/[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 664 by ph10, Mon Aug 22 14:54:38 2011 UTC revision 1316 by zherczeg, Sun Apr 28 08:54:42 2013 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2008 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) 2010-2011                        Copyright (c) 2010-2013
13    
14  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
15  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 46  POSSIBILITY OF SUCH DAMAGE. Line 46  POSSIBILITY OF SUCH DAMAGE.
46    
47  #include "pcre_internal.h"  #include "pcre_internal.h"
48    
49  #ifdef SUPPORT_JIT  #if defined SUPPORT_JIT
50    
51  /* All-in-one: Since we use the JIT compiler only from here,  /* All-in-one: Since we use the JIT compiler only from here,
52  we just include it. This way we don't need to touch the build  we just include it. This way we don't need to touch the build
53  system files. */  system files. */
54    
55    #define SLJIT_MALLOC(size) (PUBL(malloc))(size)
56    #define SLJIT_FREE(ptr) (PUBL(free))(ptr)
57  #define SLJIT_CONFIG_AUTO 1  #define SLJIT_CONFIG_AUTO 1
58    #define SLJIT_CONFIG_STATIC 1
59  #define SLJIT_VERBOSE 0  #define SLJIT_VERBOSE 0
60  #define SLJIT_DEBUG 0  #define SLJIT_DEBUG 0
61    
62  #include "sljit/sljitLir.c"  #include "sljit/sljitLir.c"
63    
64  #if defined SLJIT_CONFIG_UNSUPPORTED && SLJIT_CONFIG_UNSUPPORTED  #if defined SLJIT_CONFIG_UNSUPPORTED && SLJIT_CONFIG_UNSUPPORTED
65  #error "Unsupported architecture"  #error Unsupported architecture
66  #endif  #endif
67    
68  /* Allocate memory on the stack. Fast, but limited size. */  /* Defines for debugging purposes. */
 #define LOCAL_SPACE_SIZE 32768  
69    
70    /* 1 - Use unoptimized capturing brackets.
71       2 - Enable capture_last_ptr (includes option 1). */
72    /* #define DEBUG_FORCE_UNOPTIMIZED_CBRAS 2 */
73    
74    /* 1 - Always have a control head. */
75    /* #define DEBUG_FORCE_CONTROL_HEAD 1 */
76    
77    /* Allocate memory for the regex stack on the real machine stack.
78    Fast, but limited size. */
79    #define MACHINE_STACK_SIZE 32768
80    
81    /* Growth rate for stack allocated by the OS. Should be the multiply
82    of page size. */
83  #define STACK_GROWTH_RATE 8192  #define STACK_GROWTH_RATE 8192
84    
85  /* Enable to check that the allocation could destroy temporaries. */  /* Enable to check that the allocation could destroy temporaries. */
# Line 79  The code generator follows the recursive Line 94  The code generator follows the recursive
94  expressions. The basic blocks of regular expressions are condition checkers  expressions. The basic blocks of regular expressions are condition checkers
95  whose execute different commands depending on the result of the condition check.  whose execute different commands depending on the result of the condition check.
96  The relationship between the operators can be horizontal (concatenation) and  The relationship between the operators can be horizontal (concatenation) and
97  vertical (sub-expression) (See struct fallback_common for more details).  vertical (sub-expression) (See struct backtrack_common for more details).
98    
99    'ab' - 'a' and 'b' regexps are concatenated    'ab' - 'a' and 'b' regexps are concatenated
100    'a+' - 'a' is the sub-expression of the '+' operator    'a+' - 'a' is the sub-expression of the '+' operator
101    
102  The condition checkers are boolean (true/false) checkers. Machine code is generated  The condition checkers are boolean (true/false) checkers. Machine code is generated
103  for the checker itself and for the actions depending on the result of the checker.  for the checker itself and for the actions depending on the result of the checker.
104  The 'true' case is called as the hot path (expected path), and the other is called as  The 'true' case is called as the matching path (expected path), and the other is called as
105  the 'fallback' path. Branch instructions are expesive for all CPUs, so we avoid taken  the 'backtrack' path. Branch instructions are expesive for all CPUs, so we avoid taken
106  branches on the hot path.  branches on the matching path.
107    
108   Greedy star operator (*) :   Greedy star operator (*) :
109     Hot path: match happens.     Matching path: match happens.
110     Fallback path: match failed.     Backtrack path: match failed.
111   Non-greedy star operator (*?) :   Non-greedy star operator (*?) :
112     Hot path: no need to perform a match.     Matching path: no need to perform a match.
113     Fallback path: match is required.     Backtrack path: match is required.
114    
115  The following example shows how the code generated for a capturing bracket  The following example shows how the code generated for a capturing bracket
116  with two alternatives. Let A, B, C, D are arbirary regular expressions, and  with two alternatives. Let A, B, C, D are arbirary regular expressions, and
# Line 105  we have the following regular expression Line 120  we have the following regular expression
120    
121  The generated code will be the following:  The generated code will be the following:
122    
123   A hot path   A matching path
124   '(' hot path (pushing arguments to the stack)   '(' matching path (pushing arguments to the stack)
125   B hot path   B matching path
126   ')' hot path (pushing arguments to the stack)   ')' matching path (pushing arguments to the stack)
127   D hot path   D matching path
128   return with successful match   return with successful match
129    
130   D fallback path   D backtrack path
131   ')' fallback path (If we arrived from "C" jump to the fallback of "C")   ')' backtrack path (If we arrived from "C" jump to the backtrack of "C")
132   B fallback path   B backtrack path
133   C expected path   C expected path
134   jump to D hot path   jump to D matching path
135   C fallback path   C backtrack path
136   A fallback path   A backtrack path
137    
138   Notice, that the order of fallback code paths are the opposite of the fast   Notice, that the order of backtrack code paths are the opposite of the fast
139   code paths. In this way the topmost value on the stack is always belong   code paths. In this way the topmost value on the stack is always belong
140   to the current fallback code path. The fallback code path must check   to the current backtrack code path. The backtrack path must check
141   whether there is a next alternative. If so, it needs to jump back to   whether there is a next alternative. If so, it needs to jump back to
142   the hot path eventually. Otherwise it needs to clear out its own stack   the matching path eventually. Otherwise it needs to clear out its own stack
143   frame and continue the execution on the fallback code paths.   frame and continue the execution on the backtrack code paths.
144  */  */
145    
146  /*  /*
147  Saved stack frames:  Saved stack frames:
148    
149  Atomic blocks and asserts require reloading the values of local variables  Atomic blocks and asserts require reloading the values of private data
150  when the fallback mechanism performed. Because of OP_RECURSE, the locals  when the backtrack mechanism performed. Because of OP_RECURSE, the data
151  are not necessarly known in compile time, thus we need a dynamic restore  are not necessarly known in compile time, thus we need a dynamic restore
152  mechanism.  mechanism.
153    
154  The stack frames are stored in a chain list, and have the following format:  The stack frames are stored in a chain list, and have the following format:
155  ([ capturing bracket offset ][ start value ][ end value ])+ ... [ 0 ] [ previous head ]  ([ capturing bracket offset ][ start value ][ end value ])+ ... [ 0 ] [ previous head ]
156    
157  Thus we can restore the locals to a particular point in the stack.  Thus we can restore the private data to a particular point in the stack.
158  */  */
159    
160  typedef struct jit_arguments {  typedef struct jit_arguments {
161    /* Pointers first. */    /* Pointers first. */
162    struct sljit_stack *stack;    struct sljit_stack *stack;
163    PCRE_SPTR str;    const pcre_uchar *str;
164    PCRE_SPTR begin;    const pcre_uchar *begin;
165    PCRE_SPTR end;    const pcre_uchar *end;
166    int *offsets;    int *offsets;
167    uschar *ptr;    pcre_uchar *uchar_ptr;
168      pcre_uchar *mark_ptr;
169      void *callout_data;
170    /* Everything else after. */    /* Everything else after. */
171    int offsetcount;    pcre_uint32 limit_match;
172    uschar notbol;    int real_offset_count;
173    uschar noteol;    int offset_count;
174    uschar notempty;    pcre_uint8 notbol;
175    uschar notempty_atstart;    pcre_uint8 noteol;
176      pcre_uint8 notempty;
177      pcre_uint8 notempty_atstart;
178  } jit_arguments;  } jit_arguments;
179    
180  typedef struct executable_function {  typedef struct executable_functions {
181    void *executable_func;    void *executable_funcs[JIT_NUMBER_OF_COMPILE_MODES];
182    pcre_jit_callback callback;    PUBL(jit_callback) callback;
183    void *userdata;    void *userdata;
184  } executable_function;    pcre_uint32 top_bracket;
185      pcre_uint32 limit_match;
186      sljit_uw executable_sizes[JIT_NUMBER_OF_COMPILE_MODES];
187    } executable_functions;
188    
189  typedef struct jump_list {  typedef struct jump_list {
190    struct sljit_jump *jump;    struct sljit_jump *jump;
191    struct jump_list *next;    struct jump_list *next;
192  } jump_list;  } jump_list;
193    
 enum stub_types { stack_alloc, max_index };  
   
194  typedef struct stub_list {  typedef struct stub_list {
   enum stub_types type;  
   int data;  
195    struct sljit_jump *start;    struct sljit_jump *start;
196    struct sljit_label *leave;    struct sljit_label *quit;
197    struct stub_list *next;    struct stub_list *next;
198  } stub_list;  } stub_list;
199    
200    enum frame_types {
201      no_frame = -1,
202      no_stack = -2
203    };
204    
205    enum control_types {
206      type_mark = 0,
207      type_then_trap = 1
208    };
209    
210  typedef int (SLJIT_CALL *jit_function)(jit_arguments *args);  typedef int (SLJIT_CALL *jit_function)(jit_arguments *args);
211    
212  /* The following structure is the key data type for the recursive  /* The following structure is the key data type for the recursive
213  code generator. It is allocated by compile_hotpath, and contains  code generator. It is allocated by compile_matchingpath, and contains
214  the aguments for compile_fallbackpath. Must be the first member  the aguments for compile_backtrackingpath. Must be the first member
215  of its descendants. */  of its descendants. */
216  typedef struct fallback_common {  typedef struct backtrack_common {
217    /* Concatenation stack. */    /* Concatenation stack. */
218    struct fallback_common *prev;    struct backtrack_common *prev;
219    jump_list *nextfallbacks;    jump_list *nextbacktracks;
220    /* Internal stack (for component operators). */    /* Internal stack (for component operators). */
221    struct fallback_common *top;    struct backtrack_common *top;
222    jump_list *topfallbacks;    jump_list *topbacktracks;
223    /* Opcode pointer. */    /* Opcode pointer. */
224    uschar *cc;    pcre_uchar *cc;
225  } fallback_common;  } backtrack_common;
226    
227  typedef struct assert_fallback {  typedef struct assert_backtrack {
228    fallback_common common;    backtrack_common common;
229    jump_list *condfailed;    jump_list *condfailed;
230    /* Less than 0 (-1) if a frame is not needed. */    /* Less than 0 if a frame is not needed. */
231    int framesize;    int framesize;
232    /* Points to our private memory word on the stack. */    /* Points to our private memory word on the stack. */
233    int localptr;    int private_data_ptr;
234    /* For iterators. */    /* For iterators. */
235    struct sljit_label *hotpath;    struct sljit_label *matchingpath;
236  } assert_fallback;  } assert_backtrack;
237    
238  typedef struct bracket_fallback {  typedef struct bracket_backtrack {
239    fallback_common common;    backtrack_common common;
240    /* Where to coninue if an alternative is successfully matched. */    /* Where to coninue if an alternative is successfully matched. */
241    struct sljit_label *althotpath;    struct sljit_label *alternative_matchingpath;
242    /* For rmin and rmax iterators. */    /* For rmin and rmax iterators. */
243    struct sljit_label *recursivehotpath;    struct sljit_label *recursive_matchingpath;
244    /* For greedy ? operator. */    /* For greedy ? operator. */
245    struct sljit_label *zerohotpath;    struct sljit_label *zero_matchingpath;
246    /* Contains the branches of a failed condition. */    /* Contains the branches of a failed condition. */
247    union {    union {
248      /* Both for OP_COND, OP_SCOND. */      /* Both for OP_COND, OP_SCOND. */
249      jump_list *condfailed;      jump_list *condfailed;
250      assert_fallback *assert;      assert_backtrack *assert;
251      /* For OP_ONCE. -1 if not needed. */      /* For OP_ONCE. Less than 0 if not needed. */
252      int framesize;      int framesize;
253    } u;    } u;
254    /* Points to our private memory word on the stack. */    /* Points to our private memory word on the stack. */
255    int localptr;    int private_data_ptr;
256  } bracket_fallback;  } bracket_backtrack;
257    
258  typedef struct bracketpos_fallback {  typedef struct bracketpos_backtrack {
259    fallback_common common;    backtrack_common common;
260    /* Points to our private memory word on the stack. */    /* Points to our private memory word on the stack. */
261    int localptr;    int private_data_ptr;
262    /* Reverting stack is needed. */    /* Reverting stack is needed. */
263    int framesize;    int framesize;
264    /* Allocated stack size. */    /* Allocated stack size. */
265    int stacksize;    int stacksize;
266  } bracketpos_fallback;  } bracketpos_backtrack;
267    
268  typedef struct braminzero_fallback {  typedef struct braminzero_backtrack {
269    fallback_common common;    backtrack_common common;
270    struct sljit_label *hotpath;    struct sljit_label *matchingpath;
271  } braminzero_fallback;  } braminzero_backtrack;
272    
273  typedef struct iterator_fallback {  typedef struct iterator_backtrack {
274    fallback_common common;    backtrack_common common;
275    /* Next iteration. */    /* Next iteration. */
276    struct sljit_label *hotpath;    struct sljit_label *matchingpath;
277  } iterator_fallback;  } iterator_backtrack;
278    
279  typedef struct recurse_entry {  typedef struct recurse_entry {
280    struct recurse_entry *next;    struct recurse_entry *next;
# Line 255  typedef struct recurse_entry { Line 283  typedef struct recurse_entry {
283    /* Collects the calls until the function is not created. */    /* Collects the calls until the function is not created. */
284    jump_list *calls;    jump_list *calls;
285    /* Points to the starting opcode. */    /* Points to the starting opcode. */
286    int start;    sljit_sw start;
287  } recurse_entry;  } recurse_entry;
288    
289  typedef struct recurse_fallback {  typedef struct recurse_backtrack {
290    fallback_common common;    backtrack_common common;
291  } recurse_fallback;    BOOL inlined_pattern;
292    } recurse_backtrack;
293    
294    #define OP_THEN_TRAP OP_TABLE_LENGTH
295    
296    typedef struct then_trap_backtrack {
297      backtrack_common common;
298      /* If then_trap is not NULL, this structure contains the real
299      then_trap for the backtracking path. */
300      struct then_trap_backtrack *then_trap;
301      /* Points to the starting opcode. */
302      sljit_sw start;
303      /* Exit point for the then opcodes of this alternative. */
304      jump_list *quit;
305      /* Frame size of the current alternative. */
306      int framesize;
307    } then_trap_backtrack;
308    
309    #define MAX_RANGE_SIZE 6
310    
311  typedef struct compiler_common {  typedef struct compiler_common {
312      /* The sljit ceneric compiler. */
313    struct sljit_compiler *compiler;    struct sljit_compiler *compiler;
314    uschar *start;    /* First byte code. */
315    int localsize;    pcre_uchar *start;
316    int *localptrs;    /* Maps private data offset to each opcode. */
317    const uschar *fcc;    sljit_si *private_data_ptrs;
318    sljit_w lcc;    /* Tells whether the capturing bracket is optimized. */
319    int cbraptr;    pcre_uint8 *optimized_cbracket;
320      /* Tells whether the starting offset is a target of then. */
321      pcre_uint8 *then_offsets;
322      /* Current position where a THEN must jump. */
323      then_trap_backtrack *then_trap;
324      /* Starting offset of private data for capturing brackets. */
325      int cbra_ptr;
326      /* Output vector starting point. Must be divisible by 2. */
327      int ovector_start;
328      /* Last known position of the requested byte. */
329      int req_char_ptr;
330      /* Head of the last recursion. */
331      int recursive_head_ptr;
332      /* First inspected character for partial matching. */
333      int start_used_ptr;
334      /* Starting pointer for partial soft matches. */
335      int hit_start;
336      /* End pointer of the first line. */
337      int first_line_end;
338      /* Points to the marked string. */
339      int mark_ptr;
340      /* Recursive control verb management chain. */
341      int control_head_ptr;
342      /* Points to the last matched capture block index. */
343      int capture_last_ptr;
344      /* Points to the starting position of the current match. */
345      int start_ptr;
346    
347      /* Flipped and lower case tables. */
348      const pcre_uint8 *fcc;
349      sljit_sw lcc;
350      /* Mode can be PCRE_STUDY_JIT_COMPILE and others. */
351      int mode;
352      /* \K is found in the pattern. */
353      BOOL has_set_som;
354      /* (*SKIP:arg) is found in the pattern. */
355      BOOL has_skip_arg;
356      /* (*THEN) is found in the pattern. */
357      BOOL has_then;
358      /* Needs to know the start position anytime. */
359      BOOL needs_start_ptr;
360      /* Currently in recurse or negative assert. */
361      BOOL local_exit;
362      /* Currently in a positive assert. */
363      BOOL positive_assert;
364      /* Newline control. */
365    int nltype;    int nltype;
366    int newline;    int newline;
367    int bsr_nltype;    int bsr_nltype;
368      /* Dollar endonly. */
369    int endonly;    int endonly;
370    sljit_w ctypes;    /* Tables. */
371    struct sljit_label *acceptlabel;    sljit_sw ctypes;
372      int digits[2 + MAX_RANGE_SIZE];
373      /* Named capturing brackets. */
374      sljit_uw name_table;
375      sljit_sw name_count;
376      sljit_sw name_entry_size;
377    
378      /* Labels and jump lists. */
379      struct sljit_label *partialmatchlabel;
380      struct sljit_label *quit_label;
381      struct sljit_label *forced_quit_label;
382      struct sljit_label *accept_label;
383    stub_list *stubs;    stub_list *stubs;
384    recurse_entry *entries;    recurse_entry *entries;
385    recurse_entry *currententry;    recurse_entry *currententry;
386      jump_list *partialmatch;
387      jump_list *quit;
388      jump_list *positive_assert_quit;
389      jump_list *forced_quit;
390    jump_list *accept;    jump_list *accept;
391      jump_list *calllimit;
392    jump_list *stackalloc;    jump_list *stackalloc;
393    jump_list *revertframes;    jump_list *revertframes;
394    jump_list *wordboundary;    jump_list *wordboundary;
# Line 288  typedef struct compiler_common { Line 397  typedef struct compiler_common {
397    jump_list *vspace;    jump_list *vspace;
398    jump_list *casefulcmp;    jump_list *casefulcmp;
399    jump_list *caselesscmp;    jump_list *caselesscmp;
400      jump_list *reset_match;
401    BOOL jscript_compat;    BOOL jscript_compat;
402  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
403    BOOL utf8;    BOOL utf;
404  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
405    BOOL useucp;    BOOL use_ucp;
406    #endif
407    #ifndef COMPILE_PCRE32
408      jump_list *utfreadchar;
409  #endif  #endif
410    jump_list *utf8readchar;  #ifdef COMPILE_PCRE8
411    jump_list *utf8readtype8;    jump_list *utfreadtype8;
412  #endif  #endif
413    #endif /* SUPPORT_UTF */
414  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
415    jump_list *getucd;    jump_list *getucd;
416  #endif  #endif
# Line 308  typedef struct compare_context { Line 422  typedef struct compare_context {
422    int length;    int length;
423    int sourcereg;    int sourcereg;
