/[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 752 by zherczeg, Sat Nov 19 15:28:29 2011 UTC revision 1310 by zherczeg, Sat Apr 6 06:51:09 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) (pcre_malloc)(size)  #define SLJIT_MALLOC(size) (PUBL(malloc))(size)
56  #define SLJIT_FREE(ptr) (pcre_free)(ptr)  #define SLJIT_FREE(ptr) (PUBL(free))(ptr)
57  #define SLJIT_CONFIG_AUTO 1  #define SLJIT_CONFIG_AUTO 1
58  #define SLJIT_CONFIG_STATIC 1  #define SLJIT_CONFIG_STATIC 1
59  #define SLJIT_VERBOSE 0  #define SLJIT_VERBOSE 0
# Line 62  system files. */ Line 62  system files. */
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 82  The code generator follows the recursive Line 94  The code generator follows the recursive
94  expressions. The basic blocks of regular expressions are condition checkers  expressions. The basic blocks of regular expressions are condition checkers
95  whose execute different commands depending on the result of the condition check.  whose execute different commands depending on the result of the condition check.
96  The relationship between the operators can be horizontal (concatenation) and  The relationship between the operators can be horizontal (concatenation) and
97  vertical (sub-expression) (See struct fallback_common for more details).  vertical (sub-expression) (See struct backtrack_common for more details).
98    
99    'ab' - 'a' and 'b' regexps are concatenated    'ab' - 'a' and 'b' regexps are concatenated
100    'a+' - 'a' is the sub-expression of the '+' operator    'a+' - 'a' is the sub-expression of the '+' operator
101    
102  The condition checkers are boolean (true/false) checkers. Machine code is generated  The condition checkers are boolean (true/false) checkers. Machine code is generated
103  for the checker itself and for the actions depending on the result of the checker.  for the checker itself and for the actions depending on the result of the checker.
104  The 'true' case is called as the hot path (expected path), and the other is called as  The 'true' case is called as the matching path (expected path), and the other is called as
105  the 'fallback' path. Branch instructions are expesive for all CPUs, so we avoid taken  the 'backtrack' path. Branch instructions are expesive for all CPUs, so we avoid taken
106  branches on the hot path.  branches on the matching path.
107    
108   Greedy star operator (*) :   Greedy star operator (*) :
109     Hot path: match happens.     Matching path: match happens.
110     Fallback path: match failed.     Backtrack path: match failed.
111   Non-greedy star operator (*?) :   Non-greedy star operator (*?) :
112     Hot path: no need to perform a match.     Matching path: no need to perform a match.
113     Fallback path: match is required.     Backtrack path: match is required.
114    
115  The following example shows how the code generated for a capturing bracket  The following example shows how the code generated for a capturing bracket
116  with two alternatives. Let A, B, C, D are arbirary regular expressions, and  with two alternatives. Let A, B, C, D are arbirary regular expressions, and
# Line 108  we have the following regular expression Line 120  we have the following regular expression
120    
121  The generated code will be the following:  The generated code will be the following:
122    
123   A hot path   A matching path
124   '(' hot path (pushing arguments to the stack)   '(' matching path (pushing arguments to the stack)
125   B hot path   B matching path
126   ')' hot path (pushing arguments to the stack)   ')' matching path (pushing arguments to the stack)
127   D hot path   D matching path
128   return with successful match   return with successful match
129    
130   D fallback path   D backtrack path
131   ')' fallback path (If we arrived from "C" jump to the fallback of "C")   ')' backtrack path (If we arrived from "C" jump to the backtrack of "C")
132   B fallback path   B backtrack path
133   C expected path   C expected path
134   jump to D hot path   jump to D matching path
135   C fallback path   C backtrack path
136   A fallback path   A backtrack path
137    
138   Notice, that the order of fallback code paths are the opposite of the fast   Notice, that the order of backtrack code paths are the opposite of the fast
139   code paths. In this way the topmost value on the stack is always belong   code paths. In this way the topmost value on the stack is always belong
140   to the current fallback code path. The fallback code path must check   to the current backtrack code path. The backtrack path must check
141   whether there is a next alternative. If so, it needs to jump back to   whether there is a next alternative. If so, it needs to jump back to
142   the hot path eventually. Otherwise it needs to clear out its own stack   the matching path eventually. Otherwise it needs to clear out its own stack
143   frame and continue the execution on the fallback code paths.   frame and continue the execution on the backtrack code paths.
