < end)
1994      {
1995      if ((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND))
1996        cc = set_then_offsets(common, cc, current_offset);
1997      else
1998        {
1999        if (*cc == OP_ALT && has_alternatives)
2000          current_offset = common->then_offsets + (cc + 1 + LINK_SIZE - common->start);
2001        if (*cc >= OP_THEN && *cc <= OP_THEN_ARG && current_offset != NULL)
2002          *current_offset = 1;
2003        cc = next_opcode(common, cc);
2004        }
2005      }
2006    
2007    return end;
2008  }  }
2009    
2010  #undef CASE_ITERATOR_PRIVATE_DATA_1  #undef CASE_ITERATOR_PRIVATE_DATA_1
# Line 1708  while (list) Line 2025  while (list)
2025    {    {
2026    /* sljit_set_label is clever enough to do nothing    /* sljit_set_label is clever enough to do nothing
2027    if either the jump or the label is NULL. */    if either the jump or the label is NULL. */
2028    sljit_set_label(list->jump, label);    SET_LABEL(list->jump, label);
2029    list = list->next;    list = list->next;
2030    }    }
2031  }  }
# Line 1724  if (list_item) Line 2041  if (list_item)
2041    }    }
2042  }  }
2043    
2044  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)
2045  {  {
2046  DEFINE_COMPILER;  DEFINE_COMPILER;
2047  stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));  stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));
2048    
2049  if (list_item)  if (list_item)
2050    {    {
   list_item->type = type;  
   list_item->data = data;  
2051    list_item->start = start;    list_item->start = start;
2052    list_item->quit = LABEL();    list_item->quit = LABEL();
2053    list_item->next = common->stubs;    list_item->next = common->stubs;
# Line 1748  stub_list* list_item = common->stubs; Line 2063  stub_list* list_item = common->stubs;
2063  while (list_item)  while (list_item)
2064    {    {
2065    JUMPHERE(list_item->start);    JUMPHERE(list_item->start);
2066    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;  
     }  
2067    JUMPTO(SLJIT_JUMP, list_item->quit);    JUMPTO(SLJIT_JUMP, list_item->quit);
2068    list_item = list_item->next;    list_item = list_item->next;
2069    }    }
2070  common->stubs = NULL;  common->stubs = NULL;
2071  }  }
2072    
2073  static SLJIT_INLINE void decrease_call_count(compiler_common *common)  static void add_label_addr(compiler_common *common)
2074    {
2075    DEFINE_COMPILER;
2076    label_addr_list *label_addr;
2077    
2078    label_addr = sljit_alloc_memory(compiler, sizeof(label_addr_list));
2079    if (label_addr == NULL)
2080      return;
2081    label_addr->label = LABEL();
2082    label_addr->addr = common->read_only_data_ptr;
2083    label_addr->next = common->label_addrs;
2084    common->label_addrs = label_addr;
2085    common->read_only_data_ptr++;
2086    }
2087    
2088    static SLJIT_INLINE void count_match(compiler_common *common)
2089  {  {
2090  DEFINE_COMPILER;  DEFINE_COMPILER;
2091    
2092  OP2(SLJIT_SUB | SLJIT_SET_E, CALL_COUNT, 0, CALL_COUNT, 0, SLJIT_IMM, 1);  OP2(SLJIT_SUB | SLJIT_SET_E, COUNT_MATCH, 0, COUNT_MATCH, 0, SLJIT_IMM, 1);
2093  add_jump(compiler, &common->calllimit, JUMP(SLJIT_C_ZERO));  add_jump(compiler, &common->calllimit, JUMP(SLJIT_C_ZERO));
2094  }  }
2095    
# Line 1781  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0); Line 2106  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);
2106  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);
2107  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);
2108  #endif  #endif
2109  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));
2110  }  }
2111    
2112  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 2120  static SLJIT_INLINE void reset_ovector(c
2120  DEFINE_COMPILER;  DEFINE_COMPILER;
2121  struct sljit_label *loop;  struct sljit_label *loop;
2122  int i;  int i;
2123    
2124  /* At this point we can freely use all temporary registers. */  /* At this point we can freely use all temporary registers. */
2125    SLJIT_ASSERT(length > 1);
2126  /* TMP1 returns with begin - 1. */  /* TMP1 returns with begin - 1. */
2127  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));
2128  if (length < 8)  if (length < 8)
2129    {    {
2130    for (i = 0; i < length; i++)    for (i = 1; i < length; i++)
2131      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);
2132    }    }
2133  else  else
2134    {    {
2135    GET_LOCAL_BASE(SLJIT_SCRATCH_REG2, 0, OVECTOR_START - sizeof(sljit_sw));    GET_LOCAL_BASE(SLJIT_SCRATCH_REG2, 0, OVECTOR_START);
2136    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, length);    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, length - 1);
2137    loop = LABEL();    loop = LABEL();
2138    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);
2139    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 2141  else
2141    }    }
2142  }  }
2143    
2144    static SLJIT_INLINE void do_reset_match(compiler_common *common, int length)
2145    {
2146    DEFINE_COMPILER;
2147    struct sljit_label *loop;
2148    int i;
2149    
2150    SLJIT_ASSERT(length > 1);
2151    /* OVECTOR(1) contains the "string begin - 1" constant. */
2152    if (length > 2)
2153      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));
2154    if (length < 8)
2155      {
2156      for (i = 2; i < length; i++)
2157        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), TMP1, 0);
2158      }
2159    else
2160      {
2161      GET_LOCAL_BASE(TMP2, 0, OVECTOR_START + sizeof(sljit_sw));
2162      OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_IMM, length - 2);
2163      loop = LABEL();
2164      OP1(SLJIT_MOVU, SLJIT_MEM1(TMP2), sizeof(sljit_sw), TMP1, 0);
2165      OP2(SLJIT_SUB | SLJIT_SET_E, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, 1);
2166      JUMPTO(SLJIT_C_NOT_ZERO, loop);
2167      }
2168    
2169    OP1(SLJIT_MOV, STACK_TOP, 0, ARGUMENTS, 0);
2170    if (common->mark_ptr != 0)
2171      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr, SLJIT_IMM, 0);
2172    if (common->control_head_ptr != 0)
2173      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->control_head_ptr, SLJIT_IMM, 0);
2174    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(jit_arguments, stack));
2175    OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_ptr);
2176    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(struct sljit_stack, base));
2177    }
2178    
2179    static sljit_sw SLJIT_CALL do_search_mark(sljit_sw *current, const pcre_uchar *skip_arg)
2180    {
2181    while (current != NULL)
2182      {
2183      switch (current[-2])
2184        {
2185        case type_then_trap:
2186        break;
2187    
2188        case type_mark:
2189        if (STRCMP_UC_UC(skip_arg, (pcre_uchar *)current[-3]) == 0)
2190          return current[-4];
2191        break;
2192    
2193        default:
2194        SLJIT_ASSERT_STOP();
2195        break;
2196        }
2197      current = (sljit_sw*)current[-1];
2198      }
2199    return -1;
2200    }
2201    
2202  static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)  static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)
2203  {  {
2204  DEFINE_COMPILER;  DEFINE_COMPILER;
2205  struct sljit_label *loop;  struct sljit_label *loop;
2206  struct sljit_jump *earlyexit;  struct sljit_jump *early_quit;
2207    
2208  /* At this point we can freely use all registers. */  /* At this point we can freely use all registers. */
2209  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 2212  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_R
2212  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG1, 0, ARGUMENTS, 0);  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG1, 0, ARGUMENTS, 0);
2213  if (common->mark_ptr != 0)  if (common->mark_ptr != 0)
2214    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);
2215  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));
2216  if (common->mark_ptr != 0)  if (common->mark_ptr != 0)
2217    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);
2218  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));
2219  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));
2220  GET_LOCAL_BASE(SLJIT_SAVED_REG1, 0, OVECTOR_START);  GET_LOCAL_BASE(SLJIT_SAVED_REG1, 0, OVECTOR_START);
2221  /* Unlikely, but possible */  /* Unlikely, but possible */
2222  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);
2223  loop = LABEL();  loop = LABEL();
2224  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);
2225  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 2230  OP2(SLJIT_ASHR, SLJIT_SAVED_REG2, 0, SLJ
2230  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);
2231  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);
2232  JUMPTO(SLJIT_C_NOT_ZERO, loop);  JUMPTO(SLJIT_C_NOT_ZERO, loop);
2233  JUMPHERE(earlyexit);  JUMPHERE(early_quit);
2234    
2235  /* Calculate the return value, which is the maximum ovector value. */  /* Calculate the return value, which is the maximum ovector value. */
2236  if (topbracket > 1)  if (topbracket > 1)
# Line 1867  else Line 2252  else
2252  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)
2253  {  {
2254  DEFINE_COMPILER;  DEFINE_COMPILER;
2255    struct sljit_jump *jump;
2256    
2257  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);
2258  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
2259      && (common->mode == JIT_PARTIAL_SOFT_COMPILE ? common->hit_start != 0 : common->hit_start == 0));
2260    
2261  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, ARGUMENTS, 0);  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, ARGUMENTS, 0);
2262  OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, PCRE_ERROR_PARTIAL);  OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, PCRE_ERROR_PARTIAL);
2263  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));
2264  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);
2265    
2266  /* Store match begin and end. */  /* Store match begin and end. */
2267  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));
2268  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));
2269    
2270    jump = CMP(SLJIT_C_SIG_LESS, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 3);
2271    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);
2272    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2273    OP2(SLJIT_ASHR, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, UCHAR_SHIFT);
2274    #endif
2275    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG2), 2 * sizeof(int), SLJIT_SCRATCH_REG3, 0);
2276    JUMPHERE(jump);
2277    
2278  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);
2279  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);
2280  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
# Line 2040  return (bit < 256) ? ((0 << 8) | bit) : Line 2436  return (bit < 256) ? ((0 << 8) | bit) :
2436    
2437  static void check_partial(compiler_common *common, BOOL force)  static void check_partial(compiler_common *common, BOOL force)
2438  {  {
2439  /* Checks whether a partial matching is occured. Does not modify registers. */  /* Checks whether a partial matching is occurred. Does not modify registers. */
2440  DEFINE_COMPILER;  DEFINE_COMPILER;
2441  struct sljit_jump *jump = NULL;  struct sljit_jump *jump = NULL;
2442    
# Line 2055  else if (common->mode == JIT_PARTIAL_SOF Line 2451  else if (common->mode == JIT_PARTIAL_SOF
2451    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);
2452    
2453  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2454    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, -1);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2455  else  else
2456    {    {
2457    if (common->partialmatchlabel != NULL)    if (common->partialmatchlabel != NULL)
# Line 2068  if (jump != NULL) Line 2464  if (jump != NULL)
2464    JUMPHERE(jump);    JUMPHERE(jump);
2465  }  }
2466    
2467  static struct sljit_jump *check_str_end(compiler_common *common)  static void check_str_end(compiler_common *common, jump_list **end_reached)
2468  {  {
2469  /* Does not affect registers. Usually used in a tight spot. */  /* Does not affect registers. Usually used in a tight spot. */
2470  DEFINE_COMPILER;  DEFINE_COMPILER;
2471  struct sljit_jump *jump;  struct sljit_jump *jump;
 struct sljit_jump *nohit;  
 struct sljit_jump *return_value;  
2472    
2473  if (common->mode == JIT_COMPILE)  if (common->mode == JIT_COMPILE)
2474    return CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);    {
2475      add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2476      return;
2477      }
2478    
2479  jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);  jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);
2480  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2481    {    {
2482    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));
2483    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);
2484    JUMPHERE(nohit);    add_jump(compiler, end_reached, JUMP(SLJIT_JUMP));
   return_value = JUMP(SLJIT_JUMP);  
2485    }    }
2486  else  else
2487    {    {
2488    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));
2489    if (common->partialmatchlabel != NULL)    if (common->partialmatchlabel != NULL)
2490      JUMPTO(SLJIT_JUMP, common->partialmatchlabel);      JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2491    else    else
2492      add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));      add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2493    }    }
2494  JUMPHERE(jump);  JUMPHERE(jump);
 return return_value;  
2495  }  }
2496    
2497  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 2510  jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR
2510  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));
2511  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2512    {    {
2513    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);
2514    add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));    add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));
2515    }    }
2516  else  else
# Line 2128  else Line 2523  else
2523  JUMPHERE(jump);  JUMPHERE(jump);
2524  }  }
2525    
2526  static void read_char(compiler_common *common)  static void peek_char(compiler_common *common, pcre_uint32 max)
2527  {  {
2528  /* Reads the character into TMP1, updates STR_PTR.  /* Reads the character into TMP1, keeps STR_PTR.