424  #if defined SLJIT_UNALIGNED && SLJIT_UNALIGNED  #if defined SLJIT_UNALIGNED && SLJIT_UNALIGNED
425    int byteptr;    int ucharptr;
426    union {    union {
427      int asint;      sljit_si asint;
428      short asshort;      sljit_uh asushort;
429    #if defined COMPILE_PCRE8
430      sljit_ub asbyte;      sljit_ub asbyte;
431      sljit_ub asbytes[4];      sljit_ub asuchars[4];
432    #elif defined COMPILE_PCRE16
433        sljit_uh asuchars[2];
434    #elif defined COMPILE_PCRE32
435        sljit_ui asuchars[1];
436    #endif
437    } c;    } c;
438    union {    union {
439      int asint;      sljit_si asint;
440      short asshort;      sljit_uh asushort;
441    #if defined COMPILE_PCRE8
442      sljit_ub asbyte;      sljit_ub asbyte;
443      sljit_ub asbytes[4];      sljit_ub asuchars[4];
444    #elif defined COMPILE_PCRE16
445        sljit_uh asuchars[2];
446    #elif defined COMPILE_PCRE32
447        sljit_ui asuchars[1];
448    #endif
449    } oc;    } oc;
450  #endif  #endif
451  } compare_context;  } compare_context;
452    
453  enum {  /* Undefine sljit macros. */
454    frame_end = 0,  #undef CMP
   frame_setmaxindex = -1,  
   frame_setstrbegin = -2  
 };  
455    
456  /* Used for accessing the elements of the stack. */  /* Used for accessing the elements of the stack. */
457  #define STACK(i)      ((-(i) - 1) * (int)sizeof(sljit_w))  #define STACK(i)      ((-(i) - 1) * (int)sizeof(sljit_sw))
458    
459  #define TMP1          SLJIT_TEMPORARY_REG1  #define TMP1          SLJIT_SCRATCH_REG1
460  #define TMP2          SLJIT_TEMPORARY_REG3  #define TMP2          SLJIT_SCRATCH_REG3
461  #define TMP3          SLJIT_TEMPORARY_EREG2  #define TMP3          SLJIT_TEMPORARY_EREG2
462  #define STR_PTR       SLJIT_GENERAL_REG1  #define STR_PTR       SLJIT_SAVED_REG1
463  #define STR_END       SLJIT_GENERAL_REG2  #define STR_END       SLJIT_SAVED_REG2
464  #define STACK_TOP     SLJIT_TEMPORARY_REG2  #define STACK_TOP     SLJIT_SCRATCH_REG2
465  #define STACK_LIMIT   SLJIT_GENERAL_REG3  #define STACK_LIMIT   SLJIT_SAVED_REG3
466  #define ARGUMENTS     SLJIT_GENERAL_EREG1  #define ARGUMENTS     SLJIT_SAVED_EREG1
467  #define MAX_INDEX     SLJIT_GENERAL_EREG2  #define COUNT_MATCH   SLJIT_SAVED_EREG2
468  #define RETURN_ADDR   SLJIT_TEMPORARY_EREG1  #define RETURN_ADDR   SLJIT_TEMPORARY_EREG1
469    
470  /* Locals layout. */  /* Local space layout. */
471  /* These two locals can be used by the current opcode. */  /* These two locals can be used by the current opcode. */
472  #define LOCALS0          (0 * sizeof(sljit_w))  #define LOCALS0          (0 * sizeof(sljit_sw))
473  #define LOCALS1          (1 * sizeof(sljit_w))  #define LOCALS1          (1 * sizeof(sljit_sw))
474  /* Two local variables for possessive quantifiers (char1 cannot use them). */  /* Two local variables for possessive quantifiers (char1 cannot use them). */
475  #define POSSESSIVE0      (2 * sizeof(sljit_w))  #define POSSESSIVE0      (2 * sizeof(sljit_sw))
476  #define POSSESSIVE1      (3 * sizeof(sljit_w))  #define POSSESSIVE1      (3 * sizeof(sljit_sw))
477  /* Head of the saved local variables */  /* Max limit of recursions. */
478  #define LOCALS_HEAD      (4 * sizeof(sljit_w))  #define LIMIT_MATCH      (4 * sizeof(sljit_sw))
 /* Head of the last recursion. */  
 #define RECURSIVE_HEAD   (5 * sizeof(sljit_w))  
 /* Last known position of the requested byte. */  
 #define REQ_BYTE_PTR     (6 * sizeof(sljit_w))  
 /* End pointer of the first line. */  
 #define FIRSTLINE_END    (7 * sizeof(sljit_w))  
479  /* The output vector is stored on the stack, and contains pointers  /* The output vector is stored on the stack, and contains pointers
480  to characters. The vector data is divided into two groups: the first  to characters. The vector data is divided into two groups: the first
481  group contains the start / end character pointers, and the second is  group contains the start / end character pointers, and the second is
482  the start pointers when the end of the capturing group has not yet reached. */  the start pointers when the end of the capturing group has not yet reached. */
483  #define OVECTOR_START    (8 * sizeof(sljit_w))  #define OVECTOR_START    (common->ovector_start)
484  #define OVECTOR(i)       (OVECTOR_START + (i) * sizeof(sljit_w))  #define OVECTOR(i)       (OVECTOR_START + (i) * sizeof(sljit_sw))
485  #define OVECTOR_PRIV(i)  (common->cbraptr + (i) * sizeof(sljit_w))  #define OVECTOR_PRIV(i)  (common->cbra_ptr + (i) * sizeof(sljit_sw))
486  #define PRIV(cc)         (common->localptrs[(cc) - common->start])  #define PRIVATE_DATA(cc) (common->private_data_ptrs[(cc) - common->start])
487    
488    #if defined COMPILE_PCRE8
489    #define MOV_UCHAR  SLJIT_MOV_UB
490    #define MOVU_UCHAR SLJIT_MOVU_UB
491    #elif defined COMPILE_PCRE16
492    #define MOV_UCHAR  SLJIT_MOV_UH
493    #define MOVU_UCHAR SLJIT_MOVU_UH
494    #elif defined COMPILE_PCRE32
495    #define MOV_UCHAR  SLJIT_MOV_UI
496    #define MOVU_UCHAR SLJIT_MOVU_UI
497    #else
498    #error Unsupported compiling mode
499    #endif
500    
501  /* Shortcuts. */  /* Shortcuts. */
502  #define DEFINE_COMPILER \  #define DEFINE_COMPILER \
# Line 383  the start pointers when the end of the c Line 513  the start pointers when the end of the c
513    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))
514  #define JUMPHERE(jump) \  #define JUMPHERE(jump) \
515    sljit_set_label((jump), sljit_emit_label(compiler))    sljit_set_label((jump), sljit_emit_label(compiler))
516    #define SET_LABEL(jump, label) \
517      sljit_set_label((jump), (label))
518  #define CMP(type, src1, src1w, src2, src2w) \  #define CMP(type, src1, src1w, src2, src2w) \
519    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))
520  #define CMPTO(type, src1, src1w, src2, src2w, label) \  #define CMPTO(type, src1, src1w, src2, src2w, label) \
521    sljit_set_label(sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w)), (label))    sljit_set_label(sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w)), (label))
522  #define COND_VALUE(op, dst, dstw, type) \  #define OP_FLAGS(op, dst, dstw, src, srcw, type) \
523    sljit_emit_cond_value(compiler, (op), (dst), (dstw), (type))    sljit_emit_op_flags(compiler, (op), (dst), (dstw), (src), (srcw), (type))
524    #define GET_LOCAL_BASE(dst, dstw, offset) \
525      sljit_get_local_base(compiler, (dst), (dstw), (offset))
526    
527  static uschar* bracketend(uschar* cc)  static pcre_uchar* bracketend(pcre_uchar* cc)
528  {  {
529  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));
530  do cc += GET(cc, 1); while (*cc == OP_ALT);  do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 399  cc += 1 + LINK_SIZE; Line 533  cc += 1 + LINK_SIZE;
533  return cc;  return cc;
534  }  }
535    
536  /* Functions whose might need modification for all new supported opcodes:  /* Functions whose might need modification for all new supported opcodes:
537   next_opcode   next_opcode
538   get_localspace   check_opcode_types
539   set_localptrs   set_private_data_ptrs
540   get_framesize   get_framesize
541   init_frame   init_frame
542   get_localsize   get_private_data_copy_length
543   copy_locals   copy_private_data
544   compile_hotpath   compile_matchingpath
545   compile_fallbackpath   compile_backtrackingpath
546  */  */
547    
548  static uschar *next_opcode(compiler_common *common, uschar *cc)  static pcre_uchar *next_opcode(compiler_common *common, pcre_uchar *cc)
549  {  {
550  SLJIT_UNUSED_ARG(common);  SLJIT_UNUSED_ARG(common);
551  switch(*cc)  switch(*cc)
# Line 429  switch(*cc) Line 563  switch(*cc)
563    case OP_WORDCHAR:    case OP_WORDCHAR:
564    case OP_ANY:    case OP_ANY:
565    case OP_ALLANY:    case OP_ALLANY:
566      case OP_NOTPROP:
567      case OP_PROP:
568    case OP_ANYNL:    case OP_ANYNL:
569    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
570    case OP_HSPACE:    case OP_HSPACE:
# Line 441  switch(*cc) Line 577  switch(*cc)
577    case OP_CIRCM:    case OP_CIRCM:
578    case OP_DOLL:    case OP_DOLL:
579    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:  
580    case OP_CRSTAR:    case OP_CRSTAR:
581    case OP_CRMINSTAR:    case OP_CRMINSTAR:
582    case OP_CRPLUS:    case OP_CRPLUS:
583    case OP_CRMINPLUS:    case OP_CRMINPLUS:
584    case OP_CRQUERY:    case OP_CRQUERY:
585    case OP_CRMINQUERY:    case OP_CRMINQUERY:
586      case OP_CRRANGE:
587      case OP_CRMINRANGE:
588      case OP_CLASS:
589      case OP_NCLASS:
590      case OP_REF:
591      case OP_REFI:
592      case OP_RECURSE:
593      case OP_CALLOUT:
594      case OP_ALT:
595      case OP_KET:
596      case OP_KETRMAX:
597      case OP_KETRMIN:
598      case OP_KETRPOS:
599      case OP_REVERSE:
600      case OP_ASSERT:
601      case OP_ASSERT_NOT:
602      case OP_ASSERTBACK:
603      case OP_ASSERTBACK_NOT:
604      case OP_ONCE:
605      case OP_ONCE_NC:
606      case OP_BRA:
607      case OP_BRAPOS:
608      case OP_CBRA:
609      case OP_CBRAPOS:
610      case OP_COND:
611      case OP_SBRA:
612      case OP_SBRAPOS:
613      case OP_SCBRA:
614      case OP_SCBRAPOS:
615      case OP_SCOND:
616      case OP_CREF:
617      case OP_NCREF:
618      case OP_RREF:
619      case OP_NRREF:
620    case OP_DEF:    case OP_DEF:
621    case OP_BRAZERO:    case OP_BRAZERO:
622    case OP_BRAMINZERO:    case OP_BRAMINZERO:
623    case OP_BRAPOSZERO:    case OP_BRAPOSZERO:
624      case OP_PRUNE:
625      case OP_SKIP:
626      case OP_THEN:
627      case OP_COMMIT:
628    case OP_FAIL:    case OP_FAIL:
629    case OP_ACCEPT:    case OP_ACCEPT:
630    case OP_ASSERT_ACCEPT:    case OP_ASSERT_ACCEPT:
631      case OP_CLOSE:
632    case OP_SKIPZERO:    case OP_SKIPZERO:
633    return cc + 1;    return cc + PRIV(OP_lengths)[*cc];
634    
635    case OP_CHAR:    case OP_CHAR:
636    case OP_CHARI:    case OP_CHARI:
637    case OP_NOT:    case OP_NOT:
638    case OP_NOTI:    case OP_NOTI:
   
639    case OP_STAR:    case OP_STAR:
640    case OP_MINSTAR:    case OP_MINSTAR:
641    case OP_PLUS:    case OP_PLUS:
642    case OP_MINPLUS:    case OP_MINPLUS:
643    case OP_QUERY:    case OP_QUERY:
644    case OP_MINQUERY:    case OP_MINQUERY:
645      case OP_UPTO:
646      case OP_MINUPTO:
647      case OP_EXACT:
648    case OP_POSSTAR:    case OP_POSSTAR:
649    case OP_POSPLUS:    case OP_POSPLUS:
650    case OP_POSQUERY:    case OP_POSQUERY:
651      case OP_POSUPTO:
652    case OP_STARI:    case OP_STARI:
653    case OP_MINSTARI:    case OP_MINSTARI:
654    case OP_PLUSI:    case OP_PLUSI:
655    case OP_MINPLUSI:    case OP_MINPLUSI:
656    case OP_QUERYI:    case OP_QUERYI:
657    case OP_MINQUERYI:    case OP_MINQUERYI:
658      case OP_UPTOI:
659      case OP_MINUPTOI:
660      case OP_EXACTI:
661    case OP_POSSTARI:    case OP_POSSTARI:
662    case OP_POSPLUSI:    case OP_POSPLUSI:
663    case OP_POSQUERYI:    case OP_POSQUERYI:
664      case OP_POSUPTOI:
665    case OP_NOTSTAR:    case OP_NOTSTAR:
666    case OP_NOTMINSTAR:    case OP_NOTMINSTAR:
667    case OP_NOTPLUS:    case OP_NOTPLUS:
668    case OP_NOTMINPLUS:    case OP_NOTMINPLUS:
669    case OP_NOTQUERY:    case OP_NOTQUERY:
670    case OP_NOTMINQUERY:    case OP_NOTMINQUERY:
671      case OP_NOTUPTO:
672      case OP_NOTMINUPTO:
673      case OP_NOTEXACT:
674    case OP_NOTPOSSTAR:    case OP_NOTPOSSTAR:
675    case OP_NOTPOSPLUS:    case OP_NOTPOSPLUS:
676    case OP_NOTPOSQUERY:    case OP_NOTPOSQUERY:
677      case OP_NOTPOSUPTO:
678    case OP_NOTSTARI:    case OP_NOTSTARI:
679    case OP_NOTMINSTARI:    case OP_NOTMINSTARI:
680    case OP_NOTPLUSI:    case OP_NOTPLUSI:
681    case OP_NOTMINPLUSI:    case OP_NOTMINPLUSI:
682    case OP_NOTQUERYI:    case OP_NOTQUERYI:
683    case OP_NOTMINQUERYI:    case OP_NOTMINQUERYI:
   case OP_NOTPOSSTARI:  
   case OP_NOTPOSPLUSI:  
   case OP_NOTPOSQUERYI:  
   cc += 2;  
 #ifdef SUPPORT_UTF8  
   if (common->utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
 #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:  
684    case OP_NOTUPTOI:    case OP_NOTUPTOI:
685    case OP_NOTMINUPTOI:    case OP_NOTMINUPTOI:
686    case OP_NOTEXACTI:    case OP_NOTEXACTI:
687      case OP_NOTPOSSTARI:
688      case OP_NOTPOSPLUSI:
689      case OP_NOTPOSQUERYI:
690    case OP_NOTPOSUPTOI:    case OP_NOTPOSUPTOI:
691    cc += 4;    cc += PRIV(OP_lengths)[*cc];
692  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
693    if (common->utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];    if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
694  #endif  #endif
695    return cc;    return cc;
696    
697    case OP_NOTPROP:    /* Special cases. */
698    case OP_PROP:    case OP_TYPESTAR:
699      case OP_TYPEMINSTAR:
700      case OP_TYPEPLUS:
701      case OP_TYPEMINPLUS:
702      case OP_TYPEQUERY:
703      case OP_TYPEMINQUERY:
704    case OP_TYPEUPTO:    case OP_TYPEUPTO:
705    case OP_TYPEMINUPTO:    case OP_TYPEMINUPTO:
706    case OP_TYPEEXACT:    case OP_TYPEEXACT:
707      case OP_TYPEPOSSTAR:
708      case OP_TYPEPOSPLUS:
709      case OP_TYPEPOSQUERY:
710    case OP_TYPEPOSUPTO:    case OP_TYPEPOSUPTO:
711    case OP_REF:    return cc + PRIV(OP_lengths)[*cc] - 1;
   case OP_REFI:  
   case OP_CREF:  
   case OP_CLOSE:  
   cc += 3;  
   return cc;  
712    
713    case OP_CRRANGE:    case OP_ANYBYTE:
714    case OP_CRMINRANGE:  #ifdef SUPPORT_UTF
715    return cc + 5;    if (common->utf) return NULL;
716    #endif
717    case OP_CLASS:    return cc + 1;
   case OP_NCLASS:  
   return cc + 33;  
718    
719  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
720    case OP_XCLASS:    case OP_XCLASS:
721    return cc + GET(cc, 1);    return cc + GET(cc, 1);
722  #endif  #endif
723    
724    case OP_RECURSE:    case OP_MARK:
725    case OP_ASSERT:    case OP_PRUNE_ARG:
726    case OP_ASSERT_NOT:    case OP_SKIP_ARG:
727    case OP_ASSERTBACK:    case OP_THEN_ARG:
728    case OP_ASSERTBACK_NOT:    return cc + 1 + 2 + cc[1];
   case OP_REVERSE:  
   case OP_ONCE:  
   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 + 2;  
729    
730    default:    default:
731      /* All opcodes are supported now! */
732      SLJIT_ASSERT_STOP();
733    return NULL;    return NULL;
734    }    }
735  }  }
736    
737  static int get_localspace(compiler_common *common, uschar *cc, uschar *ccend)  static BOOL check_opcode_types(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend)
738  {  {
739  int localspace = 0;  pcre_uchar *name;
740  uschar *alternative;  pcre_uchar *name2;
741    unsigned int cbra_index;
742    int i;
743    
744  /* Calculate important variables (like stack size) and checks whether all opcodes are supported. */  /* Calculate important variables (like stack size) and checks whether all opcodes are supported. */
745  while (cc < ccend)  while (cc < ccend)
746    {    {
747    switch(*cc)    switch(*cc)
748      {      {
749      case OP_ASSERT:      case OP_SET_SOM:
750      case OP_ASSERT_NOT:      common->has_set_som = TRUE;
751      case OP_ASSERTBACK:      cc += 1;
752      case OP_ASSERTBACK_NOT:      break;
753      case OP_ONCE:  
754      case OP_BRAPOS:      case OP_REF:
755      case OP_SBRA:      case OP_REFI:
756      case OP_SBRAPOS:      common->optimized_cbracket[GET2(cc, 1)] = 0;
757      case OP_SCOND:      cc += 1 + IMM2_SIZE;
     localspace += sizeof(sljit_w);  
     cc += 1 + LINK_SIZE;  
758      break;      break;
759    
760      case OP_CBRAPOS:      case OP_CBRAPOS:
761      case OP_SCBRAPOS:      case OP_SCBRAPOS:
762      localspace += sizeof(sljit_w);      common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] = 0;
763      cc += 1 + LINK_SIZE + 2;      cc += 1 + LINK_SIZE + IMM2_SIZE;
764      break;      break;
765    
766      case OP_COND:      case OP_COND:
767      /* Might be a hidden SCOND. */      case OP_SCOND:
768      alternative = cc + GET(cc, 1);      /* Only AUTO_CALLOUT can insert this opcode. We do
769      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)         not intend to support this case. */
770        localspace += sizeof(sljit_w);      if (cc[1 + LINK_SIZE] == OP_CALLOUT)
771          return FALSE;
772      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
773      break;      break;
774    
775        case OP_CREF:
776        i = GET2(cc, 1);
777        common->optimized_cbracket[i] = 0;
778        cc += 1 + IMM2_SIZE;
779        break;
780    
781        case OP_NCREF:
782        cbra_index = GET2(cc, 1);
783        name = (pcre_uchar *)common->name_table;
784        name2 = name;
785        for (i = 0; i < common->name_count; i++)
786          {
787          if (GET2(name, 0) == cbra_index) break;
788          name += common->name_entry_size;
789          }
790        SLJIT_ASSERT(i != common->name_count);
791    
792        for (i = 0; i < common->name_count; i++)
793          {
794          if (STRCMP_UC_UC(name2 + IMM2_SIZE, name + IMM2_SIZE) == 0)
795            common->optimized_cbracket[GET2(name2, 0)] = 0;
796          name2 += common->name_entry_size;
797          }
798        cc += 1 + IMM2_SIZE;
799        break;
800    
801        case OP_RECURSE:
802        /* Set its value only once. */
803        if (common->recursive_head_ptr == 0)
804          {
805          common->recursive_head_ptr = common->ovector_start;
806          common->ovector_start += sizeof(sljit_sw);
807          }
808        cc += 1 + LINK_SIZE;
809        break;
810    
811        case OP_CALLOUT:
812        if (common->capture_last_ptr == 0)
813          {
814          common->capture_last_ptr = common->ovector_start;
815          common->ovector_start += sizeof(sljit_sw);
816          }
817        cc += 2 + 2 * LINK_SIZE;
818        break;
819    
820        case OP_THEN_ARG:
821        common->has_then = TRUE;
822        common->control_head_ptr = 1;
823        /* Fall through. */
824    
825        case OP_PRUNE_ARG:
826        common->needs_start_ptr = TRUE;
827        /* Fall through. */
828    
829        case OP_MARK:
830        if (common->mark_ptr == 0)
831          {
832          common->mark_ptr = common->ovector_start;
833          common->ovector_start += sizeof(sljit_sw);
834          }
835        cc += 1 + 2 + cc[1];
836        break;
837    
838        case OP_THEN:
839        common->has_then = TRUE;
840        common->control_head_ptr = 1;
841        /* Fall through. */
842    
843        case OP_PRUNE:
844        case OP_SKIP:
845        common->needs_start_ptr = TRUE;
846        cc += 1;
847        break;
848    
849        case OP_SKIP_ARG:
850        common->control_head_ptr = 1;
851        common->has_skip_arg = TRUE;
852        cc += 1 + 2 + cc[1];
853        break;
854    
855      default:      default:
856      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
857      if (cc == NULL)      if (cc == NULL)
858        return -1;        return FALSE;
859      break;      break;
860      }      }
861    }    }
862  return localspace;  return TRUE;
863  }  }
864    
865  static void set_localptrs(compiler_common *common, int localptr, uschar *ccend)  static int get_class_iterator_size(pcre_uchar *cc)
866    {
867    switch(*cc)
868      {
869      case OP_CRSTAR:
870      case OP_CRPLUS:
871      return 2;
872    
873      case OP_CRMINSTAR:
874      case OP_CRMINPLUS:
875      case OP_CRQUERY:
876      case OP_CRMINQUERY:
877      return 1;
878    
879      case OP_CRRANGE:
880      case OP_CRMINRANGE:
881      if (GET2(cc, 1) == GET2(cc, 1 + IMM2_SIZE))
882        return 0;
883      return 2;
884    
885      default:
886      return 0;
887      }
888    }
889    
890    static BOOL detect_repeat(compiler_common *common, pcre_uchar *begin)
891    {
892    pcre_uchar *end = bracketend(begin);
893    pcre_uchar *next;
894    pcre_uchar *next_end;
895    pcre_uchar *max_end;
896    pcre_uchar type;
897    sljit_sw length = end - begin;
898    int min, max, i;
899    
900    /* Detect fixed iterations first. */
901    if (end[-(1 + LINK_SIZE)] != OP_KET)
902      return FALSE;
903    
904    /* Already detected repeat. */
905    if (common->private_data_ptrs[end - common->start - LINK_SIZE] != 0)
906      return TRUE;
907    
908    next = end;
909    min = 1;
910    while (1)
911      {
912      if (*next != *begin)
913        break;
914      next_end = bracketend(next);
915      if (next_end - next != length || memcmp(begin, next, IN_UCHARS(length)) != 0)
916        break;
917      next = next_end;
918      min++;
919      }
920    
921    if (min == 2)
922      return FALSE;
923    
924    max = 0;
925    max_end = next;
926    if (*next == OP_BRAZERO || *next == OP_BRAMINZERO)
927      {
928      type = *next;
929      while (1)
930        {