144  */  */
145    
146  /*  /*
147  Saved stack frames:  Saved stack frames:
148    
149  Atomic blocks and asserts require reloading the values of local variables  Atomic blocks and asserts require reloading the values of private data
150  when the fallback mechanism performed. Because of OP_RECURSE, the locals  when the backtrack mechanism performed. Because of OP_RECURSE, the data
151  are not necessarly known in compile time, thus we need a dynamic restore  are not necessarly known in compile time, thus we need a dynamic restore
152  mechanism.  mechanism.
153    
154  The stack frames are stored in a chain list, and have the following format:  The stack frames are stored in a chain list, and have the following format:
155  ([ capturing bracket offset ][ start value ][ end value ])+ ... [ 0 ] [ previous head ]  ([ capturing bracket offset ][ start value ][ end value ])+ ... [ 0 ] [ previous head ]
156    
157  Thus we can restore the locals to a particular point in the stack.  Thus we can restore the private data to a particular point in the stack.
158  */  */
159    
160  typedef struct jit_arguments {  typedef struct jit_arguments {
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;    int real_offset_count;
172    int calllimit;    int offset_count;
173    uschar notbol;    int call_limit;
174    uschar noteol;    pcre_uint8 notbol;
175    uschar notempty;    pcre_uint8 noteol;
176    uschar notempty_atstart;    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      sljit_uw executable_sizes[JIT_NUMBER_OF_COMPILE_MODES];
186    } executable_functions;
187    
188  typedef struct jump_list {  typedef struct jump_list {
189    struct sljit_jump *jump;    struct sljit_jump *jump;
190    struct jump_list *next;    struct jump_list *next;
191  } jump_list;  } jump_list;
192    
 enum stub_types { stack_alloc };  
   
193  typedef struct stub_list {  typedef struct stub_list {
   enum stub_types type;  
   int data;  
194    struct sljit_jump *start;    struct sljit_jump *start;
195    struct sljit_label *leave;    struct sljit_label *quit;
196    struct stub_list *next;    struct stub_list *next;
197  } stub_list;  } stub_list;
198    
199    enum frame_types {
200      no_frame = -1,
201      no_stack = -2
202    };
203    
204    enum control_types {
205      type_mark = 0,
206      type_then_trap = 1
207    };
208    
209  typedef int (SLJIT_CALL *jit_function)(jit_arguments *args);  typedef int (SLJIT_CALL *jit_function)(jit_arguments *args);
210    
211  /* The following structure is the key data type for the recursive  /* The following structure is the key data type for the recursive
212  code generator. It is allocated by compile_hotpath, and contains  code generator. It is allocated by compile_matchingpath, and contains
213  the aguments for compile_fallbackpath. Must be the first member  the aguments for compile_backtrackingpath. Must be the first member
214  of its descendants. */  of its descendants. */
215  typedef struct fallback_common {  typedef struct backtrack_common {
216    /* Concatenation stack. */    /* Concatenation stack. */
217    struct fallback_common *prev;    struct backtrack_common *prev;
218    jump_list *nextfallbacks;    jump_list *nextbacktracks;
219    /* Internal stack (for component operators). */    /* Internal stack (for component operators). */
220    struct fallback_common *top;    struct backtrack_common *top;
221    jump_list *topfallbacks;    jump_list *topbacktracks;
222    /* Opcode pointer. */    /* Opcode pointer. */
223    uschar *cc;    pcre_uchar *cc;
224  } fallback_common;  } backtrack_common;
225    
226  typedef struct assert_fallback {  typedef struct assert_backtrack {
227    fallback_common common;    backtrack_common common;
228    jump_list *condfailed;    jump_list *condfailed;
229    /* Less than 0 (-1) if a frame is not needed. */    /* Less than 0 if a frame is not needed. */
230    int framesize;    int framesize;
231    /* Points to our private memory word on the stack. */    /* Points to our private memory word on the stack. */