2529  Does not check STR_END. TMP2 Destroyed. */  Does not check STR_END. TMP2 Destroyed. */
2530  DEFINE_COMPILER;  DEFINE_COMPILER;
2531  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2532  struct sljit_jump *jump;  struct sljit_jump *jump;
2533  #endif  #endif
2534    
2535    SLJIT_UNUSED_ARG(max);
2536    
2537  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2538  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2539  if (common->utf)  if (common->utf)
2540    {    {
2541  #if defined COMPILE_PCRE8    if (max < 128) return;
2542    
2543    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2544  #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] */  
2545    add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2546      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2547    JUMPHERE(jump);    JUMPHERE(jump);
2548    }    }
2549  #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */  #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2550    
2551    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2552    if (common->utf)
2553      {
2554      if (max < 0xd800) return;
2555    
2556      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2557      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2558      /* TMP2 contains the high surrogate. */
2559      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2560      OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2561      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2562      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2563      OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2564      JUMPHERE(jump);
2565      }
2566    #endif
2567    }
2568    
2569    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2570    
2571    static BOOL is_char7_bitset(const pcre_uint8 *bitset, BOOL nclass)
2572    {
2573    /* Tells whether the character codes below 128 are enough
2574    to determine a match. */
2575    const pcre_uint8 value = nclass ? 0xff : 0;
2576    const pcre_uint8* end = bitset + 32;
2577    
2578    bitset += 16;
2579    do
2580      {
2581      if (*bitset++ != value)
2582        return FALSE;
2583      }
2584    while (bitset < end);
2585    return TRUE;
2586    }
2587    
2588    static void read_char7_type(compiler_common *common, BOOL full_read)
2589    {
2590    /* Reads the precise character type of a character into TMP1, if the character
2591    is less than 128. Otherwise it returns with zero. Does not check STR_END. The
2592    full_read argument tells whether characters above max are accepted or not. */
2593    DEFINE_COMPILER;
2594    struct sljit_jump *jump;
2595    
2596    SLJIT_ASSERT(common->utf);
2597    
2598    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2599  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));
2600    
2601    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2602    
2603    if (full_read)
2604      {
2605      jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2606      OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2607      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2608      JUMPHERE(jump);
2609      }
2610  }  }
2611    
2612  static void peek_char(compiler_common *common)  #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2613    
2614    static void read_char_range(compiler_common *common, pcre_uint32 min, pcre_uint32 max, BOOL update_str_ptr)
2615  {  {
2616  /* Reads the character into TMP1, keeps STR_PTR.  /* Reads the precise value of a character into TMP1, if the character is
2617  Does not check STR_END. TMP2 Destroyed. */  between min and max (c >= min && c <= max). Otherwise it returns with a value
2618    outside the range. Does not check STR_END. */
2619  DEFINE_COMPILER;  DEFINE_COMPILER;
2620  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2621  struct sljit_jump *jump;  struct sljit_jump *jump;
2622  #endif  #endif
2623    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2624    struct sljit_jump *jump2;
2625    #endif
2626    
2627  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  SLJIT_UNUSED_ARG(update_str_ptr);
2628  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32  SLJIT_UNUSED_ARG(min);
2629    SLJIT_UNUSED_ARG(max);
2630    SLJIT_ASSERT(min <= max);
2631    
2632    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2633    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2634    
2635    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2636  if (common->utf)  if (common->utf)
2637    {    {
2638  #if defined COMPILE_PCRE8    if (max < 128 && !update_str_ptr) return;
2639    
2640    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2641  #elif defined COMPILE_PCRE16    if (min >= 0x10000)
2642    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);      {
2643  #endif /* COMPILE_PCRE[8|16] */      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xf0);
2644    add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));      if (update_str_ptr)
2645    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);
2646        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2647        jump2 = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0x7);
2648        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2649        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2650        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2651        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2652        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2653        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2654        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2655        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2656        if (!update_str_ptr)
2657          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2658        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2659        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2660        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2661        JUMPHERE(jump2);
2662        if (update_str_ptr)
2663          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2664        }
2665      else if (min >= 0x800 && max <= 0xffff)
2666        {
2667        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xe0);
2668        if (update_str_ptr)
2669          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2670        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2671        jump2 = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xf);
2672        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2673        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2674        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2675        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2676        if (!update_str_ptr)
2677          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2678        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2679        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2680        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2681        JUMPHERE(jump2);
2682        if (update_str_ptr)
2683          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2684        }
2685      else if (max >= 0x800)
2686        add_jump(compiler, (max < 0x10000) ? &common->utfreadchar16 : &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2687      else if (max < 128)
2688        {
2689        OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2690        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2691        }
2692      else
2693        {
2694        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2695        if (!update_str_ptr)
2696          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2697        else
2698          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2699        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2700        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2701        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2702        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2703        if (update_str_ptr)
2704          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2705        }
2706    JUMPHERE(jump);    JUMPHERE(jump);
2707    }    }
2708  #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */  #endif
2709    
2710    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2711    if (common->utf)
2712      {
2713      if (max >= 0x10000)
2714        {
2715        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2716        jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2717        /* TMP2 contains the high surrogate. */
2718        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2719        OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2720        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2721        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2722        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2723        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2724        JUMPHERE(jump);
2725        return;
2726        }
2727    
2728      if (max < 0xd800 && !update_str_ptr) return;
2729    
2730      /* Skip low surrogate if necessary. */
2731      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2732      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2733      if (update_str_ptr)
2734        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2735      if (max >= 0xd800)
2736        OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0x10000);
2737      JUMPHERE(jump);
2738      }
2739    #endif
2740    }
2741    
2742    static SLJIT_INLINE void read_char(compiler_common *common)
2743    {
2744    read_char_range(common, 0, READ_CHAR_MAX, TRUE);
2745  }  }
2746    
2747  static void read_char8_type(compiler_common *common)  static void read_char8_type(compiler_common *common, BOOL update_str_ptr)
2748  {  {
2749  /* 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. */
2750  DEFINE_COMPILER;  DEFINE_COMPILER;
2751  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2752  struct sljit_jump *jump;  struct sljit_jump *jump;
2753  #endif  #endif
2754    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2755    struct sljit_jump *jump2;
2756    #endif
2757    
2758  #ifdef SUPPORT_UTF  SLJIT_UNUSED_ARG(update_str_ptr);
2759    
2760    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2761    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2762    
2763    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2764  if (common->utf)  if (common->utf)
2765    {    {
   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  
2766    /* This can be an extra read in some situations, but hopefully    /* This can be an extra read in some situations, but hopefully
2767    it is needed in most cases. */    it is needed in most cases. */
2768    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2769    jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);    jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2770    add_jump(compiler, &common->utfreadtype8, JUMP(SLJIT_FAST_CALL));    if (!update_str_ptr)