931        if (next[0] != type || next[1] != OP_BRA || next[2 + LINK_SIZE] != *begin)
932          break;
933        next_end = bracketend(next + 2 + LINK_SIZE);
934        if (next_end - next != (length + 2 + LINK_SIZE) || memcmp(begin, next + 2 + LINK_SIZE, IN_UCHARS(length)) != 0)
935          break;
936        next = next_end;
937        max++;
938        }
939    
940      if (next[0] == type && next[1] == *begin && max >= 1)
941        {
942        next_end = bracketend(next + 1);
943        if (next_end - next == (length + 1) && memcmp(begin, next + 1, IN_UCHARS(length)) == 0)
944          {
945          for (i = 0; i < max; i++, next_end += 1 + LINK_SIZE)
946            if (*next_end != OP_KET)
947              break;
948    
949          if (i == max)
950            {
951            common->private_data_ptrs[max_end - common->start - LINK_SIZE] = next_end - max_end;
952            common->private_data_ptrs[max_end - common->start - LINK_SIZE + 1] = (type == OP_BRAZERO) ? OP_UPTO : OP_MINUPTO;
953            /* +2 the original and the last. */
954            common->private_data_ptrs[max_end - common->start - LINK_SIZE + 2] = max + 2;
955            if (min == 1)
956              return TRUE;
957            min--;
958            max_end -= (1 + LINK_SIZE) + GET(max_end, -LINK_SIZE);
959            }
960          }
961        }
962      }
963    
964    if (min >= 3)
965      {
966      common->private_data_ptrs[end - common->start - LINK_SIZE] = max_end - end;
967      common->private_data_ptrs[end - common->start - LINK_SIZE + 1] = OP_EXACT;
968      common->private_data_ptrs[end - common->start - LINK_SIZE + 2] = min;
969      return TRUE;
970      }
971    
972    return FALSE;
973    }
974    
975    #define CASE_ITERATOR_PRIVATE_DATA_1 \
976        case OP_MINSTAR: \
977        case OP_MINPLUS: \
978        case OP_QUERY: \
979        case OP_MINQUERY: \
980        case OP_MINSTARI: \
981        case OP_MINPLUSI: \
982        case OP_QUERYI: \
983        case OP_MINQUERYI: \
984        case OP_NOTMINSTAR: \
985        case OP_NOTMINPLUS: \
986        case OP_NOTQUERY: \
987        case OP_NOTMINQUERY: \
988        case OP_NOTMINSTARI: \
989        case OP_NOTMINPLUSI: \
990        case OP_NOTQUERYI: \
991        case OP_NOTMINQUERYI:
992    
993    #define CASE_ITERATOR_PRIVATE_DATA_2A \
994        case OP_STAR: \
995        case OP_PLUS: \
996        case OP_STARI: \
997        case OP_PLUSI: \
998        case OP_NOTSTAR: \
999        case OP_NOTPLUS: \
1000        case OP_NOTSTARI: \
1001        case OP_NOTPLUSI:
1002    
1003    #define CASE_ITERATOR_PRIVATE_DATA_2B \
1004        case OP_UPTO: \
1005        case OP_MINUPTO: \
1006        case OP_UPTOI: \
1007        case OP_MINUPTOI: \
1008        case OP_NOTUPTO: \
1009        case OP_NOTMINUPTO: \
1010        case OP_NOTUPTOI: \
1011        case OP_NOTMINUPTOI:
1012    
1013    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_1 \
1014        case OP_TYPEMINSTAR: \
1015        case OP_TYPEMINPLUS: \
1016        case OP_TYPEQUERY: \
1017        case OP_TYPEMINQUERY:
1018    
1019    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2A \
1020        case OP_TYPESTAR: \
1021        case OP_TYPEPLUS:
1022    
1023    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2B \
1024        case OP_TYPEUPTO: \
1025        case OP_TYPEMINUPTO:
1026    
1027    static void set_private_data_ptrs(compiler_common *common, int *private_data_start, pcre_uchar *ccend)
1028  {  {
1029  uschar *cc = common->start;  pcre_uchar *cc = common->start;
1030  uschar *alternative;  pcre_uchar *alternative;
1031    pcre_uchar *end = NULL;
1032    int private_data_ptr = *private_data_start;
1033    int space, size, bracketlen;
1034    
1035  while (cc < ccend)  while (cc < ccend)
1036    {    {
1037      space = 0;
1038      size = 0;
1039      bracketlen = 0;
1040      if (private_data_ptr > SLJIT_MAX_LOCAL_SIZE)
1041        return;
1042    
1043      if (*cc == OP_ONCE || *cc == OP_ONCE_NC || *cc == OP_BRA || *cc == OP_CBRA || *cc == OP_COND)
1044        if (detect_repeat(common, cc))
1045          {
1046          /* These brackets are converted to repeats, so no global
1047          based single character repeat is allowed. */
1048          if (cc >= end)
1049            end = bracketend(cc);
1050          }
1051    
1052    switch(*cc)    switch(*cc)
1053      {      {
1054        case OP_KET:
1055        if (common->private_data_ptrs[cc + 1 - common->start] != 0)
1056          {
1057          common->private_data_ptrs[cc - common->start] = private_data_ptr;
1058          private_data_ptr += sizeof(sljit_sw);
1059          cc += common->private_data_ptrs[cc + 1 - common->start];
1060          }
1061        cc += 1 + LINK_SIZE;
1062        break;
1063    
1064      case OP_ASSERT:      case OP_ASSERT:
1065      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1066      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1067      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1068      case OP_ONCE:      case OP_ONCE:
1069        case OP_ONCE_NC:
1070      case OP_BRAPOS:      case OP_BRAPOS:
1071      case OP_SBRA:      case OP_SBRA:
1072      case OP_SBRAPOS:      case OP_SBRAPOS:
1073      case OP_SCOND:      case OP_SCOND:
1074      common->localptrs[cc - common->start] = localptr;      common->private_data_ptrs[cc - common->start] = private_data_ptr;
1075      localptr += sizeof(sljit_w);      private_data_ptr += sizeof(sljit_sw);
1076      cc += 1 + LINK_SIZE;      bracketlen = 1 + LINK_SIZE;
1077      break;      break;
1078    
1079      case OP_CBRAPOS:      case OP_CBRAPOS:
1080      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1081      common->localptrs[cc - common->start] = localptr;      common->private_data_ptrs[cc - common->start] = private_data_ptr;
1082      localptr += sizeof(sljit_w);      private_data_ptr += sizeof(sljit_sw);
1083      cc += 1 + LINK_SIZE + 2;      bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
1084      break;      break;
1085    
1086      case OP_COND:      case OP_COND:
# Line 672  while (cc < ccend) Line 1088  while (cc < ccend)
1088      alternative = cc + GET(cc, 1);      alternative = cc + GET(cc, 1);
1089      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1090        {        {
1091        common->localptrs[cc - common->start] = localptr;        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1092        localptr += sizeof(sljit_w);        private_data_ptr += sizeof(sljit_sw);
1093        }        }
1094      cc += 1 + LINK_SIZE;      bracketlen = 1 + LINK_SIZE;
1095        break;
1096    
1097        case OP_BRA:
1098        bracketlen = 1 + LINK_SIZE;
1099        break;
1100    
1101        case OP_CBRA:
1102        case OP_SCBRA:
1103        bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
1104        break;
1105    
1106        CASE_ITERATOR_PRIVATE_DATA_1
1107        space = 1;
1108        size = -2;
1109        break;
1110    
1111        CASE_ITERATOR_PRIVATE_DATA_2A
1112        space = 2;
1113        size = -2;
1114        break;
1115    
1116        CASE_ITERATOR_PRIVATE_DATA_2B
1117        space = 2;
1118        size = -(2 + IMM2_SIZE);
1119      break;      break;
1120    
1121        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1122        space = 1;
1123        size = 1;
1124        break;
1125    
1126        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1127        if (cc[1] != OP_ANYNL && cc[1] != OP_EXTUNI)
1128          space = 2;
1129        size = 1;
1130        break;
1131    
1132        CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1133        if (cc[1 + IMM2_SIZE] != OP_ANYNL && cc[1 + IMM2_SIZE] != OP_EXTUNI)
1134          space = 2;
1135        size = 1 + IMM2_SIZE;
1136        break;
1137    
1138        case OP_CLASS:
1139        case OP_NCLASS:
1140        size += 1 + 32 / sizeof(pcre_uchar);
1141        space = get_class_iterator_size(cc + size);
1142        break;
1143    
1144    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1145        case OP_XCLASS:
1146        size = GET(cc, 1);
1147        space = get_class_iterator_size(cc + size);
1148        break;
1149    #endif
1150    
1151      default:      default:
1152      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
1153      SLJIT_ASSERT(cc != NULL);      SLJIT_ASSERT(cc != NULL);
1154      break;      break;
1155      }      }
1156    
1157      /* Character iterators, which are not inside a repeated bracket,
1158         gets a private slot instead of allocating it on the stack. */
1159      if (space > 0 && cc >= end)
1160        {
1161        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1162        private_data_ptr += sizeof(sljit_sw) * space;
1163        }
1164    
1165      if (size != 0)
1166        {
1167        if (size < 0)
1168          {
1169          cc += -size;
1170    #ifdef SUPPORT_UTF
1171          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1172    #endif
1173          }
1174        else
1175          cc += size;
1176        }
1177    
1178      if (bracketlen > 0)
1179        {
1180        if (cc >= end)
1181          {
1182          end = bracketend(cc);
1183          if (end[-1 - LINK_SIZE] == OP_KET)
1184            end = NULL;
1185          }
1186        cc += bracketlen;
1187        }
1188    }    }
1189    *private_data_start = private_data_ptr;
1190  }  }
1191    
1192  /* Returns with -1 if no need for frame. */  /* Returns with a frame_types (always < 0) if no need for frame. */
1193  static int get_framesize(compiler_common *common, uschar *cc, BOOL recursive)  static int get_framesize(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, BOOL recursive, BOOL* needs_control_head)
1194  {  {
 uschar *ccend = bracketend(cc);  
 uschar *end;  
1195  int length = 0;  int length = 0;
1196  BOOL possessive = FALSE;  int possessive = 0;
1197  BOOL needs_frame = FALSE;  BOOL stack_restore = FALSE;
1198  BOOL needs_maxindex = FALSE;  BOOL setsom_found = recursive;
1199  BOOL setsom_found = FALSE;  BOOL setmark_found = recursive;
1200    /* The last capture is a local variable even for recursions. */
1201  if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))  BOOL capture_last_found = FALSE;
1202    {  
1203    length = 3 + 2;  #if defined DEBUG_FORCE_CONTROL_HEAD && DEBUG_FORCE_CONTROL_HEAD
1204    needs_maxindex = TRUE;  SLJIT_ASSERT(common->control_head_ptr != 0);
1205    possessive = TRUE;  *needs_control_head = TRUE;
1206    #else
1207    *needs_control_head = FALSE;
1208    #endif
1209    
1210    if (ccend == NULL)
1211      {
1212      ccend = bracketend(cc) - (1 + LINK_SIZE);
1213      if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))
1214        {
1215        possessive = length = (common->capture_last_ptr != 0) ? 5 : 3;
1216        /* This is correct regardless of common->capture_last_ptr. */
1217        capture_last_found = TRUE;
1218        }
1219      cc = next_opcode(common, cc);
1220    }    }
1221    
 cc = next_opcode(common, cc);  
1222  SLJIT_ASSERT(cc != NULL);  SLJIT_ASSERT(cc != NULL);
1223  while (cc < ccend)  while (cc < ccend)
1224    switch(*cc)    switch(*cc)
1225      {      {
1226      case OP_SET_SOM:      case OP_SET_SOM:
1227      case OP_RECURSE:      SLJIT_ASSERT(common->has_set_som);
1228        stack_restore = TRUE;
1229      if (!setsom_found)      if (!setsom_found)
1230        {        {
1231        length += 2;        length += 2;
1232        setsom_found = TRUE;        setsom_found = TRUE;
1233        }        }
1234      cc += (*cc == OP_SET_SOM) ? 1 : 1 + LINK_SIZE;      cc += 1;
1235      break;      break;
1236    
1237      case OP_ASSERT:      case OP_MARK:
1238      case OP_ASSERT_NOT:      case OP_PRUNE_ARG:
1239      case OP_ASSERTBACK:      case OP_THEN_ARG:
1240      case OP_ASSERTBACK_NOT:      SLJIT_ASSERT(common->mark_ptr != 0);
1241      case OP_ONCE:      stack_restore = TRUE;
1242      if (needs_frame || length > 0)      if (!setmark_found)
1243        {        {
1244        cc = bracketend(cc);        length += 2;
1245        break;        setmark_found = TRUE;
1246        }        }
1247      /* Check whether a frame must be created. */      if (common->control_head_ptr != 0)
1248      end = bracketend(cc);        *needs_control_head = TRUE;
1249      while (cc < end)      cc += 1 + 2 + cc[1];
1250        {      break;
1251        if (*cc == OP_SET_SOM || *cc == OP_CBRA || *cc == OP_CBRAPOS  
1252            || *cc == OP_SCBRA || *cc == OP_SCBRAPOS || *cc == OP_RECURSE)      case OP_RECURSE:
1253          needs_frame = TRUE;      stack_restore = TRUE;
1254        cc = next_opcode(common, cc);      if (common->has_set_som && !setsom_found)
1255        SLJIT_ASSERT(cc != NULL);        {
1256          length += 2;
1257          setsom_found = TRUE;
1258          }
1259        if (common->mark_ptr != 0 && !setmark_found)
1260          {
1261          length += 2;
1262          setmark_found = TRUE;
1263          }
1264        if (common->capture_last_ptr != 0 && !capture_last_found)
1265          {
1266          length += 2;
1267          capture_last_found = TRUE;
1268        }        }
1269        cc += 1 + LINK_SIZE;
1270      break;      break;
1271    
1272      case OP_CBRA:      case OP_CBRA:
1273      case OP_CBRAPOS:      case OP_CBRAPOS:
1274      case OP_SCBRA:      case OP_SCBRA:
1275      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1276      if (!needs_maxindex)      stack_restore = TRUE;
1277        if (common->capture_last_ptr != 0 && !capture_last_found)
1278        {        {
       needs_maxindex = TRUE;  
1279        length += 2;        length += 2;
1280          capture_last_found = TRUE;
1281        }        }
1282      length += 3;      length += 3;
1283      cc += 1 + LINK_SIZE + 2;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1284      break;      break;
1285    
1286      default:      default:
1287        stack_restore = TRUE;
1288        /* Fall through. */
1289    
1290        case OP_NOT_WORD_BOUNDARY:
1291        case OP_WORD_BOUNDARY:
1292        case OP_NOT_DIGIT:
1293        case OP_DIGIT:
1294        case OP_NOT_WHITESPACE:
1295        case OP_WHITESPACE:
1296        case OP_NOT_WORDCHAR:
1297        case OP_WORDCHAR:
1298        case OP_ANY:
1299        case OP_ALLANY:
1300        case OP_ANYBYTE:
1301        case OP_NOTPROP:
1302        case OP_PROP:
1303        case OP_ANYNL:
1304        case OP_NOT_HSPACE:
1305        case OP_HSPACE:
1306        case OP_NOT_VSPACE:
1307        case OP_VSPACE:
1308        case OP_EXTUNI:
1309        case OP_EODN:
1310        case OP_EOD:
1311        case OP_CIRC:
1312        case OP_CIRCM:
1313        case OP_DOLL:
1314        case OP_DOLLM:
1315        case OP_CHAR:
1316        case OP_CHARI:
1317        case OP_NOT:
1318        case OP_NOTI:
1319    
1320        case OP_EXACT:
1321        case OP_POSSTAR:
1322        case OP_POSPLUS:
1323        case OP_POSQUERY:
1324        case OP_POSUPTO:
1325    
1326        case OP_EXACTI:
1327        case OP_POSSTARI:
1328        case OP_POSPLUSI:
1329        case OP_POSQUERYI:
1330        case OP_POSUPTOI:
1331    
1332        case OP_NOTEXACT:
1333        case OP_NOTPOSSTAR:
1334        case OP_NOTPOSPLUS:
1335        case OP_NOTPOSQUERY:
1336        case OP_NOTPOSUPTO:
1337    
1338        case OP_NOTEXACTI:
1339        case OP_NOTPOSSTARI:
1340        case OP_NOTPOSPLUSI:
1341        case OP_NOTPOSQUERYI:
1342        case OP_NOTPOSUPTOI:
1343    
1344        case OP_TYPEEXACT:
1345        case OP_TYPEPOSSTAR:
1346        case OP_TYPEPOSPLUS:
1347        case OP_TYPEPOSQUERY:
1348        case OP_TYPEPOSUPTO:
1349    
1350        case OP_CLASS:
1351        case OP_NCLASS:
1352        case OP_XCLASS:
1353    
1354      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
1355      SLJIT_ASSERT(cc != NULL);      SLJIT_ASSERT(cc != NULL);
1356      break;      break;
1357      }      }
1358    
1359  /* Possessive quantifiers can use a special case. */  /* Possessive quantifiers can use a special case. */
1360  if (SLJIT_UNLIKELY(possessive) && !needs_frame && length == 3 + 2)  if (SLJIT_UNLIKELY(possessive == length))
1361    return -1;    return stack_restore ? no_frame : no_stack;
1362    
1363  if (length > 0)  if (length > 0)
1364    return length + 2;    return length + 1;
1365  return needs_frame ? 0 : -1;  return stack_restore ? no_frame : no_stack;
1366  }  }
1367    
1368  static void init_frame(compiler_common *common, uschar *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)
1369  {  {
 /* TMP2 must contain STACK_TOP - (-STACK(stackpos)) */  
1370  DEFINE_COMPILER;  DEFINE_COMPILER;
1371  uschar *ccend = bracketend(cc);  BOOL setsom_found = recursive;
1372  BOOL needs_maxindex = FALSE;  BOOL setmark_found = recursive;
1373  BOOL setsom_found = FALSE;  /* The last capture is a local variable even for recursions. */
1374    BOOL capture_last_found = FALSE;
1375  int offset;  int offset;
1376    
1377  if (stackpos < stacktop)  /* >= 1 + shortest item size (2) */
1378    {  SLJIT_UNUSED_ARG(stacktop);
1379    SLJIT_ASSERT(stackpos + 1 == stacktop);  SLJIT_ASSERT(stackpos >= stacktop + 2);
   return;  
   }  
1380    
1381  stackpos = STACK(stackpos);  stackpos = STACK(stackpos);
1382  OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS_HEAD);  if (ccend == NULL)
1383  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS_HEAD, TMP2, 0);    {
1384  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), STACK(stacktop), TMP1, 0);    ccend = bracketend(cc) - (1 + LINK_SIZE);
1385      if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))
1386        cc = next_opcode(common, cc);
1387      }
1388    
 if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))  
   cc = next_opcode(common, cc);  
1389  SLJIT_ASSERT(cc != NULL);  SLJIT_ASSERT(cc != NULL);
1390  while (cc < ccend)  while (cc < ccend)
1391    switch(*cc)    switch(*cc)
1392      {      {
1393      case OP_SET_SOM:      case OP_SET_SOM:
1394      case OP_RECURSE:      SLJIT_ASSERT(common->has_set_som);
1395      if (!setsom_found)      if (!setsom_found)
1396        {        {
1397        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));
1398        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setstrbegin);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1399        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1400        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1401        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1402        setsom_found = TRUE;        setsom_found = TRUE;
1403        }        }
1404      cc += (*cc == OP_SET_SOM) ? 1 : 1 + LINK_SIZE;      cc += 1;
1405      break;      break;
1406    
1407      case OP_ASSERT:      case OP_MARK:
1408      case OP_ASSERT_NOT:      case OP_PRUNE_ARG:
1409      case OP_ASSERTBACK:      case OP_THEN_ARG:
1410      case OP_ASSERTBACK_NOT:      SLJIT_ASSERT(common->mark_ptr != 0);
1411      case OP_ONCE:      if (!setmark_found)
1412      cc = bracketend(cc);        {
1413          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);
1414          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
1415          stackpos += (int)sizeof(sljit_sw);
1416          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1417          stackpos += (int)sizeof(sljit_sw);
1418          setmark_found = TRUE;
1419          }
1420        cc += 1 + 2 + cc[1];
1421        break;
1422    
1423        case OP_RECURSE:
1424        if (common->has_set_som && !setsom_found)
1425          {
1426          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));
1427          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1428          stackpos += (int)sizeof(sljit_sw);
1429          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1430          stackpos += (int)sizeof(sljit_sw);
1431          setsom_found = TRUE;
1432          }
1433        if (common->mark_ptr != 0 && !setmark_found)
1434          {
1435          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);
1436          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
1437          stackpos += (int)sizeof(sljit_sw);
1438          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1439          stackpos += (int)sizeof(sljit_sw);
1440          setmark_found = TRUE;
1441          }
1442        if (common->capture_last_ptr != 0 && !capture_last_found)
1443          {
1444          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->capture_last_ptr);
1445          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1446          stackpos += (int)sizeof(sljit_sw);
1447          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1448          stackpos += (int)sizeof(sljit_sw);