2771    JUMPHERE(jump);      {
2772  #elif defined COMPILE_PCRE16      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2773    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2774    jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);      OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2775    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2776    JUMPHERE(jump);      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2777    /* Skip low surrogate if necessary. */      OP2(SLJIT_OR, TMP2, 0, TMP2, 0, TMP1, 0);
2778    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0xfc00);      OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2779    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0xd800);      jump2 = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);
2780    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2781    OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 1);      JUMPHERE(jump2);
2782    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);      }
2783  #elif defined COMPILE_PCRE32    else
2784    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);  
2785    JUMPHERE(jump);    JUMPHERE(jump);
 #endif /* COMPILE_PCRE[8|16|32] */  
2786    return;    return;
2787    }    }
2788  #endif /* SUPPORT_UTF */  #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2789  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);  
2790  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  
2791  /* The ctypes array contains only 256 values. */  /* The ctypes array contains only 256 values. */
2792  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2793  jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);  jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);
2794  #endif  #endif
2795  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2796  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32  #if !defined COMPILE_PCRE8
2797  JUMPHERE(jump);  JUMPHERE(jump);
2798  #endif  #endif
2799    
2800    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2801    if (common->utf && update_str_ptr)
2802      {
2803      /* Skip low surrogate if necessary. */
2804      OP2(SLJIT_SUB, TMP2, 0, TMP2, 0, SLJIT_IMM, 0xd800);
2805      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2806      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2807      JUMPHERE(jump);
2808      }
2809    #endif /* SUPPORT_UTF && COMPILE_PCRE16 */
2810  }  }
2811    
2812  static void skip_char_back(compiler_common *common)  static void skip_char_back(compiler_common *common)
# Line 2266  if (common->utf) Line 2844  if (common->utf)
2844  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));
2845  }  }
2846    
2847  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)
2848  {  {
2849  /* 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. */
2850  DEFINE_COMPILER;  DEFINE_COMPILER;
2851    struct sljit_jump *jump;
2852    
2853  if (nltype == NLTYPE_ANY)  if (nltype == NLTYPE_ANY)
2854    {    {
2855    add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));
2856    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));
2857    }    }
2858  else if (nltype == NLTYPE_ANYCRLF)  else if (nltype == NLTYPE_ANYCRLF)
2859    {    {
2860    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_CR);    if (jumpifmatch)
2861    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);      {
2862    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));
2863    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));
2864    add_jump(compiler, backtracks, JUMP(jumpiftrue ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));      }
2865      else
2866        {
2867        jump = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR);
2868        add_jump(compiler, backtracks, CMP(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL));
2869        JUMPHERE(jump);
2870        }
2871    }    }
2872  else  else
2873    {    {
2874    SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline < 256);    SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline < 256);
2875    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));
2876    }    }
2877  }  }
2878    
# Line 2297  else Line 2882  else
2882  static void do_utfreadchar(compiler_common *common)  static void do_utfreadchar(compiler_common *common)
2883  {  {
2884  /* Fast decoding a UTF-8 character. TMP1 contains the first byte  /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2885  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. */
2886  DEFINE_COMPILER;  DEFINE_COMPILER;
2887  struct sljit_jump *jump;  struct sljit_jump *jump;
2888    
2889  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2890    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2891    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2892    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2893    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2894    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2895    
2896  /* Searching for the first zero. */  /* Searching for the first zero. */
2897  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);
2898  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_C_NOT_ZERO);
2899  /* Two byte sequence. */  /* Two byte sequence. */
 OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  
2900  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));
2901  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x1f);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(2));
2902    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2903    
2904    JUMPHERE(jump);
2905    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2906    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
2907  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2908  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2909  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);  
2910    
2911  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);
2912  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_C_NOT_ZERO);
2913  /* Three byte sequence. */  /* Three byte sequence. */
2914  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2915  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x0f);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(3));
2916  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 12);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2917    
2918    /* Four byte sequence. */
2919    JUMPHERE(jump);
2920    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2921    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2922    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2923    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2924  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);  
2925  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2926  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(4));
2927  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2928    }
2929    
2930    static void do_utfreadchar16(compiler_common *common)
2931    {
2932    /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2933    of the character (>= 0xc0). Return value in TMP1. */
2934    DEFINE_COMPILER;
2935    struct sljit_jump *jump;
2936    
2937    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2938    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2939    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2940    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2941  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2942  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2943  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(2));  
2944    /* Searching for the first zero. */
2945    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
2946    jump = JUMP(SLJIT_C_NOT_ZERO);
2947    /* Two byte sequence. */
2948    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2949  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
 JUMPHERE(jump);  
2950    
2951  /* Four byte sequence. */  JUMPHERE(jump);
2952    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x400);
2953    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_NOT_ZERO);
2954    /* This code runs only in 8 bit mode. No need to shift the value. */
2955    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2956  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2957  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x07);  OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
2958  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 18);  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
 OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  
 OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 12);  
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
 OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));  
 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(3));  
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));  
2959  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2960  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2961  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(3));  /* Three byte sequence. */
2962    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2963  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2964  }  }
2965    
# Line 2368  jump = JUMP(SLJIT_C_NOT_ZERO); Line 2979  jump = JUMP(SLJIT_C_NOT_ZERO);
2979  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2980  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));
2981  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);
2982    /* The upper 5 bits are known at this point. */
2983    compare = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0x3);
2984  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2985  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2986  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);  
2987  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2988  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2989    
2990  JUMPHERE(compare);  JUMPHERE(compare);
2991  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2992  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
 JUMPHERE(jump);  
2993    
2994  /* 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);  
   
2995  JUMPHERE(jump);  JUMPHERE(jump);
2996  /* Combine two 16 bit characters. */  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2997  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2998  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);  
2999  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3000  }  }
3001    
3002  #endif /* COMPILE_PCRE[8|16] */  #endif /* COMPILE_PCRE8 */
3003    
3004  #endif /* SUPPORT_UTF */  #endif /* SUPPORT_UTF */
3005    
# Line 2489  if (firstline) Line 3074  if (firstline)
3074      mainloop = LABEL();      mainloop = LABEL();
3075      /* Continual stores does not cause data dependency. */      /* Continual stores does not cause data dependency. */
3076      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);