1449          capture_last_found = TRUE;
1450          }
1451        cc += 1 + LINK_SIZE;
1452      break;      break;
1453    
1454      case OP_CBRA:      case OP_CBRA:
1455      case OP_CBRAPOS:      case OP_CBRAPOS:
1456      case OP_SCBRA:      case OP_SCBRA:
1457      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1458      if (!needs_maxindex)      if (common->capture_last_ptr != 0 && !capture_last_found)
1459        {        {
1460        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setmaxindex);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->capture_last_ptr);
1461        stackpos += (int)sizeof(sljit_w);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1462        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, MAX_INDEX, 0);        stackpos += (int)sizeof(sljit_sw);
1463        stackpos += (int)sizeof(sljit_w);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1464        needs_maxindex = TRUE;        stackpos += (int)sizeof(sljit_sw);
1465          capture_last_found = TRUE;
1466        }        }
1467      offset = (GET2(cc, 1 + LINK_SIZE)) << 1;      offset = (GET2(cc, 1 + LINK_SIZE)) << 1;
1468      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));
1469      stackpos += (int)sizeof(sljit_w);      stackpos += (int)sizeof(sljit_sw);
1470      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset));      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset));
1471      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset + 1));      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset + 1));
1472      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1473      stackpos += (int)sizeof(sljit_w);      stackpos += (int)sizeof(sljit_sw);
1474      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);
1475      stackpos += (int)sizeof(sljit_w);      stackpos += (int)sizeof(sljit_sw);
1476    
1477      cc += 1 + LINK_SIZE + 2;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1478      break;      break;
1479    
1480      default:      default:
# Line 848  while (cc < ccend) Line 1483  while (cc < ccend)
1483      break;      break;
1484      }      }
1485    
1486  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_end);  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, 0);
1487  SLJIT_ASSERT(stackpos == STACK(stacktop + 1));  SLJIT_ASSERT(stackpos == STACK(stacktop));
1488  }  }
1489    
1490  static SLJIT_INLINE int get_localsize(compiler_common *common, uschar *cc, uschar *ccend)  static SLJIT_INLINE int get_private_data_copy_length(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, BOOL needs_control_head)
1491  {  {
1492  int localsize = 2;  int private_data_length = needs_control_head ? 3 : 2;
1493  uschar *alternative;  int size;
1494  /* Calculate the sum of the local variables. */  pcre_uchar *alternative;
1495    /* Calculate the sum of the private machine words. */
1496  while (cc < ccend)  while (cc < ccend)
1497    {    {
1498      size = 0;
1499    switch(*cc)    switch(*cc)
1500      {      {
1501        case OP_KET:
1502        if (PRIVATE_DATA(cc) != 0)
1503          private_data_length++;
1504        cc += 1 + LINK_SIZE;
1505        break;
1506    
1507      case OP_ASSERT:      case OP_ASSERT:
1508      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1509      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1510      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1511      case OP_ONCE:      case OP_ONCE:
1512        case OP_ONCE_NC:
1513      case OP_BRAPOS:      case OP_BRAPOS:
1514      case OP_SBRA:      case OP_SBRA:
1515      case OP_SBRAPOS:      case OP_SBRAPOS:
1516      case OP_SCOND:      case OP_SCOND:
1517      localsize++;      private_data_length++;
1518      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1519      break;      break;
1520    
1521      case OP_CBRA:      case OP_CBRA:
1522      case OP_SCBRA:      case OP_SCBRA:
1523      localsize++;      if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1524      cc += 1 + LINK_SIZE + 2;        private_data_length++;
1525        cc += 1 + LINK_SIZE + IMM2_SIZE;
1526      break;      break;
1527    
1528      case OP_CBRAPOS:      case OP_CBRAPOS:
1529      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1530      localsize += 2;      private_data_length += 2;
1531      cc += 1 + LINK_SIZE + 2;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1532      break;      break;
1533    
1534      case OP_COND:      case OP_COND:
1535      /* Might be a hidden SCOND. */      /* Might be a hidden SCOND. */
1536      alternative = cc + GET(cc, 1);      alternative = cc + GET(cc, 1);
1537      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1538        localsize++;        private_data_length++;
1539      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1540      break;      break;
1541    
1542        CASE_ITERATOR_PRIVATE_DATA_1
1543        if (PRIVATE_DATA(cc))
1544          private_data_length++;
1545        cc += 2;
1546    #ifdef SUPPORT_UTF
1547        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1548    #endif
1549        break;
1550    
1551        CASE_ITERATOR_PRIVATE_DATA_2A
1552        if (PRIVATE_DATA(cc))
1553          private_data_length += 2;
1554        cc += 2;
1555    #ifdef SUPPORT_UTF
1556        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1557    #endif
1558        break;
1559    
1560        CASE_ITERATOR_PRIVATE_DATA_2B
1561        if (PRIVATE_DATA(cc))
1562          private_data_length += 2;
1563        cc += 2 + IMM2_SIZE;
1564    #ifdef SUPPORT_UTF
1565        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1566    #endif
1567        break;
1568    
1569        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1570        if (PRIVATE_DATA(cc))
1571          private_data_length++;
1572        cc += 1;
1573        break;
1574    
1575        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1576        if (PRIVATE_DATA(cc))
1577          private_data_length += 2;
1578        cc += 1;
1579        break;
1580    
1581        CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1582        if (PRIVATE_DATA(cc))
1583          private_data_length += 2;
1584        cc += 1 + IMM2_SIZE;
1585        break;
1586    
1587        case OP_CLASS:
1588        case OP_NCLASS:
1589    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1590        case OP_XCLASS:
1591        size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
1592    #else
1593        size = 1 + 32 / (int)sizeof(pcre_uchar);
1594    #endif
1595        if (PRIVATE_DATA(cc))
1596          private_data_length += get_class_iterator_size(cc + size);
1597        cc += size;
1598        break;
1599    
1600      default:      default:
1601      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
1602      SLJIT_ASSERT(cc != NULL);      SLJIT_ASSERT(cc != NULL);
# Line 901  while (cc < ccend) Line 1604  while (cc < ccend)
1604      }      }
1605    }    }
1606  SLJIT_ASSERT(cc == ccend);  SLJIT_ASSERT(cc == ccend);
1607  return localsize;  return private_data_length;
1608  }  }
1609    
1610  static void copy_locals(compiler_common *common, uschar *cc, uschar *ccend,  static void copy_private_data(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend,
1611    BOOL save, int stackptr, int stacktop)    BOOL save, int stackptr, int stacktop, BOOL needs_control_head)
1612  {  {
1613  DEFINE_COMPILER;  DEFINE_COMPILER;
1614  int srcw[2];  int srcw[2];
1615  int count;  int count, size;
1616  BOOL tmp1next = TRUE;  BOOL tmp1next = TRUE;
1617  BOOL tmp1empty = TRUE;  BOOL tmp1empty = TRUE;
1618  BOOL tmp2empty = TRUE;  BOOL tmp2empty = TRUE;
1619  uschar *alternative;  pcre_uchar *alternative;
1620  enum {  enum {
1621    start,    start,
1622    loop,    loop,
# Line 926  stacktop = STACK(stacktop - 1); Line 1629  stacktop = STACK(stacktop - 1);
1629    
1630  if (!save)  if (!save)
1631    {    {
1632    stackptr += sizeof(sljit_w);    stackptr += (needs_control_head ? 2 : 1) * sizeof(sljit_sw);
1633    if (stackptr < stacktop)    if (stackptr < stacktop)
1634      {      {
1635      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1636      stackptr += sizeof(sljit_w);      stackptr += sizeof(sljit_sw);
1637      tmp1empty = FALSE;      tmp1empty = FALSE;
1638      }      }
1639    if (stackptr < stacktop)    if (stackptr < stacktop)
1640      {      {
1641      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1642      stackptr += sizeof(sljit_w);      stackptr += sizeof(sljit_sw);
1643      tmp2empty = FALSE;      tmp2empty = FALSE;
1644      }      }
1645    /* The tmp1next must be TRUE in either way. */    /* The tmp1next must be TRUE in either way. */
1646    }    }
1647    
1648  while (status != end)  do
1649    {    {
1650    count = 0;    count = 0;
1651    switch(status)    switch(status)
1652      {      {
1653      case start:      case start:
1654      SLJIT_ASSERT(save);      SLJIT_ASSERT(save && common->recursive_head_ptr != 0);
1655      count = 1;      count = 1;
1656      srcw[0] = RECURSIVE_HEAD;      srcw[0] = common->recursive_head_ptr;
1657        if (needs_control_head)
1658          {
1659          SLJIT_ASSERT(common->control_head_ptr != 0);
1660          count = 2;
1661          srcw[1] = common->control_head_ptr;
1662          }
1663      status = loop;      status = loop;
1664      break;      break;
1665    
# Line 963  while (status != end) Line 1672  while (status != end)
1672    
1673      switch(*cc)      switch(*cc)
1674        {        {
1675          case OP_KET:
1676          if (PRIVATE_DATA(cc) != 0)
1677            {
1678            count = 1;
1679            srcw[0] = PRIVATE_DATA(cc);
1680            }
1681          cc += 1 + LINK_SIZE;
1682          break;
1683    
1684        case OP_ASSERT:        case OP_ASSERT:
1685        case OP_ASSERT_NOT:        case OP_ASSERT_NOT:
1686        case OP_ASSERTBACK:        case OP_ASSERTBACK:
1687        case OP_ASSERTBACK_NOT:        case OP_ASSERTBACK_NOT:
1688        case OP_ONCE:        case OP_ONCE:
1689          case OP_ONCE_NC:
1690        case OP_BRAPOS:        case OP_BRAPOS:
1691        case OP_SBRA:        case OP_SBRA:
1692        case OP_SBRAPOS:        case OP_SBRAPOS:
1693        case OP_SCOND:        case OP_SCOND:
1694        count = 1;        count = 1;
1695        srcw[0] = PRIV(cc);        srcw[0] = PRIVATE_DATA(cc);
1696        SLJIT_ASSERT(srcw[0] != 0);        SLJIT_ASSERT(srcw[0] != 0);
1697        cc += 1 + LINK_SIZE;        cc += 1 + LINK_SIZE;
1698        break;        break;
1699    
1700        case OP_CBRA:        case OP_CBRA:
1701        case OP_SCBRA:        case OP_SCBRA:
1702        count = 1;        if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1703        srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));          {
1704        cc += 1 + LINK_SIZE + 2;          count = 1;
1705            srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1706            }
1707          cc += 1 + LINK_SIZE + IMM2_SIZE;
1708        break;        break;
1709    
1710        case OP_CBRAPOS:        case OP_CBRAPOS:
1711        case OP_SCBRAPOS:        case OP_SCBRAPOS:
1712        count = 2;        count = 2;
1713          srcw[0] = PRIVATE_DATA(cc);
1714        srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));        srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1715        srcw[0] = PRIV(cc);        SLJIT_ASSERT(srcw[0] != 0 && srcw[1] != 0);
1716        SLJIT_ASSERT(srcw[0] != 0);        cc += 1 + LINK_SIZE + IMM2_SIZE;
       cc += 1 + LINK_SIZE + 2;  
1717        break;        break;
1718    
1719        case OP_COND:        case OP_COND:
# Line 1000  while (status != end) Line 1722  while (status != end)
1722        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1723          {          {
1724          count = 1;          count = 1;
1725          srcw[0] = PRIV(cc);          srcw[0] = PRIVATE_DATA(cc);
1726          SLJIT_ASSERT(srcw[0] != 0);          SLJIT_ASSERT(srcw[0] != 0);
1727          }          }
1728        cc += 1 + LINK_SIZE;        cc += 1 + LINK_SIZE;
1729        break;        break;
1730    
1731          CASE_ITERATOR_PRIVATE_DATA_1
1732          if (PRIVATE_DATA(cc))
1733            {
1734            count = 1;
1735            srcw[0] = PRIVATE_DATA(cc);
1736            }
1737          cc += 2;
1738    #ifdef SUPPORT_UTF
1739          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1740    #endif
1741          break;
1742    
1743          CASE_ITERATOR_PRIVATE_DATA_2A
1744          if (PRIVATE_DATA(cc))
1745            {
1746            count = 2;
1747            srcw[0] = PRIVATE_DATA(cc);
1748            srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1749            }
1750          cc += 2;
1751    #ifdef SUPPORT_UTF
1752          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1753    #endif
1754          break;
1755    
1756          CASE_ITERATOR_PRIVATE_DATA_2B
1757          if (PRIVATE_DATA(cc))
1758            {
1759            count = 2;
1760            srcw[0] = PRIVATE_DATA(cc);
1761            srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1762            }
1763          cc += 2 + IMM2_SIZE;
1764    #ifdef SUPPORT_UTF
1765          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1766    #endif
1767          break;
1768    
1769          CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1770          if (PRIVATE_DATA(cc))
1771            {
1772            count = 1;
1773            srcw[0] = PRIVATE_DATA(cc);
1774            }
1775          cc += 1;
1776          break;
1777    
1778          CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1779          if (PRIVATE_DATA(cc))
1780            {
1781            count = 2;
1782            srcw[0] = PRIVATE_DATA(cc);
1783            srcw[1] = srcw[0] + sizeof(sljit_sw);
1784            }
1785          cc += 1;
1786          break;
1787    
1788          CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1789          if (PRIVATE_DATA(cc))
1790            {
1791            count = 2;
1792            srcw[0] = PRIVATE_DATA(cc);
1793            srcw[1] = srcw[0] + sizeof(sljit_sw);
1794            }
1795          cc += 1 + IMM2_SIZE;
1796          break;
1797    
1798          case OP_CLASS:
1799          case OP_NCLASS:
1800    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1801          case OP_XCLASS:
1802          size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
1803    #else
1804          size = 1 + 32 / (int)sizeof(pcre_uchar);
1805    #endif
1806          if (PRIVATE_DATA(cc))
1807            switch(get_class_iterator_size(cc + size))
1808              {
1809              case 1:
1810              count = 1;
1811              srcw[0] = PRIVATE_DATA(cc);
1812              break;
1813    
1814              case 2:
1815              count = 2;
1816              srcw[0] = PRIVATE_DATA(cc);
1817              srcw[1] = srcw[0] + sizeof(sljit_sw);
1818              break;
1819    
1820              default:
1821              SLJIT_ASSERT_STOP();
1822              break;
1823              }
1824          cc += size;
1825          break;
1826    
1827        default:        default:
1828        cc = next_opcode(common, cc);        cc = next_opcode(common, cc);
1829        SLJIT_ASSERT(cc != NULL);        SLJIT_ASSERT(cc != NULL);
# Line 1028  while (status != end) Line 1846  while (status != end)
1846          if (!tmp1empty)          if (!tmp1empty)
1847            {            {
1848            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1849            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1850            }            }
1851          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);
1852          tmp1empty = FALSE;          tmp1empty = FALSE;
# Line 1039  while (status != end) Line 1857  while (status != end)
1857          if (!tmp2empty)          if (!tmp2empty)
1858            {            {
1859            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1860            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1861            }            }
1862          OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);          OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);
1863          tmp2empty = FALSE;          tmp2empty = FALSE;
# Line 1056  while (status != end) Line 1874  while (status != end)
1874          if (!tmp1empty)          if (!tmp1empty)
1875            {            {
1876            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1877            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1878            }            }
1879          tmp1next = FALSE;          tmp1next = FALSE;
1880          }          }
# Line 1068  while (status != end) Line 1886  while (status != end)
1886          if (!tmp2empty)          if (!tmp2empty)
1887            {            {
1888            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1889            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1890            }            }
1891          tmp1next = TRUE;          tmp1next = TRUE;
1892          }          }
1893        }        }
1894      }      }
1895    }    }
1896    while (status != end);
1897    
1898  if (save)  if (save)
1899    {    {
# Line 1083  if (save) Line 1902  if (save)
1902      if (!tmp1empty)      if (!tmp1empty)
1903        {        {
1904        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1905        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1906        }        }
1907      if (!tmp2empty)      if (!tmp2empty)
1908        {        {
1909        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1910        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1911        }        }
1912      }      }
1913    else    else
# Line 1096  if (save) Line 1915  if (save)
1915      if (!tmp2empty)      if (!tmp2empty)
1916        {        {
1917        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1918        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1919        }        }
1920      if (!tmp1empty)      if (!tmp1empty)
1921        {        {
1922        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1923        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1924        }        }
1925      }      }
1926    }    }
1927  SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));  SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));
1928  }  }
1929    
1930  static SLJIT_INLINE BOOL ispowerof2(unsigned int value)  static SLJIT_INLINE pcre_uchar *set_then_offsets(compiler_common *common, pcre_uchar *cc, pcre_uint8 *current_offset)
1931    {
1932    pcre_uchar *end = bracketend(cc);
1933    BOOL has_alternatives = cc[GET(cc, 1)] == OP_ALT;
1934    
1935    /* Assert captures then. */
1936    if (*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT)
1937      current_offset = NULL;
1938    /* Conditional block does not. */
1939    if (*cc == OP_COND || *cc == OP_SCOND)
1940      has_alternatives = FALSE;
1941    
1942    cc = next_opcode(common, cc);
1943    if (has_alternatives)
1944      current_offset = common->then_offsets + (cc - common->start);
1945    
1946    while (cc < end)
1947      {
1948      if ((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND))
1949        cc = set_then_offsets(common, cc, current_offset);
1950      else
1951        {
1952        if (*cc == OP_ALT && has_alternatives)
1953          current_offset = common->then_offsets + (cc + 1 + LINK_SIZE - common->start);
1954        if (*cc >= OP_THEN && *cc <= OP_THEN_ARG && current_offset != NULL)
1955          *current_offset = 1;
1956        cc = next_opcode(common, cc);
1957        }
1958      }
1959    
1960    return end;
1961    }
1962    
1963    #undef CASE_ITERATOR_PRIVATE_DATA_1
1964    #undef CASE_ITERATOR_PRIVATE_DATA_2A
1965    #undef CASE_ITERATOR_PRIVATE_DATA_2B
1966    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1967    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1968    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1969    
1970    static SLJIT_INLINE BOOL is_powerof2(unsigned int value)
1971  {  {
1972  return (value & (value - 1)) == 0;  return (value & (value - 1)) == 0;
1973  }  }
# Line 1118  static SLJIT_INLINE void set_jumps(jump_ Line 1977  static SLJIT_INLINE void set_jumps(jump_
1977  while (list)  while (list)
1978    {    {
1979    /* sljit_set_label is clever enough to do nothing    /* sljit_set_label is clever enough to do nothing
1980    if either the jump or the label is NULL */    if either the jump or the label is NULL. */