3077      read_char(common);      read_char_range(common, common->nlmin, common->nlmax, TRUE);
3078      check_newlinechar(common, common->nltype, &newline, TRUE);      check_newlinechar(common, common->nltype, &newline, TRUE);
3079      CMPTO(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0, mainloop);      CMPTO(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0, mainloop);
3080      JUMPHERE(end);      JUMPHERE(end);
# Line 2556  if (common->utf) Line 3141  if (common->utf)
3141  #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */  #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
3142  JUMPHERE(start);  JUMPHERE(start);
3143    
3144  if (newlinecheck)  if (newlinecheck)
3145    {    {
3146    JUMPHERE(end);    JUMPHERE(end);
3147    JUMPHERE(nl);    JUMPHERE(nl);
3148    }    }
3149    
3150    return mainloop;
3151    }
3152    
3153    #define MAX_N_CHARS 16
3154    #define MAX_N_BYTES 8
3155    
3156    static SLJIT_INLINE void add_prefix_byte(pcre_uint8 byte, pcre_uint8 *bytes)
3157    {
3158    pcre_uint8 len = bytes[0];
3159    int i;
3160    
3161    if (len == 255)
3162      return;
3163    
3164    if (len == 0)
3165      {
3166      bytes[0] = 1;
3167      bytes[1] = byte;
3168      return;
3169      }
3170    
3171    for (i = len; i > 0; i--)
3172      if (bytes[i] == byte)
3173        return;
3174    
3175    if (len >= MAX_N_BYTES - 1)
3176      {
3177      bytes[0] = 255;
3178      return;
3179      }
3180    
3181    len++;
3182    bytes[len] = byte;
3183    bytes[0] = len;
3184    }
3185    
3186    static int scan_prefix(compiler_common *common, pcre_uchar *cc, pcre_uint32 *chars, pcre_uint8 *bytes, int max_chars)
3187    {
3188    /* Recursive function, which scans prefix literals. */
3189    BOOL last, any, caseless;
3190    int len, repeat, len_save, consumed = 0;
3191    pcre_uint32 chr, mask;
3192    pcre_uchar *alternative, *cc_save, *oc;
3193    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3194    pcre_uchar othercase[8];
3195    #elif defined SUPPORT_UTF && defined COMPILE_PCRE16
3196    pcre_uchar othercase[2];
3197    #else
3198    pcre_uchar othercase[1];
3199    #endif
3200    
3201    repeat = 1;
3202    while (TRUE)
3203      {
3204      last = TRUE;
3205      any = FALSE;
3206      caseless = FALSE;
3207      switch (*cc)
3208        {
3209        case OP_CHARI:
3210        caseless = TRUE;
3211        case OP_CHAR:
3212        last = FALSE;
3213        cc++;
3214        break;
3215    
3216        case OP_SOD:
3217        case OP_SOM:
3218        case OP_SET_SOM:
3219        case OP_NOT_WORD_BOUNDARY:
3220        case OP_WORD_BOUNDARY:
3221        case OP_EODN:
3222        case OP_EOD:
3223        case OP_CIRC:
3224        case OP_CIRCM:
3225        case OP_DOLL:
3226        case OP_DOLLM:
3227        /* Zero width assertions. */
3228        cc++;
3229        continue;
3230    
3231        case OP_ASSERT:
3232        case OP_ASSERT_NOT:
3233        case OP_ASSERTBACK:
3234        case OP_ASSERTBACK_NOT:
3235        cc = bracketend(cc);
3236        continue;
3237    
3238        case OP_PLUSI:
3239        case OP_MINPLUSI:
3240        case OP_POSPLUSI:
3241        caseless = TRUE;
3242        case OP_PLUS:
3243        case OP_MINPLUS:
3244        case OP_POSPLUS:
3245        cc++;
3246        break;
3247    
3248        case OP_EXACTI:
3249        caseless = TRUE;
3250        case OP_EXACT:
3251        repeat = GET2(cc, 1);
3252        last = FALSE;
3253        cc += 1 + IMM2_SIZE;
3254        break;
3255    
3256        case OP_QUERYI:
3257        case OP_MINQUERYI:
3258        case OP_POSQUERYI:
3259        caseless = TRUE;
3260        case OP_QUERY:
3261        case OP_MINQUERY:
3262        case OP_POSQUERY:
3263        len = 1;
3264        cc++;
3265    #ifdef SUPPORT_UTF
3266        if (common->utf && HAS_EXTRALEN(*cc)) len += GET_EXTRALEN(*cc);
3267    #endif
3268        max_chars = scan_prefix(common, cc + len, chars, bytes, max_chars);
3269        if (max_chars == 0)
3270          return consumed;
3271        last = FALSE;
3272        break;
3273    
3274        case OP_KET:
3275        cc += 1 + LINK_SIZE;
3276        continue;
3277    
3278        case OP_ALT:
3279        cc += GET(cc, 1);
3280        continue;
3281    
3282        case OP_ONCE:
3283        case OP_ONCE_NC:
3284        case OP_BRA:
3285        case OP_BRAPOS:
3286        case OP_CBRA:
3287        case OP_CBRAPOS:
3288        alternative = cc + GET(cc, 1);
3289        while (*alternative == OP_ALT)
3290          {
3291          max_chars = scan_prefix(common, alternative + 1 + LINK_SIZE, chars, bytes, max_chars);
3292          if (max_chars == 0)
3293            return consumed;
3294          alternative += GET(alternative, 1);
3295          }
3296    
3297        if (*cc == OP_CBRA || *cc == OP_CBRAPOS)
3298          cc += IMM2_SIZE;
3299        cc += 1 + LINK_SIZE;
3300        continue;
3301    
3302        case OP_CLASS:
3303    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3304        if (common->utf && !is_char7_bitset((const pcre_uint8 *)(cc + 1), FALSE)) return consumed;
3305    #endif
3306        any = TRUE;
3307        cc += 1 + 32 / sizeof(pcre_uchar);
3308        break;
3309    
3310        case OP_NCLASS:
3311    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3312        if (common->utf) return consumed;
3313    #endif
3314        any = TRUE;
3315        cc += 1 + 32 / sizeof(pcre_uchar);
3316        break;
3317    
3318    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3319        case OP_XCLASS:
3320    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3321        if (common->utf) return consumed;
3322    #endif
3323        any = TRUE;
3324        cc += GET(cc, 1);
3325        break;
3326    #endif
3327    
3328        case OP_DIGIT:
3329    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3330        if (common->utf && !is_char7_bitset((const pcre_uint8 *)common->ctypes - cbit_length + cbit_digit, FALSE))
3331          return consumed;
3332    #endif
3333        any = TRUE;
3334        cc++;
3335        break;
3336    
3337        case OP_WHITESPACE:
3338    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3339        if (common->utf && !is_char7_bitset((const pcre_uint8 *)common->ctypes - cbit_length + cbit_space, FALSE))
3340          return consumed;
3341    #endif
3342        any = TRUE;
3343        cc++;
3344        break;
3345    
3346        case OP_WORDCHAR:
3347    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3348        if (common->utf && !is_char7_bitset((const pcre_uint8 *)common->ctypes - cbit_length + cbit_word, FALSE))
3349          return consumed;
3350    #endif
3351        any = TRUE;
3352        cc++;
3353        break;
3354    
3355        case OP_NOT:
3356        case OP_NOTI:
3357        cc++;
3358        /* Fall through. */
3359        case OP_NOT_DIGIT:
3360        case OP_NOT_WHITESPACE:
3361        case OP_NOT_WORDCHAR:
3362        case OP_ANY:
3363        case OP_ALLANY:
3364    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3365        if (common->utf) return consumed;
3366    #endif
3367        any = TRUE;
3368        cc++;
3369        break;
3370    
3371    #ifdef SUPPORT_UCP
3372        case OP_NOTPROP:
3373        case OP_PROP:
3374    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3375        if (common->utf) return consumed;
3376    #endif
3377        any = TRUE;
3378        cc += 1 + 2;
3379        break;
3380    #endif
3381    
3382        case OP_TYPEEXACT:
3383        repeat = GET2(cc, 1);
3384        cc += 1 + IMM2_SIZE;
3385        continue;
3386    
3387        case OP_NOTEXACT:
3388        case OP_NOTEXACTI:
3389    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3390        if (common->utf) return consumed;
3391    #endif
3392        any = TRUE;
3393        repeat = GET2(cc, 1);
3394        cc += 1 + IMM2_SIZE + 1;
3395        break;
3396    
3397        default:
3398        return consumed;
3399        }
3400    
3401      if (any)
3402        {
3403    #if defined COMPILE_PCRE8
3404        mask = 0xff;
3405    #elif defined COMPILE_PCRE16
3406        mask = 0xffff;
3407    #elif defined COMPILE_PCRE32
3408        mask = 0xffffffff;
3409    #else
3410        SLJIT_ASSERT_STOP();
3411    #endif
3412    
3413        do
3414          {
3415          chars[0] = mask;
3416          chars[1] = mask;
3417          bytes[0] = 255;
3418    
3419          consumed++;
3420          if (--max_chars == 0)
3421            return consumed;
3422          chars += 2;
3423          bytes += MAX_N_BYTES;
3424          }
3425        while (--repeat > 0);
3426    
3427        repeat = 1;
3428        continue;
3429        }
3430    
3431      len = 1;
3432    #ifdef SUPPORT_UTF
3433      if (common->utf && HAS_EXTRALEN(*cc)) len += GET_EXTRALEN(*cc);
3434    #endif
3435    
3436      if (caseless && char_has_othercase(common, cc))
3437        {
3438    #ifdef SUPPORT_UTF
3439        if (common->utf)
3440          {
3441          GETCHAR(chr, cc);
3442          if ((int)PRIV(ord2utf)(char_othercase(common, chr), othercase) != len)
3443            return consumed;
3444          }
3445        else
3446    #endif
3447          {
3448          chr = *cc;
3449          othercase[0] = TABLE_GET(chr, common->fcc, chr);
3450          }
3451        }
3452      else
3453        caseless = FALSE;
3454    
3455      len_save = len;
3456      cc_save = cc;
3457      while (TRUE)
3458        {
3459        oc = othercase;
3460        do
3461          {
3462          chr = *cc;
3463    #ifdef COMPILE_PCRE32
3464          if (SLJIT_UNLIKELY(chr == NOTACHAR))
3465            return consumed;
3466    #endif
3467          add_prefix_byte((pcre_uint8)chr, bytes);
3468    
3469          mask = 0;
3470          if (caseless)
3471            {
3472            add_prefix_byte((pcre_uint8)*oc, bytes);
3473            mask = *cc ^ *oc;
3474            chr |= mask;
3475            }
3476    
3477    #ifdef COMPILE_PCRE32
3478          if (chars[0] == NOTACHAR && chars[1] == 0)
3479    #else
3480          if (chars[0] == NOTACHAR)
3481    #endif
3482            {
3483            chars[0] = chr;
3484            chars[1] = mask;
3485            }
3486          else
3487            {
3488            mask |= chars[0] ^ chr;
3489            chr |= mask;
3490            chars[0] = chr;
3491            chars[1] |= mask;
3492            }
3493    
3494          len--;
3495          consumed++;
3496          if (--max_chars == 0)
3497            return consumed;
3498          chars += 2;
3499          bytes += MAX_N_BYTES;
3500          cc++;
3501          oc++;
3502          }
3503        while (len > 0);
3504    
3505        if (--repeat == 0)
3506          break;
3507    
3508  return mainloop;      len = len_save;
3509  }      cc = cc_save;
3510        }
3511    
3512  #define MAX_N_CHARS 3    repeat = 1;
3513      if (last)
3514        return consumed;
3515      }
3516    }
3517    
3518  static SLJIT_INLINE BOOL fast_forward_first_n_chars(compiler_common *common, BOOL firstline)  static SLJIT_INLINE BOOL fast_forward_first_n_chars(compiler_common *common, BOOL firstline)