1981    sljit_set_label(list->jump, label);    SET_LABEL(list->jump, label);
1982    list = list->next;    list = list->next;
1983    }    }
1984  }  }
# Line 1135  if (list_item) Line 1994  if (list_item)
1994    }    }
1995  }  }
1996    
1997  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)
1998  {  {
1999  DEFINE_COMPILER;  DEFINE_COMPILER;
2000  stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));  stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));
2001    
2002  if (list_item)  if (list_item)
2003    {    {
   list_item->type = type;  
   list_item->data = data;  
2004    list_item->start = start;    list_item->start = start;
2005    list_item->leave = LABEL();    list_item->quit = LABEL();
2006    list_item->next = common->stubs;    list_item->next = common->stubs;
2007    common->stubs = list_item;    common->stubs = list_item;
2008    }    }
# Line 1159  stub_list* list_item = common->stubs; Line 2016  stub_list* list_item = common->stubs;
2016  while (list_item)  while (list_item)
2017    {    {
2018    JUMPHERE(list_item->start);    JUMPHERE(list_item->start);
2019    switch(list_item->type)    add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));
2020      {    JUMPTO(SLJIT_JUMP, list_item->quit);
     case stack_alloc:  
     add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));  
     break;  
   
     case max_index:  
     OP1(SLJIT_MOV, MAX_INDEX, 0, SLJIT_IMM, list_item->data);  
     break;  
     }  
   JUMPTO(SLJIT_JUMP, list_item->leave);  
2021    list_item = list_item->next;    list_item = list_item->next;
2022    }    }
2023  common->stubs = NULL;  common->stubs = NULL;
2024  }  }
2025    
2026    static SLJIT_INLINE void count_match(compiler_common *common)
2027    {
2028    DEFINE_COMPILER;
2029    
2030    OP2(SLJIT_SUB | SLJIT_SET_E, COUNT_MATCH, 0, COUNT_MATCH, 0, SLJIT_IMM, 1);
2031    add_jump(compiler, &common->calllimit, JUMP(SLJIT_C_ZERO));
2032    }
2033    
2034  static SLJIT_INLINE void allocate_stack(compiler_common *common, int size)  static SLJIT_INLINE void allocate_stack(compiler_common *common, int size)
2035  {  {
2036  /* May destroy all locals and registers except TMP2. */  /* May destroy all locals and registers except TMP2. */
2037  DEFINE_COMPILER;  DEFINE_COMPILER;
2038    
2039  OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_w));  OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_sw));
2040  #ifdef DESTROY_REGISTERS  #ifdef DESTROY_REGISTERS
2041  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);
2042  OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);  OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);
# Line 1188  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0); Line 2044  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);
2044  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);
2045  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);
2046  #endif  #endif
2047  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));
2048  }  }
2049    
2050  static SLJIT_INLINE void free_stack(compiler_common *common, int size)  static SLJIT_INLINE void free_stack(compiler_common *common, int size)
2051  {  {
2052  DEFINE_COMPILER;  DEFINE_COMPILER;
2053  OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_w));  OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_sw));
2054  }  }
2055    
2056  static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)  static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)
# Line 1202  static SLJIT_INLINE void reset_ovector(c Line 2058  static SLJIT_INLINE void reset_ovector(c
2058  DEFINE_COMPILER;  DEFINE_COMPILER;
2059  struct sljit_label *loop;  struct sljit_label *loop;
2060  int i;  int i;
2061    
2062  /* At this point we can freely use all temporary registers. */  /* At this point we can freely use all temporary registers. */
2063    SLJIT_ASSERT(length > 1);
2064  /* TMP1 returns with begin - 1. */  /* TMP1 returns with begin - 1. */
2065  OP1(SLJIT_MOV, MAX_INDEX, 0, SLJIT_IMM, 1);  OP2(SLJIT_SUB, SLJIT_SCRATCH_REG1, 0, SLJIT_MEM1(SLJIT_SAVED_REG1), SLJIT_OFFSETOF(jit_arguments, begin), SLJIT_IMM, IN_UCHARS(1));
2066  OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG1, 0, SLJIT_MEM1(SLJIT_GENERAL_REG1), SLJIT_OFFSETOF(jit_arguments, begin), SLJIT_IMM, 1);  if (length < 8)
2067      {
2068      for (i = 1; i < length; i++)
2069        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), SLJIT_SCRATCH_REG1, 0);
2070      }
2071    else
2072      {
2073      GET_LOCAL_BASE(SLJIT_SCRATCH_REG2, 0, OVECTOR_START);
2074      OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, length - 1);
2075      loop = LABEL();
2076      OP1(SLJIT_MOVU, SLJIT_MEM1(SLJIT_SCRATCH_REG2), sizeof(sljit_sw), SLJIT_SCRATCH_REG1, 0);
2077      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 1);
2078      JUMPTO(SLJIT_C_NOT_ZERO, loop);
2079      }
2080    }
2081    
2082    static SLJIT_INLINE void do_reset_match(compiler_common *common, int length)
2083    {
2084    DEFINE_COMPILER;
2085    struct sljit_label *loop;
2086    int i;
2087    
2088    SLJIT_ASSERT(length > 1);
2089    /* OVECTOR(1) contains the "string begin - 1" constant. */
2090    if (length > 2)
2091      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));
2092  if (length < 8)  if (length < 8)
2093    {    {
2094    for (i = 0; i < length; i++)    for (i = 2; i < length; i++)
2095      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), SLJIT_TEMPORARY_REG1, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), TMP1, 0);
2096    }    }
2097  else  else
2098    {    {
2099    OP2(SLJIT_ADD, SLJIT_TEMPORARY_REG2, 0, SLJIT_LOCALS_REG, 0, SLJIT_IMM, OVECTOR_START - sizeof(sljit_w));    GET_LOCAL_BASE(TMP2, 0, OVECTOR_START + sizeof(sljit_sw));
2100    OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, length);    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_IMM, length - 2);
2101    loop = LABEL();    loop = LABEL();
2102    OP1(SLJIT_MOVU, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), sizeof(sljit_w), SLJIT_TEMPORARY_REG1, 0);    OP1(SLJIT_MOVU, SLJIT_MEM1(TMP2), sizeof(sljit_sw), TMP1, 0);
2103    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_TEMPORARY_REG3, 0, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, 1);    OP2(SLJIT_SUB | SLJIT_SET_E, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, 1);
2104    JUMPTO(SLJIT_C_NOT_ZERO, loop);    JUMPTO(SLJIT_C_NOT_ZERO, loop);
2105    }    }
2106    
2107    OP1(SLJIT_MOV, STACK_TOP, 0, ARGUMENTS, 0);
2108    if (common->mark_ptr != 0)
2109      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr, SLJIT_IMM, 0);
2110    if (common->control_head_ptr != 0)
2111      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->control_head_ptr, SLJIT_IMM, 0);
2112    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(jit_arguments, stack));
2113    OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_ptr);
2114    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(struct sljit_stack, base));
2115    }
2116    
2117    static sljit_sw SLJIT_CALL do_search_mark(sljit_sw *current, const pcre_uchar *skip_arg)
2118    {
2119    while (current != NULL)
2120      {
2121      switch (current[-2])
2122        {
2123        case type_then_trap:
2124        break;
2125    
2126        case type_mark:
2127        if (STRCMP_UC_UC(skip_arg, (pcre_uchar *)current[-3]) == 0)
2128          return current[-4];
2129        break;
2130    
2131        default:
2132        SLJIT_ASSERT_STOP();
2133        break;
2134        }
2135      current = (sljit_sw*)current[-1];
2136      }
2137    return -1;
2138  }  }
2139    
2140  static SLJIT_INLINE void copy_ovector(compiler_common *common)  static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)
2141  {  {
2142  DEFINE_COMPILER;  DEFINE_COMPILER;
2143  struct sljit_label *loop;  struct sljit_label *loop;
2144  struct sljit_jump *earlyexit;  struct sljit_jump *early_quit;
2145    
2146  /* At this point we can freely use all registers. */  /* At this point we can freely use all registers. */
2147  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, ARGUMENTS, 0);  OP1(SLJIT_MOV, SLJIT_SAVED_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));
2148  OP1(SLJIT_MOV_SI, SLJIT_TEMPORARY_REG2, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, offsetcount));  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1), STR_PTR, 0);
2149  OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));  
2150  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, begin));  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG1, 0, ARGUMENTS, 0);
2151  OP2(SLJIT_ADD, SLJIT_GENERAL_REG1, 0, SLJIT_LOCALS_REG, 0, SLJIT_IMM, OVECTOR_START);  if (common->mark_ptr != 0)
2152      OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);
2153    OP1(SLJIT_MOV_SI, SLJIT_SCRATCH_REG2, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, offset_count));
2154    if (common->mark_ptr != 0)
2155      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, mark_ptr), SLJIT_SCRATCH_REG3, 0);
2156    OP2(SLJIT_SUB, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));
2157    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG1, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, begin));
2158    GET_LOCAL_BASE(SLJIT_SAVED_REG1, 0, OVECTOR_START);
2159  /* Unlikely, but possible */  /* Unlikely, but possible */
2160  earlyexit = CMP(SLJIT_C_EQUAL, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, 0);  early_quit = CMP(SLJIT_C_EQUAL, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, 0);
2161  loop = LABEL();  loop = LABEL();
2162  OP2(SLJIT_SUB, SLJIT_GENERAL_REG2, 0, SLJIT_MEM1(SLJIT_GENERAL_REG1), 0, SLJIT_TEMPORARY_REG1, 0);  OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, SLJIT_MEM1(SLJIT_SAVED_REG1), 0, SLJIT_SCRATCH_REG1, 0);
2163  OP2(SLJIT_ADD, SLJIT_GENERAL_REG1, 0, SLJIT_GENERAL_REG1, 0, SLJIT_IMM, sizeof(sljit_w));  OP2(SLJIT_ADD, SLJIT_SAVED_REG1, 0, SLJIT_SAVED_REG1, 0, SLJIT_IMM, sizeof(sljit_sw));
2164  /* Copy the integer value to the output buffer */  /* Copy the integer value to the output buffer */
2165  OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_TEMPORARY_REG3), sizeof(int), SLJIT_GENERAL_REG2, 0);  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2166  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_TEMPORARY_REG2, 0, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, 1);  OP2(SLJIT_ASHR, SLJIT_SAVED_REG2, 0, SLJIT_SAVED_REG2, 0, SLJIT_IMM, UCHAR_SHIFT);
2167    #endif
2168    OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG3), sizeof(int), SLJIT_SAVED_REG2, 0);
2169    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_SCRATCH_REG2, 0, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, 1);
2170  JUMPTO(SLJIT_C_NOT_ZERO, loop);  JUMPTO(SLJIT_C_NOT_ZERO, loop);
2171  JUMPHERE(earlyexit);  JUMPHERE(early_quit);
2172    
2173    /* Calculate the return value, which is the maximum ovector value. */
2174    if (topbracket > 1)
2175      {
2176      GET_LOCAL_BASE(SLJIT_SCRATCH_REG1, 0, OVECTOR_START + topbracket * 2 * sizeof(sljit_sw));
2177      OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, topbracket + 1);
2178    
2179      /* OVECTOR(0) is never equal to SLJIT_SAVED_REG3. */
2180      loop = LABEL();
2181      OP1(SLJIT_MOVU, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), -(2 * (sljit_sw)sizeof(sljit_sw)));
2182      OP2(SLJIT_SUB, SLJIT_SCRATCH_REG2, 0, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, 1);
2183      CMPTO(SLJIT_C_EQUAL, SLJIT_SCRATCH_REG3, 0, SLJIT_SAVED_REG3, 0, loop);
2184      OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_SCRATCH_REG2, 0);
2185      }
2186    else
2187      OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, 1);
2188    }
2189    
2190    static SLJIT_INLINE void return_with_partial_match(compiler_common *common, struct sljit_label *quit)
2191    {
2192    DEFINE_COMPILER;
2193    struct sljit_jump *jump;
2194    
2195    SLJIT_COMPILE_ASSERT(STR_END == SLJIT_SAVED_REG2, str_end_must_be_saved_reg2);
2196    SLJIT_ASSERT(common->start_used_ptr != 0 && common->start_ptr != 0
2197      && (common->mode == JIT_PARTIAL_SOFT_COMPILE ? common->hit_start != 0 : common->hit_start == 0));
2198    
2199    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, ARGUMENTS, 0);
2200    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, PCRE_ERROR_PARTIAL);
2201    OP1(SLJIT_MOV_SI, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, real_offset_count));
2202    CMPTO(SLJIT_C_SIG_LESS, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 2, quit);
2203    
2204    /* Store match begin and end. */
2205    OP1(SLJIT_MOV, SLJIT_SAVED_REG1, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, begin));
2206    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, offsets));
2207    
2208    jump = CMP(SLJIT_C_SIG_LESS, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 3);
2209    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);
2210    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2211    OP2(SLJIT_ASHR, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, UCHAR_SHIFT);
2212    #endif
2213    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG2), 2 * sizeof(int), SLJIT_SCRATCH_REG3, 0);
2214    JUMPHERE(jump);
2215    
2216    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);
2217    OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, STR_END, 0, SLJIT_SAVED_REG1, 0);
2218    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2219    OP2(SLJIT_ASHR, SLJIT_SAVED_REG2, 0, SLJIT_SAVED_REG2, 0, SLJIT_IMM, UCHAR_SHIFT);
2220    #endif
2221    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG2), sizeof(int), SLJIT_SAVED_REG2, 0);
2222    
2223    OP2(SLJIT_SUB, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_SAVED_REG1, 0);
2224    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2225    OP2(SLJIT_ASHR, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, UCHAR_SHIFT);
2226    #endif
2227    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG2), 0, SLJIT_SCRATCH_REG3, 0);
2228    
2229    JUMPTO(SLJIT_JUMP, quit);
2230  }  }
2231    
2232  static SLJIT_INLINE BOOL char_has_othercase(compiler_common *common, uschar* cc)  static SLJIT_INLINE void check_start_used_ptr(compiler_common *common)
2233    {
2234    /* May destroy TMP1. */
2235    DEFINE_COMPILER;
2236    struct sljit_jump *jump;
2237    
2238    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2239      {
2240      /* The value of -1 must be kept for start_used_ptr! */
2241      OP2(SLJIT_ADD, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, SLJIT_IMM, 1);
2242      /* Jumps if start_used_ptr < STR_PTR, or start_used_ptr == -1. Although overwriting
2243      is not necessary if start_used_ptr == STR_PTR, it does not hurt as well. */
2244      jump = CMP(SLJIT_C_LESS_EQUAL, TMP1, 0, STR_PTR, 0);
2245      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);
2246      JUMPHERE(jump);
2247      }
2248    else if (common->mode == JIT_PARTIAL_HARD_COMPILE)
2249      {
2250      jump = CMP(SLJIT_C_LESS_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);
2251      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);
2252      JUMPHERE(jump);
2253      }
2254    }
2255    
2256    static SLJIT_INLINE BOOL char_has_othercase(compiler_common *common, pcre_uchar* cc)
2257  {  {
2258  /* Detects if the character has an othercase. */  /* Detects if the character has an othercase. */
2259  unsigned int c;  unsigned int c;
2260    
2261  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2262  if (common->utf8)  if (common->utf)
2263    {    {
2264    GETCHAR(c, cc);    GETCHAR(c, cc);
2265    if (c > 127)    if (c > 127)
# Line 1263  if (common->utf8) Line 2270  if (common->utf8)
2270      return FALSE;      return FALSE;
2271  #endif  #endif
2272      }      }
2273    #ifndef COMPILE_PCRE8
2274      return common->fcc[c] != c;
2275    #endif
2276    }    }
2277  else  else
2278  #endif  #endif
2279    c = *cc;    c = *cc;
2280  return common->fcc[c] != c;  return MAX_255(c) ? common->fcc[c] != c : FALSE;
2281  }  }
2282    
2283  static SLJIT_INLINE unsigned int char_othercase(compiler_common *common, unsigned int c)  static SLJIT_INLINE unsigned int char_othercase(compiler_common *common, unsigned int c)
2284  {  {
2285  /* Returns with the othercase. */  /* Returns with the othercase. */
2286  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2287  if (common->utf8 && c > 127)  if (common->utf && c > 127)
2288    {    {
2289  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2290    return UCD_OTHERCASE(c);    return UCD_OTHERCASE(c);
# Line 1283  if (common->utf8 && c > 127) Line 2293  if (common->utf8 && c > 127)
2293  #endif  #endif
2294    }    }
2295  #endif  #endif
2296  return common->fcc[c];  return TABLE_GET(c, common->fcc, c);
2297  }  }
2298    
2299  static unsigned int char_get_othercase_bit(compiler_common *common, uschar* cc)  static unsigned int char_get_othercase_bit(compiler_common *common, pcre_uchar* cc)
2300  {  {
2301  /* Detects if the character and its othercase has only 1 bit difference. */  /* Detects if the character and its othercase has only 1 bit difference. */
2302  unsigned int c, oc, bit;  unsigned int c, oc, bit;
2303  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2304  int n;  int n;
2305  #endif  #endif
2306    
2307  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2308  if (common->utf8)  if (common->utf)
2309    {    {
2310    GETCHAR(c, cc);    GETCHAR(c, cc);
2311    if (c <= 127)    if (c <= 127)
# Line 1312  if (common->utf8) Line 2322  if (common->utf8)
2322  else  else
2323    {    {
2324    c = *cc;    c = *cc;
2325    oc = common->fcc[c];    oc = TABLE_GET(c, common->fcc, c);
2326    }    }
2327  #else  #else
2328  c = *cc;  c = *cc;
2329  oc = common->fcc[c];  oc = TABLE_GET(c, common->fcc, c);
2330  #endif  #endif
2331    
2332  SLJIT_ASSERT(c != oc);  SLJIT_ASSERT(c != oc);
# Line 1327  if (c <= 127 && bit == 0x20) Line 2337  if (c <= 127 && bit == 0x20)
2337    return (0 << 8) | 0x20;    return (0 << 8) | 0x20;
2338    
2339  /* Since c != oc, they must have at least 1 bit difference. */  /* Since c != oc, they must have at least 1 bit difference. */
2340  if (!ispowerof2(bit))  if (!is_powerof2(bit))
2341    return 0;    return 0;
2342    
2343  #ifdef SUPPORT_UTF8  #if defined COMPILE_PCRE8
2344  if (common->utf8 && c > 127)  
2345    #ifdef SUPPORT_UTF
2346    if (common->utf && c > 127)
2347    {    {
2348    n = _pcre_utf8_table4[*cc & 0x3f];    n = GET_EXTRALEN(*cc);
2349    while ((bit & 0x3f) == 0)    while ((bit & 0x3f) == 0)
2350      {      {
2351      n--;      n--;
# Line 1341  if (common->utf8 && c > 127) Line 2353  if (common->utf8 && c > 127)
2353      }      }
2354    return (n << 8) | bit;    return (n << 8) | bit;
2355    }    }
2356  #endif  #endif /* SUPPORT_UTF */
2357  return (0 << 8) | bit;  return (0 << 8) | bit;
2358    
2359    #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2360    
2361    #ifdef SUPPORT_UTF
2362    if (common->utf && c > 65535)
2363      {
2364      if (bit >= (1 << 10))
2365        bit >>= 10;
2366      else
2367        return (bit < 256) ? ((2 << 8) | bit) : ((3 << 8) | (bit >> 8));
2368      }
2369    #endif /* SUPPORT_UTF */
2370    return (bit < 256) ? ((0 << 8) | bit) : ((1 << 8) | (bit >> 8));
2371    
2372    #endif /* COMPILE_PCRE[8|16|32] */
2373    }
2374    
2375    static void check_partial(compiler_common *common, BOOL force)
2376    {
2377    /* Checks whether a partial matching is occured. Does not modify registers. */
2378    DEFINE_COMPILER;
2379    struct sljit_jump *jump = NULL;
2380    
2381    SLJIT_ASSERT(!force || common->mode != JIT_COMPILE);
2382    
2383    if (common->mode == JIT_COMPILE)
2384      return;
2385    
2386    if (!force)
2387      jump = CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);
2388    else if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2389      jump = CMP(SLJIT_C_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, SLJIT_IMM, -1);
2390    
2391    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2392      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2393    else
2394      {
2395      if (common->partialmatchlabel != NULL)
2396        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2397      else
2398        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2399      }
2400    
2401    if (jump != NULL)
2402      JUMPHERE(jump);
2403    }
2404    
2405    static void check_str_end(compiler_common *common, jump_list **end_reached)
2406    {
2407    /* Does not affect registers. Usually used in a tight spot. */
2408    DEFINE_COMPILER;
2409    struct sljit_jump *jump;
2410    
2411    if (common->mode == JIT_COMPILE)
2412      {
2413      add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2414      return;
2415      }
2416    
2417    jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);
2418    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2419      {
2420      add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));
2421      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2422      add_jump(compiler, end_reached, JUMP(SLJIT_JUMP));
2423      }
2424    else
2425      {
2426      add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));
2427      if (common->partialmatchlabel != NULL)
2428        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2429      else
2430        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2431      }
2432    JUMPHERE(jump);
2433  }  }
2434    
2435  static SLJIT_INLINE void check_input_end(compiler_common *common, jump_list **fallbacks)  static void detect_partial_match(compiler_common *common, jump_list **backtracks)
2436  {  {
2437  DEFINE_COMPILER;  DEFINE_COMPILER;
2438  add_jump(compiler, fallbacks, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));  struct sljit_jump *jump;
2439    
2440    if (common->mode == JIT_COMPILE)
2441      {
2442      add_jump(compiler, backtracks, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2443      return;
2444      }
2445    
2446    /* Partial matching mode. */
2447    jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);
2448    add_jump(compiler, backtracks, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));
2449    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2450      {
2451      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2452      add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));
2453      }
2454    else
2455      {
2456      if (common->partialmatchlabel != NULL)
2457        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2458      else
2459        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2460      }
2461    JUMPHERE(jump);
2462  }  }
2463    
2464  static void read_char(compiler_common *common)  static void read_char(compiler_common *common)
# Line 1356  static void read_char(compiler_common *c Line 2466  static void read_char(compiler_common *c
2466  /* Reads the character into TMP1, updates STR_PTR.  /* Reads the character into TMP1, updates STR_PTR.
2467  Does not check STR_END. TMP2 Destroyed. */  Does not check STR_END. TMP2 Destroyed. */
2468  DEFINE_COMPILER;  DEFINE_COMPILER;
2469  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2470  struct sljit_jump *jump;  struct sljit_jump *jump;
2471  #endif  #endif
2472    
2473  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2474  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2475  if (common->utf8)  if (common->utf)
2476    {    {
2477    /* Should not found a value between 128 and 192 here. */  #if defined COMPILE_PCRE8
2478    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 192);    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2479    add_jump(compiler, &common->utf8readchar, JUMP(SLJIT_FAST_CALL));  #elif defined COMPILE_PCRE16
2480      jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);
2481    #endif /* COMPILE_PCRE[8|16] */
2482      add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2483    JUMPHERE(jump);    JUMPHERE(jump);
2484    }    }
2485  #endif  #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2486  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2487  }  }
2488    
2489  static void peek_char(compiler_common *common)  static void peek_char(compiler_common *common)
# Line 1378  static void peek_char(compiler_common *c Line 2491  static void peek_char(compiler_common *c
2491  /* Reads the character into TMP1, keeps STR_PTR.  /* Reads the character into TMP1, keeps STR_PTR.
2492  Does not check STR_END. TMP2 Destroyed. */  Does not check STR_END. TMP2 Destroyed. */
2493  DEFINE_COMPILER;  DEFINE_COMPILER;
2494  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2495  struct sljit_jump *jump;  struct sljit_jump *jump;
2496  #endif  #endif
2497    
2498  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2499  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2500  if (common->utf8)  if (common->utf)
2501    {    {
2502    /* Should not found a value between 128 and 192 here. */  #if defined COMPILE_PCRE8
2503    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 192);    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2504    add_jump(compiler, &common->utf8readchar, JUMP(SLJIT_FAST_CALL));  #elif defined COMPILE_PCRE16
2505      jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);
2506    #endif /* COMPILE_PCRE[8|16] */
2507      add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2508    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2509    JUMPHERE(jump);    JUMPHERE(jump);
2510    }    }
2511  #endif  #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2512  }  }
2513    
2514  static void read_char8_type(compiler_common *common)  static void read_char8_type(compiler_common *common)
2515  {  {
2516  /* 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. */
2517  DEFINE_COMPILER;  DEFINE_COMPILER;
2518  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2519  struct sljit_jump *jump;  struct sljit_jump *jump;
2520  #endif  #endif
2521    
2522  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2523  if (common->utf8)  if (common->utf)
2524    {    {
2525    OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2526    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2527    #if defined COMPILE_PCRE8
2528    /* This can be an extra read in some situations, but hopefully    /* This can be an extra read in some situations, but hopefully
2529    it is a clever early read in most cases. */    it is needed in most cases. */
2530      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2531      jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2532      add_jump(compiler, &common->utfreadtype8, JUMP(SLJIT_FAST_CALL));
2533      JUMPHERE(jump);
2534    #elif defined COMPILE_PCRE16
2535      OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2536      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);
2537    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
   /* Should not found a value between 128 and 192 here. */  
   jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 192);  
   add_jump(compiler, &common->utf8readtype8, JUMP(SLJIT_FAST_CALL));  
2538    JUMPHERE(jump);    JUMPHERE(jump);
2539      /* Skip low surrogate if necessary. */
2540      OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0xfc00);
2541      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0xd800);
2542      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
2543      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 1);
2544      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2545    #elif defined COMPILE_PCRE32
2546      OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2547      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);
2548      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2549      JUMPHERE(jump);
2550    #endif /* COMPILE_PCRE[8|16|32] */
2551    return;    return;
2552    }    }
2553    #endif /* SUPPORT_UTF */
2554    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2555    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2556    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2557    /* The ctypes array contains only 256 values. */
2558    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2559    jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);
2560    #endif
2561    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2562    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2563    JUMPHERE(jump);
2564  #endif  #endif
 OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);  
 OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), common->ctypes);  
2565  }  }
2566    
2567  static void skip_char_back(compiler_common *common)  static void skip_char_back(compiler_common *common)
2568  {  {
2569  /* Goes one character back. Only affects STR_PTR. Does not check begin. */  /* Goes one character back. Affects STR_PTR and TMP1. Does not check begin. */
2570  DEFINE_COMPILER;  DEFINE_COMPILER;
2571  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2572    #if defined COMPILE_PCRE8
2573  struct sljit_label *label;  struct sljit_label *label;
2574    
2575  if (common->utf8)  if (common->utf)
2576    {    {
2577    label = LABEL();    label = LABEL();
2578    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
2579    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2580    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xc0);    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xc0);
2581    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, 0x80, label);    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, 0x80, label);
2582    return;    return;
2583    }    }
2584  #endif  #elif defined COMPILE_PCRE16
2585  OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);  if (common->utf)
2586      {
2587      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
2588      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2589      /* Skip low surrogate if necessary. */
2590      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
2591      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xdc00);
2592      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
2593      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
2594      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2595      return;
2596      }
2597    #endif /* COMPILE_PCRE[8|16] */
2598    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2599    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2600  }  }
2601    
2602  static void check_newlinechar(compiler_common *common, int nltype, jump_list **fallbacks, BOOL jumpiftrue)  static void check_newlinechar(compiler_common *common, int nltype, jump_list **backtracks, BOOL jumpiftrue)
2603  {  {
2604  /* 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. */
2605  DEFINE_COMPILER;  DEFINE_COMPILER;
# Line 1451  DEFINE_COMPILER; Line 2607  DEFINE_COMPILER;
2607  if (nltype == NLTYPE_ANY)  if (nltype == NLTYPE_ANY)
2608    {    {
2609    add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));
2610    add_jump(compiler, fallbacks, JUMP(jumpiftrue ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));    add_jump(compiler, backtracks, JUMP(jumpiftrue ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));
2611    }    }
2612  else if (nltype == NLTYPE_ANYCRLF)  else if (nltype == NLTYPE_ANYCRLF)
2613    {    {
2614    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_CR);    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_CR);
2615    COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
2616    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_NL);    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_NL);
2617    COND_VALUE(SLJIT_OR | SLJIT_SET_E, TMP2, 0, SLJIT_C_EQUAL);    OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);
2618    add_jump(compiler, fallbacks, JUMP(jumpiftrue ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));    add_jump(compiler, backtracks, JUMP(jumpiftrue ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));
2619    }    }
2620  else  else
2621    {    {
2622    SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline <= 255);    SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline < 256);
2623    add_jump(compiler, fallbacks, CMP(jumpiftrue ? SLJIT_C_EQUAL : SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, common->newline));    add_jump(compiler, backtracks, CMP(jumpiftrue ? SLJIT_C_EQUAL : SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, common->newline));
2624    }    }
2625  }  }
2626    
2627  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2628  static void do_utf8readchar(compiler_common *common)  
2629    #if defined COMPILE_PCRE8
2630    static void do_utfreadchar(compiler_common *common)
2631  {  {
2632  /* Fast decoding an utf8 character. TMP1 contains the first byte  /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2633  of the character (>= 192). Return char value in TMP1, length - 1 in TMP2. */  of the character (>= 0xc0). Return char value in TMP1, length - 1 in TMP2. */
2634  DEFINE_COMPILER;  DEFINE_COMPILER;
2635  struct sljit_jump *jump;  struct sljit_jump *jump;
2636    
2637  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2638  /* Searching for the first zero. */  /* Searching for the first zero. */
2639  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, 0x20);
2640  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_C_NOT_ZERO);
2641  /* 2 byte sequence */  /* Two byte sequence. */
2642  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 1);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2643  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2644  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x1f);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x1f);
2645  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2646  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2647  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2648  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 1);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(1));
 sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  
 JUMPHERE(jump);  
   
 OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x10);  
 jump = JUMP(SLJIT_C_NOT_ZERO);  
 /* 3 byte sequence */  
 OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 1);  
 OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x0f);  
 OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 12);  
 OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  
 OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
 OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 2);  
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 2);  
 OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
 OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 2);  
2649  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2650  JUMPHERE(jump);  JUMPHERE(jump);
2651    
2652  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x08);  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x10);
2653  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_C_NOT_ZERO);
2654  /* 4 byte sequence */  /* Three byte sequence. */
2655  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 1);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2656  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x07);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x0f);
2657  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 18);  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 12);
 OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  
 OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 12);  
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
 OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 2);  
2658  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2659  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2660  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2661  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 3);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2662  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 3);  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2663  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2664  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2665  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 3);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(2));
2666  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2667  JUMPHERE(jump);  JUMPHERE(jump);
2668    
2669  /* 5 byte sequence */  /* Four byte sequence. */
2670  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 1);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2671  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x03);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x07);
2672  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 24);  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 18);
 OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  
 OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 18);  
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
 OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 2);  
2673  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2674  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 12);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 12);
2675  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2676  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 3);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2677  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2678  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2679  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2680  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 4);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(3));
2681  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 4);  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2682  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2683  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2684  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 4);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(3));
2685  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2686  }  }
2687    
2688  static void do_utf8readtype8(compiler_common *common)  static void do_utfreadtype8(compiler_common *common)
2689  {  {
2690  /* Fast decoding an utf8 character type. TMP2 contains the first byte  /* Fast decoding a UTF-8 character type. TMP2 contains the first byte
2691  of the character (>= 192) and TMP1 is destroyed. Return value in TMP1. */  of the character (>= 0xc0). Return value in TMP1. */
2692  DEFINE_COMPILER;  DEFINE_COMPILER;
2693  struct sljit_jump *jump;  struct sljit_jump *jump;
2694  struct sljit_jump *compare;  struct sljit_jump *compare;
2695    
2696  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2697    
2698  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0x20);  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0x20);
2699  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_C_NOT_ZERO);
2700  /* 2 byte sequence */  /* Two byte sequence. */
2701  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2702  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2703  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);
2704  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2705  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
# Line 1581  sljit_emit_fast_return(compiler, RETURN_ Line 2714  sljit_emit_fast_return(compiler, RETURN_
2714  JUMPHERE(jump);  JUMPHERE(jump);
2715    
2716  /* We only have types for characters less than 256. */  /* We only have types for characters less than 256. */
2717  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), (sljit_w)_pcre_utf8_char_sizes);  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
 OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);  
2718  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2719  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2720  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2721  }  }
2722    
2723  #endif  #elif defined COMPILE_PCRE16
2724    
2725    static void do_utfreadchar(compiler_common *common)
2726    {
2727    /* Fast decoding a UTF-16 character. TMP1 contains the first 16 bit char
2728    of the character (>= 0xd800). Return char value in TMP1, length - 1 in TMP2. */
2729    DEFINE_COMPILER;
2730    struct sljit_jump *jump;
2731    
2732    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2733    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xdc00);
2734    /* Do nothing, only return. */
2735    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2736    
2737    JUMPHERE(jump);
2738    /* Combine two 16 bit characters. */
2739    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2740    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2741    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2742    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 10);
2743    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3ff);