3519  {  {
# Line 2573  DEFINE_COMPILER; Line 3521  DEFINE_COMPILER;
3521  struct sljit_label *start;  struct sljit_label *start;
3522  struct sljit_jump *quit;  struct sljit_jump *quit;
3523  pcre_uint32 chars[MAX_N_CHARS * 2];  pcre_uint32 chars[MAX_N_CHARS * 2];
3524  pcre_uchar *cc = common->start + 1 + LINK_SIZE;  pcre_uint8 bytes[MAX_N_CHARS * MAX_N_BYTES];
3525  int location = 0;  pcre_uint8 ones[MAX_N_CHARS];
3526  pcre_int32 len, c, bit, caseless;  int offsets[3];
3527  int must_stop;  pcre_uint32 mask;
3528    pcre_uint8 *byte_set, *byte_set_end;
3529    int i, max, from;
3530    int range_right = -1, range_len = 3 - 1;
3531    sljit_ub *update_table = NULL;
3532    BOOL in_range;
3533    
3534    /* This is even TRUE, if both are NULL. */
3535    SLJIT_ASSERT(common->read_only_data_ptr == common->read_only_data);
3536    
3537    for (i = 0; i < MAX_N_CHARS; i++)
3538      {
3539      chars[i << 1] = NOTACHAR;
3540      chars[(i << 1) + 1] = 0;
3541      bytes[i * MAX_N_BYTES] = 0;
3542      }
3543    
3544  /* We do not support alternatives now. */  max = scan_prefix(common, common->start, chars, bytes, MAX_N_CHARS);
3545  if (*(common->start + GET(common->start, 1)) == OP_ALT)  
3546    if (max <= 1)
3547    return FALSE;    return FALSE;
3548    
3549  while (TRUE)  for (i = 0; i < max; i++)
3550    {    {
3551    caseless = 0;    mask = chars[(i << 1) + 1];
3552    must_stop = 1;    ones[i] = ones_in_half_byte[mask & 0xf];
3553    switch(*cc)    mask >>= 4;
3554      while (mask != 0)
3555      {      {
3556      case OP_CHAR:      ones[i] += ones_in_half_byte[mask & 0xf];
3557      must_stop = 0;      mask >>= 4;
3558      cc++;      }
3559      break;    }
3560    
3561      case OP_CHARI:  in_range = FALSE;
3562      caseless = 1;  for (i = 0; i <= max; i++)
3563      must_stop = 0;    {
3564      cc++;    if (in_range && (i - from) > range_len && (bytes[(i - 1) * MAX_N_BYTES] <= 4))
3565      break;      {
3566        range_len = i - from;
3567        range_right = i - 1;
3568        }
3569    
3570      case OP_SOD:    if (i < max && bytes[i * MAX_N_BYTES] < 255)
3571      case OP_SOM:      {
3572      case OP_SET_SOM:      if (!in_range)
3573      case OP_NOT_WORD_BOUNDARY:        {
3574      case OP_WORD_BOUNDARY:        in_range = TRUE;
3575      case OP_EODN:        from = i;
3576      case OP_EOD:        }
3577      case OP_CIRC:      }
3578      case OP_CIRCM:    else if (in_range)
3579      case OP_DOLL:      in_range = FALSE;
3580      case OP_DOLLM:    }
     /* Zero width assertions. */  
     cc++;  
     continue;  
3581    
3582      case OP_PLUS:  if (range_right >= 0)
3583      case OP_MINPLUS:    {
3584      case OP_POSPLUS:    /* Since no data is consumed (see the assert in the beginning
3585      cc++;    of this function), this space can be reallocated. */
3586      break;    if (common->read_only_data)
3587        SLJIT_FREE(common->read_only_data);
3588    
3589      case OP_EXACT:    common->read_only_data_size += 256;
3590      cc += 1 + IMM2_SIZE;    common->read_only_data = (sljit_uw *)SLJIT_MALLOC(common->read_only_data_size);
3591      break;    if (common->read_only_data == NULL)
3592        return TRUE;
3593    
3594      case OP_PLUSI:    update_table = (sljit_ub *)common->read_only_data;
3595      case OP_MINPLUSI:    common->read_only_data_ptr = (sljit_uw *)(update_table + 256);
3596      case OP_POSPLUSI:    memset(update_table, IN_UCHARS(range_len), 256);
     caseless = 1;  
     cc++;  
     break;  
3597    
3598      case OP_EXACTI:    for (i = 0; i < range_len; i++)
3599      caseless = 1;      {
3600      cc += 1 + IMM2_SIZE;      byte_set = bytes + ((range_right - i) * MAX_N_BYTES);
3601      break;      SLJIT_ASSERT(byte_set[0] > 0 && byte_set[0] < 255);
3602        byte_set_end = byte_set + byte_set[0];
3603        byte_set++;
3604        while (byte_set <= byte_set_end)
3605          {
3606          if (update_table[*byte_set] > IN_UCHARS(i))
3607            update_table[*byte_set] = IN_UCHARS(i);
3608          byte_set++;
3609          }
3610        }
3611      }
3612    
3613      default:  offsets[0] = -1;
3614      must_stop = 2;  /* Scan forward. */
3615    for (i = 0; i < max; i++)
3616      if (ones[i] <= 2) {
3617        offsets[0] = i;
3618      break;      break;
3619      }    }
3620    
3621    if (offsets[0] < 0 && range_right < 0)
3622      return FALSE;
3623    
3624    if (must_stop == 2)  if (offsets[0] >= 0)
3625      {
3626      /* Scan backward. */
3627      offsets[1] = -1;
3628      for (i = max - 1; i > offsets[0]; i--)
3629        if (ones[i] <= 2 && i != range_right)
3630          {
3631          offsets[1] = i;
3632        break;        break;
3633          }
3634    
3635    len = 1;    /* This case is handled better by fast_forward_first_char. */
3636  #ifdef SUPPORT_UTF    if (offsets[1] == -1 && offsets[0] == 0 && range_right < 0)
3637    if (common->utf && HAS_EXTRALEN(cc[0])) len += GET_EXTRALEN(cc[0]);      return FALSE;
 #endif  
3638    
3639    if (caseless && char_has_othercase(common, cc))    offsets[2] = -1;
3640      {    /* We only search for a middle character if there is no range check. */
3641      caseless = char_get_othercase_bit(common, cc);    if (offsets[1] >= 0 && range_right == -1)
3642      if (caseless == 0)      {
3643        return FALSE;      /* Scan from middle. */
3644  #ifdef COMPILE_PCRE8      for (i = (offsets[0] + offsets[1]) / 2 + 1; i < offsets[1]; i++)
3645      caseless = ((caseless & 0xff) << 8) | (len - (caseless >> 8));        if (ones[i] <= 2)
3646  #else          {
3647      if ((caseless & 0x100) != 0)          offsets[2] = i;
3648        caseless = ((caseless & 0xff) << 16) | (len - (caseless >> 9));          break;
3649      else          }
       caseless = ((caseless & 0xff) << 8) | (len - (caseless >> 9));  
 #endif  
     }  
   else  
     caseless = 0;  
3650    
3651    while (len > 0 && location < MAX_N_CHARS * 2)      if (offsets[2] == -1)
     {  
     c = *cc;  
     bit = 0;  
     if (len == (caseless & 0xff))  
3652        {        {
3653        bit = caseless >> 8;        for (i = (offsets[0] + offsets[1]) / 2; i > offsets[0]; i--)
3654        c |= bit;          if (ones[i] <= 2)
3655              {
3656              offsets[2] = i;
3657              break;
3658              }
3659        }        }
3660        }
3661    
3662      chars[location] = c;    SLJIT_ASSERT(offsets[1] == -1 || (offsets[0] < offsets[1]));
3663      chars[location + 1] = bit;    SLJIT_ASSERT(offsets[2] == -1 || (offsets[0] < offsets[2] && offsets[1] > offsets[2]));
3664    
3665      len--;    chars[0] = chars[offsets[0] << 1];
3666      location += 2;    chars[1] = chars[(offsets[0] << 1) + 1];
3667      cc++;    if (offsets[2] >= 0)
3668        {
3669        chars[2] = chars[offsets[2] << 1];
3670        chars[3] = chars[(offsets[2] << 1) + 1];
3671        }
3672      if (offsets[1] >= 0)
3673        {
3674        chars[4] = chars[offsets[1] << 1];
3675        chars[5] = chars[(offsets[1] << 1) + 1];
3676      }      }
   
   if (location >= MAX_N_CHARS * 2 || must_stop != 0)  
     break;  
3677    }    }
3678    
3679  /* At least two characters are required. */  max -= 1;
 if (location < 2 * 2)  
     return FALSE;  
   
3680  if (firstline)  if (firstline)
3681    {    {
3682    SLJIT_ASSERT(common->first_line_end != 0);    SLJIT_ASSERT(common->first_line_end != 0);
3683      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end);
3684    OP1(SLJIT_MOV, TMP3, 0, STR_END, 0);    OP1(SLJIT_MOV, TMP3, 0, STR_END, 0);
3685    OP2(SLJIT_SUB, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, SLJIT_IMM, IN_UCHARS((location >> 1) - 1));    OP2(SLJIT_SUB, STR_END, 0, STR_END, 0, SLJIT_IMM, IN_UCHARS(max));
3686      quit = CMP(SLJIT_C_LESS_EQUAL, STR_END, 0, TMP1, 0);
3687      OP1(SLJIT_MOV, STR_END, 0, TMP1, 0);
3688      JUMPHERE(quit);
3689    }    }
3690  else  else
3691    OP2(SLJIT_SUB, STR_END, 0, STR_END, 0, SLJIT_IMM, IN_UCHARS((location >> 1) - 1));    OP2(SLJIT_SUB, STR_END, 0, STR_END, 0, SLJIT_IMM, IN_UCHARS(max));
3692    
3693    #if !(defined SLJIT_CONFIG_X86_32 && SLJIT_CONFIG_X86_32)
3694    if (range_right >= 0)
3695      OP1(SLJIT_MOV, RETURN_ADDR, 0, SLJIT_IMM, (sljit_sw)update_table);
3696    #endif
3697    
3698  start = LABEL();  start = LABEL();
3699  quit = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);  quit = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3700    
3701  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));  SLJIT_ASSERT(range_right >= 0 || offsets[0] >= 0);
3702  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  
3703  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  if (range_right >= 0)
3704  if (chars[1] != 0)    {
3705    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, chars[1]);  #if defined COMPILE_PCRE8 || (defined SLJIT_LITTLE_ENDIAN && SLJIT_LITTLE_ENDIAN)
3706  CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, chars[0], start);    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(range_right));
3707  if (location > 2 * 2)  #else
3708    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(range_right + 1) - 1);
3709  if (chars[3] != 0)  #endif
3710    OP2(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_IMM, chars[3]);  
3711  CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, chars[2], start);  #if !(defined SLJIT_CONFIG_X86_32 && SLJIT_CONFIG_X86_32)
3712  if (location > 2 * 2)    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM2(RETURN_ADDR, TMP1), 0);
3713    {  #else
3714    if (chars[5] != 0)    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)update_table);
3715      OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, chars[5]);  #endif
3716    CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, chars[4], start);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3717      CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, 0, start);
3718      }
3719    
3720    if (offsets[0] >= 0)
3721      {
3722      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(offsets[0]));
3723      if (offsets[1] >= 0)
3724        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(offsets[1]));
3725      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3726    
3727      if (chars[1] != 0)
3728        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, chars[1]);
3729      CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, chars[0], start);
3730      if (offsets[2] >= 0)
3731        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(offsets[2] - 1));
3732    
3733      if (offsets[1] >= 0)
3734        {
3735        if (chars[5] != 0)
3736          OP2(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_IMM, chars[5]);
3737        CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, chars[4], start);
3738        }
3739    
3740      if (offsets[2] >= 0)
3741        {
3742        if (chars[3] != 0)
3743          OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, chars[3]);
3744        CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, chars[2], start);
3745        }
3746      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3747    }    }
 OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  
3748    
3749  JUMPHERE(quit);  JUMPHERE(quit);
3750    
3751  if (firstline)  if (firstline)
3752      {
3753      if (range_right >= 0)
3754        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end);
3755    OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);    OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);
3756      if (range_right >= 0)
3757        {
3758        quit = CMP(SLJIT_C_LESS_EQUAL, STR_PTR, 0, TMP1, 0);
3759        OP1(SLJIT_MOV, STR_PTR, 0, TMP1, 0);
3760        JUMPHERE(quit);
3761        }
3762      }
3763  else  else
3764    OP2(SLJIT_ADD, STR_END, 0, STR_END, 0, SLJIT_IMM, IN_UCHARS((location >> 1) - 1));    OP2(SLJIT_ADD, STR_END, 0, STR_END, 0, SLJIT_IMM, IN_UCHARS(max));
3765  return TRUE;  return TRUE;
3766  }  }
3767    
3768  #undef MAX_N_CHARS  #undef MAX_N_CHARS
3769    #undef MAX_N_BYTES
3770    
3771  static SLJIT_INLINE void fast_forward_first_char(compiler_common *common, pcre_uchar first_char, BOOL caseless, BOOL firstline)  static SLJIT_INLINE void fast_forward_first_char(compiler_common *common, pcre_uchar first_char, BOOL caseless, BOOL firstline)
3772  {  {
# Line 2838  if (common->nltype == NLTYPE_FIXED && co Line 3872  if (common->nltype == NLTYPE_FIXED && co
3872    JUMPHERE(lastchar);    JUMPHERE(lastchar);
3873    
3874    if (firstline)    if (firstline)
3875      OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), POSSESSIVE0);      OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);
3876    return;    return;
3877    }    }
3878    
# Line 2848  firstchar = CMP(SLJIT_C_LESS_EQUAL, STR_ Line 3882  firstchar = CMP(SLJIT_C_LESS_EQUAL, STR_
3882  skip_char_back(common);  skip_char_back(common);
3883    
3884  loop = LABEL();  loop = LABEL();
3885  read_char(common);  common->ff_newline_shortcut = loop;
3886    
3887    read_char_range(common, common->nlmin, common->nlmax, TRUE);
3888  lastchar = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);  lastchar = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3889  if (common->nltype == NLTYPE_ANY || common->nltype == NLTYPE_ANYCRLF)  if (common->nltype == NLTYPE_ANY || common->nltype == NLTYPE_ANYCRLF)
3890    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 3913  if (firstline)
3913    OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);    OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);
3914  }  }
3915    
3916  static BOOL check_class_ranges(compiler_common *common, const pcre_uint8 *bits, BOOL nclass, jump_list **backtracks);  static BOOL check_class_ranges(compiler_common *common, const pcre_uint8 *bits, BOOL nclass, BOOL invert, jump_list **backtracks);
3917    
3918  static SLJIT_INLINE void fast_forward_start_bits(compiler_common *common, sljit_uw start_bits, BOOL firstline)  static SLJIT_INLINE void fast_forward_start_bits(compiler_common *common, pcre_uint8 *start_bits, BOOL firstline)
3919  {  {
3920  DEFINE_COMPILER;  DEFINE_COMPILER;
3921  struct sljit_label *start;  struct sljit_label *start;
3922  struct sljit_jump *quit;  struct sljit_jump *quit;
3923  struct sljit_jump *found = NULL;  struct sljit_jump *found = NULL;
3924  jump_list *matches = NULL;  jump_list *matches = NULL;
 pcre_uint8 inverted_start_bits[32];  
 int i;  
3925  #ifndef COMPILE_PCRE8  #ifndef COMPILE_PCRE8
3926  struct sljit_jump *jump;  struct sljit_jump *jump;
3927  #endif  #endif
3928    
 for (i = 0; i < 32; ++i)  
   inverted_start_bits[i] = ~(((pcre_uint8*)start_bits)[i]);  
   
3929  if (firstline)  if (firstline)
3930    {    {
3931    SLJIT_ASSERT(common->first_line_end != 0);    SLJIT_ASSERT(common->first_line_end != 0);
# Line 2910  if (common->utf) Line 3941  if (common->utf)
3941    OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);    OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);
3942  #endif  #endif
3943    
3944  if (!check_class_ranges(common, inverted_start_bits, (inverted_start_bits[31] & 0x80) != 0, &matches))  if (!check_class_ranges(common, start_bits, (start_bits[31] & 0x80) != 0, TRUE, &matches))
3945    {    {
3946  #ifndef COMPILE_PCRE8  #ifndef COMPILE_PCRE8
3947    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 255);    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 255);
# Line 2919  if (!check_class_ranges(common, inverted Line 3950  if (!check_class_ranges(common, inverted
3950  #endif  #endif
3951    OP2(SLJIT_AND, TMP2, 0, TMP1, 0, SLJIT_IMM, 0x7);    OP2(SLJIT_AND, TMP2, 0, TMP1, 0, SLJIT_IMM, 0x7);
3952    OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 3);    OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 3);
3953    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), start_bits);    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)start_bits);
3954    OP2(SLJIT_SHL, TMP2, 0, SLJIT_IMM, 1, TMP2, 0);    OP2(SLJIT_SHL, TMP2, 0, SLJIT_IMM, 1, TMP2, 0);
3955    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);
3956    found = JUMP(SLJIT_C_NOT_ZERO);    found = JUMP(SLJIT_C_NOT_ZERO);
# Line 3037  GET_LOCAL_BASE(TMP3, 0, 0); Line 4068  GET_LOCAL_BASE(TMP3, 0, 0);
4068  /* Drop frames until we reach STACK_TOP. */  /* Drop frames until we reach STACK_TOP. */
4069  mainloop = LABEL();  mainloop = LABEL();
4070  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(TMP1), 0);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(TMP1), 0);
4071  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);
4072    jump = JUMP(SLJIT_C_SIG_LESS_EQUAL);
4073    
4074  OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, TMP3, 0);  OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, TMP3, 0);
4075  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));
4076  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 3045  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_I Line 4078  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_I
4078  JUMPTO(SLJIT_JUMP, mainloop);  JUMPTO(SLJIT_JUMP, mainloop);
4079    
4080  JUMPHERE(jump);  JUMPHERE(jump);
4081  jump = CMP(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, frame_end);  jump = JUMP(SLJIT_C_SIG_LESS);
4082  /* End of dropping frames. */  /* End of dropping frames. */
4083  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
4084    
4085  JUMPHERE(jump);  JUMPHERE(jump);
4086  jump = CMP(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, frame_setstrbegin);  OP1(SLJIT_NEG, TMP2, 0, TMP2, 0);
4087  /* Set string begin. */  OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, TMP3, 0);
4088  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. */  
4089  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));
4090  JUMPTO(SLJIT_JUMP, mainloop);  JUMPTO(SLJIT_JUMP, mainloop);
4091  }  }
# Line 3078  static void check_wordboundary(compiler_ Line 4094  static void check_wordboundary(compiler_
4094  {  {
4095  DEFINE_COMPILER;  DEFINE_COMPILER;
4096  struct sljit_jump *skipread;  struct sljit_jump *skipread;
4097    jump_list *skipread_list = NULL;
4098  #if !(defined COMPILE_PCRE8) || defined SUPPORT_UTF  #if !(defined COMPILE_PCRE8) || defined SUPPORT_UTF
4099  struct sljit_jump *jump;  struct sljit_jump *jump;
4100  #endif  #endif
# Line 3135  else Line 4152  else
4152  JUMPHERE(skipread);  JUMPHERE(skipread);
4153    
4154  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 0);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 0);
4155  skipread = check_str_end(common);  check_str_end(common, &skipread_list);
4156  peek_char(common);  peek_char(common, READ_CHAR_MAX);
4157    
4158  /* Testing char type. This is a code duplication. */  /* Testing char type. This is a code duplication. */
4159  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 3176  else Line 4193  else
4193      JUMPHERE(jump);      JUMPHERE(jump);
4194  #endif /* COMPILE_PCRE8 */  #endif /* COMPILE_PCRE8 */
4195    }    }
4196  JUMPHERE(skipread);  set_jumps(skipread_list, LABEL());
4197    
4198  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);
4199  sljit_emit_fast_return(compiler, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0);  sljit_emit_fast_return(compiler, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0);
4200  }  }
4201    
4202  /*  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)  
4203  {  {
4204  DEFINE_COMPILER;  DEFINE_COMPILER;
4205  struct sljit_jump *jump;  int ranges[MAX_RANGE_SIZE];
4206    pcre_uint8 bit, cbit, all;
4207    int i, byte, length = 0;
4208    
4209    bit = bits[0] & 0x1;
4210    /* All bits will be zero or one (since bit is zero or one). */
4211    all = -bit;
4212    
4213    for (i = 0; i < 256; )
4214      {
4215      byte = i >> 3;
4216      if ((i & 0x7) == 0 && bits[byte] == all)
4217        i += 8;
4218      else
4219        {
4220        cbit = (bits[byte] >> (i & 0x7)) & 0x1;
4221        if (cbit != bit)
4222          {
4223          if (length >= MAX_RANGE_SIZE)
4224            return FALSE;
4225          ranges[length] = i;
4226          length++;
4227          bit = cbit;
4228          all = -cbit;
4229          }
4230        i++;
4231        }
4232      }
4233    
4234    if (((bit == 0) && nclass) || ((bit == 1) && !nclass))
4235      {
4236      if (length >= MAX_RANGE_SIZE)
4237        return FALSE;
4238      ranges[length] = 256;
4239      length++;
4240      }
4241    
4242  if (ranges[0] < 0)  if (length < 0 || length > 4)
4243    return FALSE;    return FALSE;
4244    
4245  switch(ranges[0])  bit = bits[0] & 0x1;
4246    if (invert) bit ^= 0x1;
4247    
4248    /* No character is accepted. */
4249    if (length == 0 && bit == 0)
4250      add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));
4251    
4252    switch(length)
4253    {    {
4254      case 0:
4255      /* When bit != 0, all characters are accepted. */
4256      return TRUE;
4257    
4258    case 1:    case 1:
4259    if (readch)    add_jump(compiler, backtracks, CMP(bit == 0 ? SLJIT_C_LESS : SLJIT_C_GREATER_EQUAL, TMP1, 0, SLJIT_IMM, ranges[0]));
     read_char(common);  
   add_jump(compiler, backtracks, CMP(ranges[1] == 0 ? SLJIT_C_LESS : SLJIT_C_GREATER_EQUAL, TMP1, 0, SLJIT_IMM, ranges[2]));  
4260    return TRUE;    return TRUE;
4261    
4262    case 2:    case 2:
4263    if (readch)    if (ranges[0] + 1 != ranges[1])
4264      read_char(common);      {
4265    OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[2]);      OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[0]);
4266    add_jump(compiler, backtracks, CMP(ranges[1] != 0 ? SLJIT_C_LESS : SLJIT_C_GREATER_EQUAL, TMP1, 0, SLJIT_IMM, ranges[3] - ranges[2]));      add_jump(compiler, backtracks, CMP(bit != 0 ? SLJIT_C_LESS : SLJIT_C_GREATER_EQUAL, TMP1, 0, SLJIT_IMM, ranges[1] - ranges[0]));
4267        }
4268      else
4269        add_jump(compiler, backtracks, CMP(bit != 0 ? SLJIT_C_EQUAL : SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, ranges[0]));
4270    return TRUE;    return TRUE;
4271    
4272    case 4:    case 3:
4273    if (ranges[2] + 1 == ranges[3] && ranges[4] + 1 == ranges[5])    if (bit != 0)