2744    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2745    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(1));
2746    OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2747    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2748    }
2749    
2750    #endif /* COMPILE_PCRE[8|16] */
2751    
2752    #endif /* SUPPORT_UTF */
2753    
2754  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2755    
# Line 1604  DEFINE_COMPILER; Line 2765  DEFINE_COMPILER;
2765    
2766  SLJIT_ASSERT(UCD_BLOCK_SIZE == 128 && sizeof(ucd_record) == 8);  SLJIT_ASSERT(UCD_BLOCK_SIZE == 128 && sizeof(ucd_record) == 8);
2767    
2768  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2769  OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);  OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
2770  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_w)_pcre_ucd_stage1);  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(ucd_stage1));
2771  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);
2772  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
2773  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, TMP2, 0);
2774  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, (sljit_w)_pcre_ucd_stage2);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_stage2));
2775  OP1(SLJIT_MOV_UH, TMP2, 0, SLJIT_MEM2(TMP2, TMP1), 1);  OP1(SLJIT_MOV_UH, TMP2, 0, SLJIT_MEM2(TMP2, TMP1), 1);
2776  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, (sljit_w)_pcre_ucd_records + SLJIT_OFFSETOF(ucd_record, chartype));  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_records) + SLJIT_OFFSETOF(ucd_record, chartype));
2777  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM2(TMP1, TMP2), 3);  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM2(TMP1, TMP2), 3);
2778  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2779  }  }
# Line 1626  struct sljit_label *newlinelabel = NULL; Line 2787  struct sljit_label *newlinelabel = NULL;
2787  struct sljit_jump *start;  struct sljit_jump *start;
2788  struct sljit_jump *end = NULL;  struct sljit_jump *end = NULL;
2789  struct sljit_jump *nl = NULL;  struct sljit_jump *nl = NULL;
2790    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2791    struct sljit_jump *singlechar;
2792    #endif
2793  jump_list *newline = NULL;  jump_list *newline = NULL;
2794  BOOL newlinecheck = FALSE;  BOOL newlinecheck = FALSE;
2795  BOOL readbyte = FALSE;  BOOL readuchar = FALSE;
2796    
2797  if (!(hascrorlf || firstline) && (common->nltype == NLTYPE_ANY ||  if (!(hascrorlf || firstline) && (common->nltype == NLTYPE_ANY ||
2798      common->nltype == NLTYPE_ANYCRLF || common->newline > 255))      common->nltype == NLTYPE_ANYCRLF || common->newline > 255))
# Line 1637  if (!(hascrorlf || firstline) && (common Line 2801  if (!(hascrorlf || firstline) && (common
2801  if (firstline)  if (firstline)
2802    {    {
2803    /* Search for the end of the first line. */    /* Search for the end of the first line. */
2804    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, STR_PTR, 0);    SLJIT_ASSERT(common->first_line_end != 0);
2805    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), FIRSTLINE_END, STR_END, 0);    OP1(SLJIT_MOV, TMP3, 0, STR_PTR, 0);
2806    
2807    if (common->nltype == NLTYPE_FIXED && common->newline > 255)    if (common->nltype == NLTYPE_FIXED && common->newline > 255)
2808      {      {
2809      mainloop = LABEL();      mainloop = LABEL();
2810      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2811      end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);      end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
2812      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), -1);      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(-1));
2813      OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);      OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2814      CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, mainloop);      CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, mainloop);
2815      CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, mainloop);      CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, mainloop);
2816      OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_LOCALS_REG), FIRSTLINE_END, STR_PTR, 0, SLJIT_IMM, 1);      JUMPHERE(end);
2817        OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2818      }      }
2819    else    else
2820      {      {
2821      end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);      end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
2822      mainloop = LABEL();      mainloop = LABEL();
2823      /* Continual stores does not cause data dependency. */      /* Continual stores does not cause data dependency. */
2824      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), FIRSTLINE_END, STR_PTR, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);
2825      read_char(common);      read_char(common);
2826      check_newlinechar(common, common->nltype, &newline, TRUE);      check_newlinechar(common, common->nltype, &newline, TRUE);
2827      CMPTO(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0, mainloop);      CMPTO(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0, mainloop);
2828      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), FIRSTLINE_END, STR_PTR, 0);      JUMPHERE(end);
2829        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);
2830      set_jumps(newline, LABEL());      set_jumps(newline, LABEL());
2831      }      }
2832    
2833    JUMPHERE(end);    OP1(SLJIT_MOV, STR_PTR, 0, TMP3, 0);
   OP1(SLJIT_MOV, STR_PTR, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0);  
2834    }    }
2835    
2836  start = JUMP(SLJIT_JUMP);  start = JUMP(SLJIT_JUMP);
# Line 1673  start = JUMP(SLJIT_JUMP); Line 2838  start = JUMP(SLJIT_JUMP);
2838  if (newlinecheck)  if (newlinecheck)
2839    {    {
2840    newlinelabel = LABEL();    newlinelabel = LABEL();
2841    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2842    end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);    end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
2843    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2844    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, common->newline & 0xff);    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, common->newline & 0xff);
2845    COND_VALUE(SLJIT_MOV, TMP1, 0, SLJIT_C_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
2846    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2847      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, UCHAR_SHIFT);
2848    #endif
2849    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2850    nl = JUMP(SLJIT_JUMP);    nl = JUMP(SLJIT_JUMP);
2851    }    }
# Line 1685  if (newlinecheck) Line 2853  if (newlinecheck)
2853  mainloop = LABEL();  mainloop = LABEL();
2854    
2855  /* Increasing the STR_PTR here requires one less jump in the most common case. */  /* Increasing the STR_PTR here requires one less jump in the most common case. */
2856  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2857  if (common->utf8) readbyte = TRUE;  if (common->utf) readuchar = TRUE;
2858  #endif  #endif
2859  if (newlinecheck) readbyte = TRUE;  if (newlinecheck) readuchar = TRUE;
2860    
2861  if (readbyte)  if (readuchar)
2862    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2863    
2864  if (newlinecheck)  if (newlinecheck)
2865    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, newlinelabel);    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, newlinelabel);
2866    
2867  #ifdef SUPPORT_UTF8  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2868  if (common->utf8)  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2869    #if defined COMPILE_PCRE8
2870    if (common->utf)
2871    {    {
2872    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_w)_pcre_utf8_char_sizes);    singlechar = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2873      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2874    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2875      JUMPHERE(singlechar);
2876    }    }
2877  else  #elif defined COMPILE_PCRE16
2878    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);  if (common->utf)
2879  #else    {
2880  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);    singlechar = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);
2881  #endif    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
2882      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2883      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
2884      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
2885      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2886      JUMPHERE(singlechar);
2887      }
2888    #endif /* COMPILE_PCRE[8|16] */
2889    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2890  JUMPHERE(start);  JUMPHERE(start);
2891    
2892  if (newlinecheck)  if (newlinecheck)
# Line 1718  if (newlinecheck) Line 2898  if (newlinecheck)
2898  return mainloop;  return mainloop;
2899  }  }
2900    
2901  static SLJIT_INLINE void fast_forward_first_byte(compiler_common *common, pcre_uint16 firstbyte, BOOL firstline)  #define MAX_N_CHARS 3
2902    
2903    static SLJIT_INLINE BOOL fast_forward_first_n_chars(compiler_common *common, BOOL firstline)
2904    {
2905    DEFINE_COMPILER;
2906    struct sljit_label *start;
2907    struct sljit_jump *quit;
2908    pcre_uint32 chars[MAX_N_CHARS * 2];
2909    pcre_uchar *cc = common->start + 1 + LINK_SIZE;
2910    int location = 0;
2911    pcre_int32 len, c, bit, caseless;
2912    int must_stop;
2913    
2914    /* We do not support alternatives now. */
2915    if (*(common->start + GET(common->start, 1)) == OP_ALT)
2916      return FALSE;
2917    
2918    while (TRUE)
2919      {
2920      caseless = 0;
2921      must_stop = 1;
2922      switch(*cc)
2923        {
2924        case OP_CHAR:
2925        must_stop = 0;
2926        cc++;
2927        break;
2928    
2929        case OP_CHARI:
2930        caseless = 1;
2931        must_stop = 0;
2932        cc++;
2933        break;
2934    
2935        case OP_SOD:
2936        case OP_SOM:
2937        case OP_SET_SOM:
2938        case OP_NOT_WORD_BOUNDARY:
2939        case OP_WORD_BOUNDARY:
2940        case OP_EODN:
2941        case OP_EOD:
2942        case OP_CIRC:
2943        case OP_CIRCM:
2944        case OP_DOLL:
2945        case OP_DOLLM:
2946        /* Zero width assertions. */
2947        cc++;
2948        continue;
2949    
2950        case OP_PLUS:
2951        case OP_MINPLUS:
2952        case OP_POSPLUS:
2953        cc++;
2954        break;
2955    
2956        case OP_EXACT:
2957        cc += 1 + IMM2_SIZE;
2958        break;
2959    
2960        case OP_PLUSI:
2961        case OP_MINPLUSI:
2962        case OP_POSPLUSI:
2963        caseless = 1;
2964        cc++;
2965        break;
2966    
2967        case OP_EXACTI:
2968        caseless = 1;
2969        cc += 1 + IMM2_SIZE;
2970        break;
2971    
2972        default:
2973        must_stop = 2;
2974        break;
2975        }
2976    
2977      if (must_stop == 2)
2978          break;
2979    
2980      len = 1;
2981    #ifdef SUPPORT_UTF
2982      if (common->utf && HAS_EXTRALEN(cc[0])) len += GET_EXTRALEN(cc[0]);
2983    #endif
2984    
2985      if (caseless && char_has_othercase(common, cc))
2986        {
2987        caseless = char_get_othercase_bit(common, cc);
2988        if (caseless == 0)
2989          return FALSE;
2990    #ifdef COMPILE_PCRE8
2991        caseless = ((caseless & 0xff) << 8) | (len - (caseless >> 8));
2992    #else
2993        if ((caseless & 0x100) != 0)
2994          caseless = ((caseless & 0xff) << 16) | (len - (caseless >> 9));
2995        else
2996          caseless = ((caseless & 0xff) << 8) | (len - (caseless >> 9));
2997    #endif
2998        }
2999      else
3000        caseless = 0;
3001    
3002      while (len > 0 && location < MAX_N_CHARS * 2)
3003        {
3004        c = *cc;
3005        bit = 0;
3006        if (len == (caseless & 0xff))
3007          {
3008          bit = caseless >> 8;
3009          c |= bit;
3010          }
3011    
3012        chars[location] = c;
3013        chars[location + 1] = bit;
3014    
3015        len--;
3016        location += 2;
3017        cc++;
3018        }
3019    
3020      if (location >= MAX_N_CHARS * 2 || must_stop != 0)
3021        break;
3022      }
3023    
3024    /* At least two characters are required. */
3025    if (location < 2 * 2)
3026        return FALSE;
3027    
3028    if (firstline)
3029      {
3030      SLJIT_ASSERT(common->first_line_end != 0);
3031      OP1(SLJIT_MOV, TMP3, 0, STR_END, 0);
3032      OP2(SLJIT_SUB, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, SLJIT_IMM, IN_UCHARS((location >> 1) - 1));
3033      }
3034    else
3035      OP2(SLJIT_SUB, STR_END, 0, STR_END, 0, SLJIT_IMM, IN_UCHARS((location >> 1) - 1));
3036    
3037    start = LABEL();
3038    quit = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3039    
3040    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
3041    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
3042    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3043    if (chars[1] != 0)
3044      OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, chars[1]);
3045    CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, chars[0], start);
3046    if (location > 2 * 2)
3047      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
3048    if (chars[3] != 0)
3049      OP2(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_IMM, chars[3]);
3050    CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, chars[2], start);
3051    if (location > 2 * 2)
3052      {
3053      if (chars[5] != 0)
3054        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, chars[5]);
3055      CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, chars[4], start);
3056      }
3057    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3058    
3059    JUMPHERE(quit);
3060    
3061    if (firstline)
3062      OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);
3063    else
3064      OP2(SLJIT_ADD, STR_END, 0, STR_END, 0, SLJIT_IMM, IN_UCHARS((location >> 1) - 1));
3065    return TRUE;
3066    }
3067    
3068    #undef MAX_N_CHARS
3069    
3070    static SLJIT_INLINE void fast_forward_first_char(compiler_common *common, pcre_uchar first_char, BOOL caseless, BOOL firstline)
3071  {  {
3072  DEFINE_COMPILER;  DEFINE_COMPILER;
3073  struct sljit_label *start;  struct sljit_label *start;
3074  struct sljit_jump *leave;  struct sljit_jump *quit;
3075  struct sljit_jump *found;  struct sljit_jump *found;
3076  pcre_uint16 oc, bit;  pcre_uchar oc, bit;
3077    
3078  if (firstline)  if (firstline)
3079    {    {
3080    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), POSSESSIVE0, STR_END, 0);    SLJIT_ASSERT(common->first_line_end != 0);
3081    OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), FIRSTLINE_END);    OP1(SLJIT_MOV, TMP3, 0, STR_END, 0);
3082      OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end);
3083    }    }
3084    
3085  start = LABEL();  start = LABEL();
3086  leave = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);  quit = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3087  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3088    
3089  if ((firstbyte & REQ_CASELESS) == 0)  oc = first_char;
3090    found = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, firstbyte & 0xff);  if (caseless)
3091      {
3092      oc = TABLE_GET(first_char, common->fcc, first_char);
3093    #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8)
3094      if (first_char > 127 && common->utf)
3095        oc = UCD_OTHERCASE(first_char);
3096    #endif
3097      }
3098    if (first_char == oc)
3099      found = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, first_char);
3100  else  else
3101    {    {
3102    firstbyte &= 0xff;    bit = first_char ^ oc;
3103    oc = common->fcc[firstbyte];    if (is_powerof2(bit))
   bit = firstbyte ^ oc;  
   if (ispowerof2(bit))  
3104      {      {
3105      OP2(SLJIT_OR, TMP2, 0, TMP1, 0, SLJIT_IMM, bit);      OP2(SLJIT_OR, TMP2, 0, TMP1, 0, SLJIT_IMM, bit);
3106      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, firstbyte | bit);      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, first_char | bit);
3107      }      }
3108    else    else
3109      {      {
3110      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, firstbyte);      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, first_char);
3111      COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_EQUAL);      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
3112      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, oc);      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, oc);
3113      COND_VALUE(SLJIT_OR | SLJIT_SET_E, TMP2, 0, SLJIT_C_EQUAL);      OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);
3114      found = JUMP(SLJIT_C_NOT_ZERO);      found = JUMP(SLJIT_C_NOT_ZERO);
3115      }      }
3116    }    }
3117    
3118  #ifdef SUPPORT_UTF8  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
 if (common->utf8)  
   {  
   OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_w)_pcre_utf8_char_sizes);  
   OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);  
   }  
 else  
   OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);  
 #else  
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);  
 #endif  
3119  JUMPTO(SLJIT_JUMP, start);  JUMPTO(SLJIT_JUMP, start);
3120  JUMPHERE(found);  JUMPHERE(found);
3121  JUMPHERE(leave);  JUMPHERE(quit);
3122    
3123  if (firstline)  if (firstline)
3124    OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), POSSESSIVE0);    OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);
3125  }  }
3126    
3127  static SLJIT_INLINE void fast_forward_newline(compiler_common *common, BOOL firstline)  static SLJIT_INLINE void fast_forward_newline(compiler_common *common, BOOL firstline)
# Line 1783  DEFINE_COMPILER; Line 3130  DEFINE_COMPILER;
3130  struct sljit_label *loop;  struct sljit_label *loop;
3131  struct sljit_jump *lastchar;  struct sljit_jump *lastchar;
3132  struct sljit_jump *firstchar;  struct sljit_jump *firstchar;
3133  struct sljit_jump *leave;  struct sljit_jump *quit;
3134  struct sljit_jump *foundcr = NULL;  struct sljit_jump *foundcr = NULL;
3135  struct sljit_jump *notfoundnl;  struct sljit_jump *notfoundnl;
3136  jump_list *newline = NULL;  jump_list *newline = NULL;
3137    
3138  if (firstline)  if (firstline)
3139    {    {
3140    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), POSSESSIVE0, STR_END, 0);    SLJIT_ASSERT(common->first_line_end != 0);
3141    OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), FIRSTLINE_END);    OP1(SLJIT_MOV, TMP3, 0, STR_END, 0);
3142      OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end);
3143    }    }
3144    
3145  if (common->nltype == NLTYPE_FIXED && common->newline > 255)  if (common->nltype == NLTYPE_FIXED && common->newline > 255)
# Line 1802  if (common->nltype == NLTYPE_FIXED && co Line 3150  if (common->nltype == NLTYPE_FIXED && co
3150    OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(TMP1), SLJIT_OFFSETOF(jit_arguments, begin));    OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(TMP1), SLJIT_OFFSETOF(jit_arguments, begin));
3151    firstchar = CMP(SLJIT_C_LESS_EQUAL, STR_PTR, 0, TMP2, 0);    firstchar = CMP(SLJIT_C_LESS_EQUAL, STR_PTR, 0, TMP2, 0);
3152    
3153    OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 2);    OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, IN_UCHARS(2));
3154    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, STR_PTR, 0, TMP1, 0);
3155    COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_GREATER_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_GREATER_EQUAL);
3156    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
3157      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCHAR_SHIFT);
3158    #endif
3159    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
3160    
3161    loop = LABEL();    loop = LABEL();
3162    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3163    leave = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);    quit = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3164    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), -2);    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(-2));
3165    OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), -1);    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(-1));
3166    CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, loop);    CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, loop);
3167    CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, loop);    CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, loop);
3168    
3169    JUMPHERE(leave);    JUMPHERE(quit);
3170    JUMPHERE(firstchar);    JUMPHERE(firstchar);
3171    JUMPHERE(lastchar);    JUMPHERE(lastchar);
3172    
# Line 1839  set_jumps(newline, loop); Line 3190  set_jumps(newline, loop);
3190    
3191  if (common->nltype == NLTYPE_ANY || common->nltype == NLTYPE_ANYCRLF)  if (common->nltype == NLTYPE_ANY || common->nltype == NLTYPE_ANYCRLF)
3192    {    {
3193    leave = JUMP(SLJIT_JUMP);    quit = JUMP(SLJIT_JUMP);
3194    JUMPHERE(foundcr);    JUMPHERE(foundcr);
3195    notfoundnl = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);    notfoundnl = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3196    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3197    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_NL);    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_NL);
3198    COND_VALUE(SLJIT_MOV, TMP1, 0, SLJIT_C_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
3199    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
3200      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, UCHAR_SHIFT);
3201    #endif
3202    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3203    JUMPHERE(notfoundnl);    JUMPHERE(notfoundnl);
3204    JUMPHERE(leave);    JUMPHERE(quit);
3205    }    }
3206  JUMPHERE(lastchar);  JUMPHERE(lastchar);
3207  JUMPHERE(firstchar);  JUMPHERE(firstchar);
3208    
3209  if (firstline)  if (firstline)
3210    OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), POSSESSIVE0);    OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);
3211  }  }
3212    
3213    static BOOL check_class_ranges(compiler_common *common, const pcre_uint8 *bits, BOOL nclass, jump_list **backtracks);
3214    
3215  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, sljit_uw start_bits, BOOL firstline)
3216  {  {
3217  DEFINE_COMPILER;  DEFINE_COMPILER;
3218  struct sljit_label *start;  struct sljit_label *start;
3219  struct sljit_jump *leave;  struct sljit_jump *quit;
3220  struct sljit_jump *found;  struct sljit_jump *found = NULL;
3221    jump_list *matches = NULL;
3222    pcre_uint8 inverted_start_bits[32];
3223    int i;
3224    #ifndef COMPILE_PCRE8
3225    struct sljit_jump *jump;
3226    #endif
3227    
3228    for (i = 0; i < 32; ++i)
3229      inverted_start_bits[i] = ~(((pcre_uint8*)start_bits)[i]);
3230    
3231  if (firstline)  if (firstline)
3232    {    {
3233    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), POSSESSIVE0, STR_END, 0);    SLJIT_ASSERT(common->first_line_end != 0);
3234    OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), FIRSTLINE_END);    OP1(SLJIT_MOV, RETURN_ADDR, 0, STR_END, 0);
3235      OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end);
3236    }    }
3237    
3238  start = LABEL();  start = LABEL();
3239  leave = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);  quit = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3240  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3241  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3242  if (common->utf8)  if (common->utf)
3243    OP1(SLJIT_MOV_UB, TMP3, 0, SLJIT_MEM1(TMP1), (sljit_w)_pcre_utf8_char_sizes);    OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);
3244  #endif  #endif
3245  OP2(SLJIT_AND, TMP2, 0, TMP1, 0, SLJIT_IMM, 0x7);  
3246  OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 3);  if (!check_class_ranges(common, inverted_start_bits, (inverted_start_bits[31] & 0x80) != 0, &matches))
3247  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), start_bits);    {
3248  OP2(SLJIT_SHL, TMP2, 0, SLJIT_IMM, 1, TMP2, 0);  #ifndef COMPILE_PCRE8
3249  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 255);
3250  found = JUMP(SLJIT_C_NOT_ZERO);    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 255);
3251      JUMPHERE(jump);
3252  #ifdef SUPPORT_UTF8  #endif
3253  if (common->utf8)    OP2(SLJIT_AND, TMP2, 0, TMP1, 0, SLJIT_IMM, 0x7);
3254    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP3, 0);    OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 3);
3255  else    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), start_bits);
3256    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);    OP2(SLJIT_SHL, TMP2, 0, SLJIT_IMM, 1, TMP2, 0);
3257  #else    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);
3258  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);    found = JUMP(SLJIT_C_NOT_ZERO);
3259      }
3260    
3261    #ifdef SUPPORT_UTF
3262    if (common->utf)
3263      OP1(SLJIT_MOV, TMP1, 0, TMP3, 0);
3264  #endif  #endif
3265    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3266    #ifdef SUPPORT_UTF
3267    #if defined COMPILE_PCRE8
3268    if (common->utf)
3269      {
3270      CMPTO(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0, start);
3271      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
3272      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3273      }
3274    #elif defined COMPILE_PCRE16
3275    if (common->utf)
3276      {
3277      CMPTO(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800, start);
3278      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
3279      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);
3280      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
3281      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
3282      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3283      }
3284    #endif /* COMPILE_PCRE[8|16] */
3285    #endif /* SUPPORT_UTF */
3286  JUMPTO(SLJIT_JUMP, start);  JUMPTO(SLJIT_JUMP, start);
3287  JUMPHERE(found);  if (found != NULL)
3288  JUMPHERE(leave);    JUMPHERE(found);
3289    if (matches != NULL)
3290      set_jumps(matches, LABEL());
3291    JUMPHERE(quit);
3292    
3293  if (firstline)  if (firstline)
3294    OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), POSSESSIVE0);    OP1(SLJIT_MOV, STR_END, 0, RETURN_ADDR, 0);
3295  }  }
3296    
3297  static SLJIT_INLINE struct sljit_jump *search_requested_char(compiler_common *common, pcre_uint16 reqbyte, BOOL has_firstbyte)  static SLJIT_INLINE struct sljit_jump *search_requested_char(compiler_common *common, pcre_uchar req_char, BOOL caseless, BOOL has_firstchar)
3298  {  {
3299  DEFINE_COMPILER;  DEFINE_COMPILER;
3300  struct sljit_label *loop;  struct sljit_label *loop;
# Line 1908  struct sljit_jump *alreadyfound; Line 3303  struct sljit_jump *alreadyfound;
3303  struct sljit_jump *found;  struct sljit_jump *found;
3304  struct sljit_jump *foundoc = NULL;  struct sljit_jump *foundoc = NULL;
3305  struct sljit_jump *notfound;  struct sljit_jump *notfound;
3306  pcre_uint16 oc, bit;  pcre_uint32 oc, bit;
3307    
3308  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), REQ_BYTE_PTR);  SLJIT_ASSERT(common->req_char_ptr != 0);
3309    OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->req_char_ptr);
3310  OP2(SLJIT_ADD, TMP1, 0, STR_PTR, 0, SLJIT_IMM, REQ_BYTE_MAX);  OP2(SLJIT_ADD, TMP1, 0, STR_PTR, 0, SLJIT_IMM, REQ_BYTE_MAX);
3311  toolong = CMP(SLJIT_C_LESS, TMP1, 0, STR_END, 0);  toolong = CMP(SLJIT_C_LESS, TMP1, 0, STR_END, 0);
3312  alreadyfound = CMP(SLJIT_C_LESS, STR_PTR, 0, TMP2, 0);  alreadyfound = CMP(SLJIT_C_LESS, STR_PTR, 0, TMP2, 0);
3313    
3314  if (has_firstbyte)  if (has_firstchar)
3315    OP2(SLJIT_ADD, TMP1, 0, STR_PTR, 0, SLJIT_IMM, 1);    OP2(SLJIT_ADD, TMP1, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3316  else  else
3317    OP1(SLJIT_MOV, TMP1, 0, STR_PTR, 0);    OP1(SLJIT_MOV, TMP1, 0, STR_PTR, 0);
3318    
3319  loop = LABEL();  loop = LABEL();
3320  notfound = CMP(SLJIT_C_GREATER_EQUAL, TMP1, 0, STR_END, 0);  notfound = CMP(SLJIT_C_GREATER_EQUAL, TMP1, 0, STR_END, 0);
3321    
3322  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP1), 0);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(TMP1), 0);
3323  if ((reqbyte & REQ_CASELESS) == 0)  oc = req_char;
3324    found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, reqbyte & 0xff);  if (caseless)
3325      {
3326      oc = TABLE_GET(req_char, common->fcc, req_char);
3327    #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8)
3328      if (req_char > 127 && common->utf)
3329        oc = UCD_OTHERCASE(req_char);
3330    #endif
3331      }
3332    if (req_char == oc)
3333      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, req_char);
3334  else  else
3335    {    {
3336    reqbyte &= 0xff;    bit = req_char ^ oc;
3337    oc = common->fcc[reqbyte];    if (is_powerof2(bit))
   bit = reqbyte ^ oc;  
   if (ispowerof2(bit))  
3338      {      {
3339      OP2(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_IMM, bit);      OP2(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_IMM, bit);
3340      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, reqbyte | bit);      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, req_char | bit);
3341      }      }
3342    else    else
3343      {      {
3344      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, reqbyte);      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, req_char);
3345      foundoc = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, oc);      foundoc = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, oc);
3346      }      }
3347    }    }
3348  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, IN_UCHARS(1));
3349  JUMPTO(SLJIT_JUMP, loop);  JUMPTO(SLJIT_JUMP, loop);
3350    
3351  JUMPHERE(found);  JUMPHERE(found);
3352  if (foundoc)  if (foundoc)
3353    JUMPHERE(foundoc);    JUMPHERE(foundoc);
3354  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), REQ_BYTE_PTR, TMP1, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->req_char_ptr, TMP1, 0);
3355  JUMPHERE(alreadyfound);  JUMPHERE(alreadyfound);
3356  JUMPHERE(toolong);  JUMPHERE(toolong);
3357  return notfound;  return notfound;
# Line 1957  return notfound; Line 3360  return notfound;
3360  static void do_revertframes(compiler_common *common)  static void do_revertframes(compiler_common *common)
3361  {  {
3362  DEFINE_COMPILER;  DEFINE_COMPILER;
 struct sljit_jump *earlyexit;  
3363  struct sljit_jump *jump;  struct sljit_jump *jump;
3364  struct sljit_label *mainloop;  struct sljit_label *mainloop;
3365    
3366  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
3367  OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS_HEAD);  OP1(SLJIT_MOV, TMP1, 0, STACK_TOP, 0);
3368    GET_LOCAL_BASE(TMP3, 0, 0);
3369    
3370  /* Drop frames until we reach STACK_TOP. */  /* Drop frames until we reach STACK_TOP. */
 earlyexit = CMP(SLJIT_C_LESS, TMP1, 0, STACK_TOP, 0);  
3371  mainloop = LABEL();  mainloop = LABEL();
3372  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(TMP1), 0);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(TMP1), 0);
3373  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);
3374  OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_LOCALS_REG, 0);  jump = JUMP(SLJIT_C_SIG_LESS_EQUAL);
3375  OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), 0, SLJIT_MEM1(TMP1), sizeof(sljit_w));  
3376  OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), sizeof(sljit_w), SLJIT_MEM1(TMP1), 2 * sizeof(sljit_w));  OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, TMP3, 0);
3377  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 3 * sizeof(sljit_w));  OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), 0, SLJIT_MEM1(TMP1), sizeof(sljit_sw));
3378    OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), sizeof(sljit_sw), SLJIT_MEM1(TMP1), 2 * sizeof(sljit_sw));
3379    OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 3 * sizeof(sljit_sw));
3380  JUMPTO(SLJIT_JUMP, mainloop);  JUMPTO(SLJIT_JUMP, mainloop);
3381    
3382  JUMPHERE(jump);  JUMPHERE(jump);
3383  jump = CMP(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, frame_end);  jump = JUMP(SLJIT_C_SIG_LESS);
3384  /* End of dropping frames. */  /* End of dropping frames. */
 OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(TMP1), sizeof(sljit_w));  
 OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS_HEAD, TMP1, 0);  
 CMPTO(SLJIT_C_GREATER_EQUAL, TMP1, 0, STACK_TOP, 0, mainloop);  
 JUMPHERE(earlyexit);  
3385  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3386    
3387  JUMPHERE(jump);  JUMPHERE(jump);
3388  jump = CMP(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, frame_setmaxindex);  OP1(SLJIT_NEG, TMP2, 0, TMP2, 0);
3389  /* Set max index. */  OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, TMP3, 0);
3390  OP1(SLJIT_MOV, MAX_INDEX, 0, SLJIT_MEM1(TMP1), sizeof(sljit_w));  OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), 0, SLJIT_MEM1(TMP1), sizeof(sljit_sw));
3391  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 2 * sizeof(sljit_w));  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 2 * sizeof(sljit_sw));
 JUMPTO(SLJIT_JUMP, mainloop);  
   
 JUMPHERE(jump);  
 jump = CMP(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, frame_setstrbegin);  
 /* Set max index. */  
 OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(TMP1), sizeof(sljit_w));  
 OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 2 * sizeof(sljit_w));  
 OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0), TMP2, 0);  
 JUMPTO(SLJIT_JUMP, mainloop);  
   
 JUMPHERE(jump);  
 /* Unknown command. */  
 OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 2 * sizeof(sljit_w));  
3392  JUMPTO(SLJIT_JUMP, mainloop);  JUMPTO(SLJIT_JUMP, mainloop);
3393  }  }
3394    
3395  static void check_wordboundary(compiler_common *common)  static void check_wordboundary(compiler_common *common)
3396  {  {
3397  DEFINE_COMPILER;  DEFINE_COMPILER;
3398  struct sljit_jump *beginend;  struct sljit_jump *skipread;
3399    jump_list *skipread_list = NULL;
3400    #if !(defined COMPILE_PCRE8) || defined SUPPORT_UTF
3401  struct sljit_jump *jump;  struct sljit_jump *jump;
3402    #endif
3403    
3404  SLJIT_ASSERT(ctype_word == 0x10);  SLJIT_COMPILE_ASSERT(ctype_word == 0x10, ctype_word_must_be_16);
3405    
3406  sljit_emit_fast_enter(compiler, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, 1, 5, 5, common->localsize);  sljit_emit_fast_enter(compiler, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0);
3407  /* Get type of the previous char, and put it to LOCALS1. */  /* Get type of the previous char, and put it to LOCALS1. */
3408  OP1(SLJIT_MOV, TMP1, 0, ARGUMENTS, 0);  OP1(SLJIT_MOV, TMP1, 0, ARGUMENTS, 0);
3409  OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(TMP1), SLJIT_OFFSETOF(jit_arguments, begin));  OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(TMP1), SLJIT_OFFSETOF(jit_arguments, begin));
3410  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, SLJIT_IMM, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, SLJIT_IMM, 0);
3411  beginend = CMP(SLJIT_C_LESS_EQUAL, STR_PTR, 0, TMP1, 0);  skipread = CMP(SLJIT_C_LESS_EQUAL, STR_PTR, 0, TMP1, 0);
3412  skip_char_back(common);  skip_char_back(common);
3413    check_start_used_ptr(common);
3414  read_char(common);  read_char(common);
3415    
3416  /* Testing char type. */  /* Testing char type. */
3417  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3418  if (common->useucp)  if (common->use_ucp)
3419    {    {
3420    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 1);    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 1);
3421    jump = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_UNDERSCORE);    jump = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_UNDERSCORE);
3422    add_jump(compiler, &common->getucd, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->getucd, JUMP(SLJIT_FAST_CALL));
3423    OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Ll);    OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Ll);
3424    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_Lu - ucp_Ll);    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_Lu - ucp_Ll);
3425    COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_LESS_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_LESS_EQUAL);
3426    OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Nd - ucp_Ll);    OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Nd - ucp_Ll);
3427    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_No - ucp_Nd);    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_No - ucp_Nd);
3428    COND_VALUE(SLJIT_OR, TMP2, 0, SLJIT_C_LESS_EQUAL);    OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_C_LESS_EQUAL);
3429    JUMPHERE(jump);    JUMPHERE(jump);
3430    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP2, 0);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP2, 0);
3431    }    }
3432  else  else
3433  #endif  #endif
3434    {    {
3435  #ifdef SUPPORT_UTF8  #ifndef COMPILE_PCRE8
3436      jump = CMP(SLJIT_C_GREATER, TMP1, 0, SLJIT_IMM, 255);
3437    #elif defined SUPPORT_UTF
3438    /* Here LOCALS1 has already been zeroed. */    /* Here LOCALS1 has already been zeroed. */
3439    jump = NULL;    jump = NULL;
3440    if (common->utf8)    if (common->utf)
3441      jump = CMP(SLJIT_C_GREATER, TMP1, 0, SLJIT_IMM, 255);      jump = CMP(SLJIT_C_GREATER, TMP1, 0, SLJIT_IMM, 255);
3442  #endif  #endif /* COMPILE_PCRE8 */
3443    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), common->ctypes);    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), common->ctypes);
3444    OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 4 /* ctype_word */);    OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 4 /* ctype_word */);
3445    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
3446    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);
3447  #ifdef SUPPORT_UTF8  #ifndef COMPILE_PCRE8
3448      JUMPHERE(jump);
3449    #elif defined SUPPORT_UTF
3450    if (jump != NULL)    if (jump != NULL)
3451      JUMPHERE(jump);      JUMPHERE(jump);
3452    #endif /* COMPILE_PCRE8 */
3453      }
3454    JUMPHERE(skipread);
3455    
3456    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 0);
3457    check_str_end(common, &skipread_list);
3458    peek_char(common);
3459    
3460    /* Testing char type. This is a code duplication. */
3461    #ifdef SUPPORT_UCP
3462    if (common->use_ucp)
3463      {
3464      OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 1);
3465      jump = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_UNDERSCORE);
3466      add_jump(compiler, &common->getucd, JUMP(SLJIT_FAST_CALL));
3467      OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Ll);
3468      OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_Lu - ucp_Ll);
3469      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_LESS_EQUAL);
3470      OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Nd - ucp_Ll);
3471      OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_No - ucp_Nd);
3472      OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_C_LESS_EQUAL);
3473      JUMPHERE(jump);
3474      }
3475    else
3476    #endif
3477      {
3478    #ifndef COMPILE_PCRE8
3479      /* TMP2 may be destroyed by peek_char. */
3480      OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 0);
3481      jump = CMP(SLJIT_C_GREATER, TMP1, 0, SLJIT_IMM, 255);
3482    #elif defined SUPPORT_UTF
3483      OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 0);
3484      jump = NULL;
3485      if (common->utf)
3486        jump = CMP(SLJIT_C_GREATER, TMP1, 0, SLJIT_IMM, 255);
3487  #endif  #endif
3488      OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP1), common->ctypes);
3489      OP2(SLJIT_LSHR, TMP2, 0, TMP2, 0, SLJIT_IMM, 4 /* ctype_word */);
3490      OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 1);
3491    #ifndef COMPILE_PCRE8
3492      JUMPHERE(jump);
3493    #elif defined SUPPORT_UTF
3494      if (jump != NULL)
3495        JUMPHERE(jump);
3496    #endif /* COMPILE_PCRE8 */
3497      }
3498    set_jumps(skipread_list, LABEL());
3499    
3500    OP2(SLJIT_XOR | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1);
3501    sljit_emit_fast_return(compiler, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0);
3502    }
3503    
3504    /*
3505      range format:
3506    
3507      ranges[0] = length of the range (max MAX_RANGE_SIZE, -1 means invalid range).
3508      ranges[1] = first bit (0 or 1)
3509      ranges[2-length] = position of the bit change (when the current bit is not equal to the previous)
3510    */
3511