4274      {      {
4275      if (readch)      add_jump(compiler, backtracks, CMP(SLJIT_C_GREATER_EQUAL, TMP1, 0, SLJIT_IMM, ranges[2]));
4276        read_char(common);      if (ranges[0] + 1 != ranges[1])
     if (ranges[1] != 0)  
4277        {        {
4278        add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, ranges[2]));        OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[0]);
4279        add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, ranges[4]));        add_jump(compiler, backtracks, CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, ranges[1] - ranges[0]));
4280        }        }
4281      else      else
4282        {        add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, ranges[0]));
       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);  
       }  
4283      return TRUE;      return TRUE;
4284      }      }
4285    if ((ranges[3] - ranges[2]) == (ranges[5] - ranges[4]) && is_powerof2(ranges[4] - ranges[2]))  
4286      add_jump(compiler, backtracks, CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, ranges[0]));
4287      if (ranges[1] + 1 != ranges[2])
4288      {      {
4289      if (readch)      OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[1]);
4290        read_char(common);      add_jump(compiler, backtracks, CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, ranges[2] - ranges[1]));
     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;  
4291      }      }
4292    return FALSE;    else
4293        add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, ranges[1]));
4294    default:    return TRUE;
   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;  
4295    
4296  for (i = 1; i < 256; i++)    case 4:
4297    if ((ctypes[i] & flag) != bit)    if ((ranges[1] - ranges[0]) == (ranges[3] - ranges[2])
4298          && (ranges[0] | (ranges[2] - ranges[0])) == ranges[2]
4299          && is_powerof2(ranges[2] - ranges[0]))
4300      {      {
4301      if (length >= MAX_RANGE_SIZE)      OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[2] - ranges[0]);
4302        return;      if (ranges[2] + 1 != ranges[3])
4303      ranges[2 + length] = i;        {
4304      length++;        OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[2]);
4305      bit ^= flag;        add_jump(compiler, backtracks, CMP(bit != 0 ? SLJIT_C_LESS : SLJIT_C_GREATER_EQUAL, TMP1, 0, SLJIT_IMM, ranges[3] - ranges[2]));
4306          }
4307        else
4308          add_jump(compiler, backtracks, CMP(bit != 0 ? SLJIT_C_EQUAL : SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, ranges[2]));
4309        return TRUE;
4310      }      }
4311    
4312  if (bit != 0)    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];  
 pcre_uint8 bit, cbit, all;  
 int i, byte, length = 0;  
   
 bit = bits[0] & 0x1;  
 ranges[1] = bit;  
 /* Can be 0 or 255. */  
 all = -bit;  
   
 for (i = 0; i < 256; )  
   {  
   byte = i >> 3;  
   if ((i & 0x7) == 0 && bits[byte] == all)  
     i += 8;  
   else  
4313      {      {
4314      cbit = (bits[byte] >> (i & 0x7)) & 0x1;      i = 0;
4315      if (cbit != bit)      if (ranges[0] + 1 != ranges[1])
4316        {        {
4317        if (length >= MAX_RANGE_SIZE)        OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[0]);
4318          return FALSE;        add_jump(compiler, backtracks, CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, ranges[1] - ranges[0]));
4319        ranges[2 + length] = i;        i = ranges[0];
4320        length++;        }
4321        bit = cbit;      else
4322        all = -cbit;        add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, ranges[0]));
4323    
4324        if (ranges[2] + 1 != ranges[3])
4325          {
4326          OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[2] - i);
4327          add_jump(compiler, backtracks, CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, ranges[3] - ranges[2]));
4328        }        }
4329      i++;      else
4330          add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, ranges[2] - i));
4331        return TRUE;
4332      }      }
   }  
4333    
4334  if (((bit == 0) && nclass) || ((bit == 1) && !nclass))    OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[0]);
4335    {    add_jump(compiler, backtracks, CMP(SLJIT_C_GREATER_EQUAL, TMP1, 0, SLJIT_IMM, ranges[3] - ranges[0]));
4336    if (length >= MAX_RANGE_SIZE)    if (ranges[1] + 1 != ranges[2])
4337      return FALSE;      {
4338    ranges[2 + length] = 256;      OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[1] - ranges[0]);
4339    length++;      add_jump(compiler, backtracks, CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, ranges[2] - ranges[1]));
4340    }      }
4341  ranges[0] = length;    else
4342        add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, ranges[1] - ranges[0]));
4343      return TRUE;
4344    
4345  return check_ranges(common, ranges, backtracks, FALSE);    default:
4346      SLJIT_ASSERT_STOP();
4347      return FALSE;
4348      }
4349  }  }
4350    
4351  static void check_anynewline(compiler_common *common)  static void check_anynewline(compiler_common *common)
# Line 3496  sljit_emit_fast_return(compiler, RETURN_ Line 4523  sljit_emit_fast_return(compiler, RETURN_
4523    
4524  #if defined SUPPORT_UTF && defined SUPPORT_UCP  #if defined SUPPORT_UTF && defined SUPPORT_UCP
4525    
4526  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)
4527  {  {
4528  /* This function would be ineffective to do in JIT level. */  /* This function would be ineffective to do in JIT level. */
4529  pcre_uint32 c1, c2;  pcre_uint32 c1, c2;
# Line 3685  return cc; Line 4712  return cc;
4712  #define SET_TYPE_OFFSET(value) \  #define SET_TYPE_OFFSET(value) \
4713    if ((value) != typeoffset) \    if ((value) != typeoffset) \
4714      { \      { \
4715      if ((value) > typeoffset) \      if ((value) < typeoffset) \
       OP2(SLJIT_SUB, typereg, 0, typereg, 0, SLJIT_IMM, (value) - typeoffset); \  
     else \  
4716        OP2(SLJIT_ADD, typereg, 0, typereg, 0, SLJIT_IMM, typeoffset - (value)); \        OP2(SLJIT_ADD, typereg, 0, typereg, 0, SLJIT_IMM, typeoffset - (value)); \
4717        else \
4718          OP2(SLJIT_SUB, typereg, 0, typereg, 0, SLJIT_IMM, (value) - typeoffset); \
4719      } \      } \
4720    typeoffset = (value);    typeoffset = (value);
4721    
4722  #define SET_CHAR_OFFSET(value) \  #define SET_CHAR_OFFSET(value) \
4723    if ((value) != charoffset) \    if ((value) != charoffset) \
4724      { \      { \
4725      if ((value) > charoffset) \      if ((value) < charoffset) \
4726        OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, (value) - charoffset); \        OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(charoffset - (value))); \
4727      else \      else \
4728        OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, charoffset - (value)); \        OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)((value) - charoffset)); \
4729      } \      } \
4730    charoffset = (value);    charoffset = (value);
4731    
# Line 3706  static void compile_xclass_matchingpath( Line 4733  static void compile_xclass_matchingpath(
4733  {  {
4734  DEFINE_COMPILER;  DEFINE_COMPILER;
4735  jump_list *found = NULL;  jump_list *found = NULL;
4736  jump_list **list = (*cc & XCL_NOT) == 0 ? &found : backtracks;  jump_list **list = (cc[0] & XCL_NOT) == 0 ? &found : backtracks;
4737  pcre_int32 c, charoffset;  sljit_uw c, charoffset, max = 256, min = READ_CHAR_MAX;
 const pcre_uint32 *other_cases;  
4738  struct sljit_jump *jump = NULL;  struct sljit_jump *jump = NULL;
4739  pcre_uchar *ccbegin;  pcre_uchar *ccbegin;
4740  int compares, invertcmp, numberofcmps;  int compares, invertcmp, numberofcmps;
4741    #if defined SUPPORT_UTF && (defined COMPILE_PCRE8 || defined COMPILE_PCRE16)
4742    BOOL utf = common->utf;
4743    #endif
4744    
4745  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4746  BOOL needstype = FALSE, needsscript = FALSE, needschar = FALSE;  BOOL needstype = FALSE, needsscript = FALSE, needschar = FALSE;
4747  BOOL charsaved = FALSE;  BOOL charsaved = FALSE;
4748  int typereg = TMP1, scriptreg = TMP1;  int typereg = TMP1, scriptreg = TMP1;
4749  pcre_int32 typeoffset;  const pcre_uint32 *other_cases;
4750    sljit_uw typeoffset;
4751  #endif  #endif
4752    
4753  /* Although SUPPORT_UTF must be defined, we are  /* Scanning the necessary info. */
4754     not necessary in utf mode even in 8 bit mode. */  cc++;
4755  detect_partial_match(common, backtracks);  ccbegin = cc;
4756  read_char(common);  compares = 0;
4757    if (cc[-1] & XCL_MAP)
 if ((*cc++ & XCL_MAP) != 0)  
4758    {    {
4759    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  
4760    cc += 32 / sizeof(pcre_uchar);    cc += 32 / sizeof(pcre_uchar);
4761    }    }
4762    
 /* Scanning the necessary info. */  
 ccbegin = cc;  
 compares = 0;  
4763  while (*cc != XCL_END)  while (*cc != XCL_END)
4764    {    {
4765    compares++;    compares++;
4766    if (*cc == XCL_SINGLE)    if (*cc == XCL_SINGLE)
4767      {      {
4768      cc += 2;      cc ++;
4769  #ifdef SUPPORT_UTF      GETCHARINCTEST(c, cc);
4770      if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (c > max) max = c;
4771  #endif      if (c < min) min = c;
4772  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4773      needschar = TRUE;      needschar = TRUE;
4774  #endif  #endif
4775      }      }
4776    else if (*cc == XCL_RANGE)    else if (*cc == XCL_RANGE)
4777      {      {
4778      cc += 2;      cc ++;
4779  #ifdef SUPPORT_UTF      GETCHARINCTEST(c, cc);
4780      if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);      if (c < min) min = c;
4781  #endif      GETCHARINCTEST(c, cc);
4782      cc++;      if (c > max) max = c;
 #ifdef SUPPORT_UTF  
     if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);  
 #endif  
4783  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4784      needschar = TRUE;      needschar = TRUE;
4785  #endif  #endif
# Line 3792  while (*cc != XCL_END) Line 4789  while (*cc != XCL_END)
4789      {      {
4790      SLJIT_ASSERT(*cc == XCL_PROP || *cc == XCL_NOTPROP);      SLJIT_ASSERT(*cc == XCL_PROP || *cc == XCL_NOTPROP);
4791      cc++;      cc++;
4792        if (*cc == PT_CLIST)
4793          {
4794          other_cases = PRIV(ucd_caseless_sets) + cc[1];
4795          while (*other_cases != NOTACHAR)
4796            {
4797            if (*other_cases > max) max = *other_cases;
4798            if (*other_cases < min) min = *other_cases;
4799            other_cases++;
4800            }
4801          }
4802        else
4803          {
4804          max = READ_CHAR_MAX;
4805          min = 0;
4806          }
4807    
4808      switch(*cc)      switch(*cc)
4809        {        {
4810        case PT_ANY:        case PT_ANY:
# Line 3811  while (*cc != XCL_END) Line 4824  while (*cc != XCL_END)
4824        case PT_SPACE:        case PT_SPACE:
4825        case PT_PXSPACE:        case PT_PXSPACE:
4826        case PT_WORD:        case PT_WORD:
4827          case PT_PXGRAPH:
4828          case PT_PXPRINT:
4829          case PT_PXPUNCT:
4830        needstype = TRUE;        needstype = TRUE;
4831        needschar = TRUE;        needschar = TRUE;
4832        break;        break;
4833    
4834        case PT_CLIST:        case PT_CLIST:
4835          case PT_UCNC:
4836        needschar = TRUE;        needschar = TRUE;
4837        break;        break;
4838    
# Line 3828  while (*cc != XCL_END) Line 4845  while (*cc != XCL_END)
4845  #endif  #endif
4846    }    }
4847    
4848    /* We are not necessary in utf mode even in 8 bit mode. */
4849    cc = ccbegin;
4850    detect_partial_match(common, backtracks);
4851    read_char_range(common, min, max, (cc[-1] & XCL_NOT) != 0);
4852    
4853    if ((cc[-1] & XCL_HASPROP) == 0)
4854      {
4855      if ((cc[-1] & XCL_MAP) != 0)
4856        {
4857        jump = CMP(SLJIT_C_GREATER, TMP1, 0, SLJIT_IMM, 255);
4858        if (!check_class_ranges(common, (const pcre_uint8 *)cc, (((const pcre_uint8 *)cc)[31] & 0x80) != 0, TRUE, &found))
4859          {
4860          OP2(SLJIT_AND, TMP2, 0, TMP1, 0, SLJIT_IMM, 0x7);
4861          OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 3);
4862          OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)cc);
4863          OP2(SLJIT_SHL, TMP2, 0, SLJIT_IMM, 1, TMP2, 0);
4864          OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);
4865          add_jump(compiler, &found, JUMP(SLJIT_C_NOT_ZERO));
4866          }
4867    
4868        add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));
4869        JUMPHERE(jump);
4870    
4871        cc += 32 / sizeof(pcre_uchar);
4872        }
4873      else
4874        {
4875        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, min);
4876        add_jump(compiler, (cc[-1] & XCL_NOT) == 0 ? backtracks : &found, CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, max - min));
4877        }
4878      }
4879    else if ((cc[-1] & XCL_MAP) != 0)
4880      {
4881      OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);
4882    #ifdef SUPPORT_UCP
4883      charsaved = TRUE;
4884    #endif
4885      if (!check_class_ranges(common, (const pcre_uint8 *)cc, FALSE, TRUE, list))
4886        {
4887    #ifdef COMPILE_PCRE8
4888        SLJIT_ASSERT(common->utf);
4889    #endif
4890        jump = CMP(SLJIT_C_GREATER, TMP1, 0, SLJIT_IMM, 255);
4891    
4892        OP2(SLJIT_AND, TMP2, 0, TMP1, 0, SLJIT_IMM, 0x7);
4893        OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 3);
4894        OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)cc);
4895        OP2(SLJIT_SHL, TMP2, 0, SLJIT_IMM, 1, TMP2, 0);
4896        OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);
4897        add_jump(compiler, list, JUMP(SLJIT_C_NOT_ZERO));
4898    
4899        JUMPHERE(jump);
4900        }
4901    
4902      OP1(SLJIT_MOV, TMP1, 0, TMP3, 0);
4903      cc += 32 / sizeof(pcre_uchar);
4904      }
4905    
4906  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
4907  /* Simple register allocation. TMP1 is preferred if possible. */  /* Simple register allocation. TMP1 is preferred if possible. */
4908  if (needstype || needsscript)  if (needstype || needsscript)
# Line 3869  if (needstype || needsscript) Line 4944  if (needstype || needsscript)
4944  #endif  #endif
4945    
4946  /* Generating code. */  /* Generating code. */
 cc = ccbegin;  
4947  charoffset = 0;  charoffset = 0;
4948  numberofcmps = 0;  numberofcmps = 0;
4949  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 3885  while (*cc != XCL_END) Line 4959  while (*cc != XCL_END)
4959    if (*cc == XCL_SINGLE)    if (*cc == XCL_SINGLE)
4960      {      {
4961      cc ++;      cc ++;
4962  #ifdef SUPPORT_UTF      GETCHARINCTEST(c, cc);
     if (common->utf)  
       {  
       GETCHARINC(c, cc);  
       }  
     else  
 #endif  
       c = *cc++;  
4963    
4964      if (numberofcmps < 3 && (*cc == XCL_SINGLE || *cc == XCL_RANGE))      if (numberofcmps < 3 && (*cc == XCL_SINGLE || *cc == XCL_RANGE))
4965        {        {
4966        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));
4967        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);
4968        numberofcmps++;        numberofcmps++;
4969        }        }
4970      else if (numberofcmps > 0)      else if (numberofcmps > 0)
4971        {        {
4972        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));
4973        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);
4974        jump = JUMP(SLJIT_C_NOT_ZERO ^ invertcmp);        jump = JUMP(SLJIT_C_NOT_ZERO ^ invertcmp);
4975        numberofcmps = 0;        numberofcmps = 0;
4976        }        }
4977      else      else
4978        {        {
4979        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));
4980        numberofcmps = 0;        numberofcmps = 0;
4981        }        }
4982      }      }
4983    else if (*cc == XCL_RANGE)    else if (*cc == XCL_RANGE)
4984      {      {
4985      cc ++;      cc ++;
4986  #ifdef SUPPORT_UTF      GETCHARINCTEST(c, cc);
     if (common->utf)  
       {  
       GETCHARINC(c, cc);  
       }  
     else  
 #endif  
       c = *cc++;  
4987      SET_CHAR_OFFSET(c);      SET_CHAR_OFFSET(c);
4988  #ifdef SUPPORT_UTF      GETCHARINCTEST(c, cc);
4989      if (common->utf)  
       {  
       GETCHARINC(c, cc);  
       }  
     else  
 #endif  
       c = *cc++;  
4990      if (numberofcmps < 3 && (*cc == XCL_SINGLE || *cc == XCL_RANGE))      if (numberofcmps < 3 && (*cc == XCL_SINGLE || *cc == XCL_RANGE))
4991        {        {
4992        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));
4993        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);
4994        numberofcmps++;        numberofcmps++;
4995        }        }
4996      else if (numberofcmps > 0)      else if (numberofcmps > 0)
4997        {        {
4998        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));
4999        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);
5000        jump = JUMP(SLJIT_C_NOT_ZERO ^ invertcmp);        jump = JUMP(SLJIT_C_NOT_ZERO ^ invertcmp);
5001        numberofcmps = 0;        numberofcmps = 0;
5002        }        }
5003      else      else
5004        {        {
5005        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));
5006        numberofcmps = 0;        numberofcmps = 0;
5007        }        }
5008      }      }
# Line 3997  while (*cc != XCL_END) Line 5051  while (*cc != XCL_END)
5051    
5052        case PT_SPACE:        case PT_SPACE:
5053        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);  
         }  
5054        SET_CHAR_OFFSET(9);        SET_CHAR_OFFSET(9);
5055        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);
5056        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);
5057        if (*cc == PT_SPACE)  
5058          JUMPHERE(jump);        OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x85 - 0x9);
5059          OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);
5060    
5061          OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x180e - 0x9);
5062          OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);
5063    
5064        SET_TYPE_OFFSET(ucp_Zl);        SET_TYPE_OFFSET(ucp_Zl);
5065        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 4015  while (*cc != XCL_END) Line 5068  while (*cc != XCL_END)
5068        break;        break;
5069    
5070        case PT_WORD:        case PT_WORD:
5071        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));
5072        OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);        OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
5073        /* ... fall through */        /* Fall through. */
5074    
5075        case PT_ALNUM:        case PT_ALNUM:
5076        SET_TYPE_OFFSET(ucp_Ll);        SET_TYPE_OFFSET(ucp_Ll);
# Line 4063  while (*cc != XCL_END) Line 5116  while (*cc != XCL_END)
5116          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]);
5117          OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);          OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
5118    
5119          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));
5120          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);
5121    
5122          other_cases += 3;          other_cases += 3;
5123          }          }
5124        else        else
5125          {          {
5126          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));
5127          OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);          OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
5128          }          }
5129    
5130        while (*other_cases != NOTACHAR)        while (*other_cases != NOTACHAR)
5131          {          {
5132          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));
5133          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);
5134          }          }
5135        jump = JUMP(SLJIT_C_NOT_ZERO ^ invertcmp);        jump = JUMP(SLJIT_C_NOT_ZERO ^ invertcmp);
5136        break;        break;
5137    
5138          case PT_UCNC:
5139          OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(CHAR_DOLLAR_SIGN - charoffset));
5140          OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
5141          OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(CHAR_COMMERCIAL_AT - charoffset));
5142          OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);
5143          OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(CHAR_GRAVE_ACCENT - charoffset));
5144          OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);
5145    
5146          SET_CHAR_OFFSET(0xa0);
5147          OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, (sljit_sw)(0xd7ff - charoffset));
5148          OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_C_LESS_EQUAL);
5149          SET_CHAR_OFFSET(0);
5150          OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xe000 - 0);
5151          OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_C_GREATER_EQUAL);
5152          jump = JUMP(SLJIT_C_NOT_ZERO ^ invertcmp);
5153          break;
5154    
5155          case PT_PXGRAPH:
5156          /* C and Z groups are the farthest two groups. */
5157          SET_TYPE_OFFSET(ucp_Ll);
5158          OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, typereg, 0, SLJIT_IMM, ucp_So - ucp_Ll);
5159          OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_GREATER);
5160    
5161          jump = CMP(SLJIT_C_NOT_EQUAL, typereg, 0, SLJIT_IMM, ucp_Cf - ucp_Ll);
5162    
5163          /* In case of ucp_Cf, we overwrite the result. */
5164          SET_CHAR_OFFSET(0x2066);
5165          OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x2069 - 0x2066);
5166          OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_LESS_EQUAL);
5167    
5168          OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x061c - 0x2066);
5169          OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);
5170    
5171          OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x180e - 0x2066);
5172          OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);
5173