232    int localptr;    int private_data_ptr;
233    /* For iterators. */    /* For iterators. */
234    struct sljit_label *hotpath;    struct sljit_label *matchingpath;
235  } assert_fallback;  } assert_backtrack;
236    
237  typedef struct bracket_fallback {  typedef struct bracket_backtrack {
238    fallback_common common;    backtrack_common common;
239    /* Where to coninue if an alternative is successfully matched. */    /* Where to coninue if an alternative is successfully matched. */
240    struct sljit_label *althotpath;    struct sljit_label *alternative_matchingpath;
241    /* For rmin and rmax iterators. */    /* For rmin and rmax iterators. */
242    struct sljit_label *recursivehotpath;    struct sljit_label *recursive_matchingpath;
243    /* For greedy ? operator. */    /* For greedy ? operator. */
244    struct sljit_label *zerohotpath;    struct sljit_label *zero_matchingpath;
245    /* Contains the branches of a failed condition. */    /* Contains the branches of a failed condition. */
246    union {    union {
247      /* Both for OP_COND, OP_SCOND. */      /* Both for OP_COND, OP_SCOND. */
248      jump_list *condfailed;      jump_list *condfailed;
249      assert_fallback *assert;      assert_backtrack *assert;
250      /* For OP_ONCE. -1 if not needed. */      /* For OP_ONCE. Less than 0 if not needed. */
251      int framesize;      int framesize;
252    } u;    } u;
253    /* Points to our private memory word on the stack. */    /* Points to our private memory word on the stack. */
254    int localptr;    int private_data_ptr;
255  } bracket_fallback;  } bracket_backtrack;
256    
257  typedef struct bracketpos_fallback {  typedef struct bracketpos_backtrack {
258    fallback_common common;    backtrack_common common;
259    /* Points to our private memory word on the stack. */    /* Points to our private memory word on the stack. */
260    int localptr;    int private_data_ptr;
261    /* Reverting stack is needed. */    /* Reverting stack is needed. */
262    int framesize;    int framesize;
263    /* Allocated stack size. */    /* Allocated stack size. */
264    int stacksize;    int stacksize;
265  } bracketpos_fallback;  } bracketpos_backtrack;
266    
267  typedef struct braminzero_fallback {  typedef struct braminzero_backtrack {
268    fallback_common common;    backtrack_common common;
269    struct sljit_label *hotpath;    struct sljit_label *matchingpath;
270  } braminzero_fallback;  } braminzero_backtrack;
271    
272  typedef struct iterator_fallback {  typedef struct iterator_backtrack {
273    fallback_common common;    backtrack_common common;
274    /* Next iteration. */    /* Next iteration. */
275    struct sljit_label *hotpath;    struct sljit_label *matchingpath;
276  } iterator_fallback;  } iterator_backtrack;
277    
278  typedef struct recurse_entry {  typedef struct recurse_entry {
279    struct recurse_entry *next;    struct recurse_entry *next;
# Line 259  typedef struct recurse_entry { Line 282  typedef struct recurse_entry {
282    /* Collects the calls until the function is not created. */    /* Collects the calls until the function is not created. */
283    jump_list *calls;    jump_list *calls;
284    /* Points to the starting opcode. */    /* Points to the starting opcode. */
285    int start;    sljit_sw start;
286  } recurse_entry;  } recurse_entry;
287    
288  typedef struct recurse_fallback {  typedef struct recurse_backtrack {
289    fallback_common common;    backtrack_common common;
290  } recurse_fallback;    BOOL inlined_pattern;
291    } recurse_backtrack;
292    
293    #define OP_THEN_TRAP OP_TABLE_LENGTH
294    
295    typedef struct then_trap_backtrack {
296      backtrack_common common;
297      /* If then_trap is not NULL, this structure contains the real
298      then_trap for the backtracking path. */
299      struct then_trap_backtrack *then_trap;
300      /* Points to the starting opcode. */
301      sljit_sw start;
302      /* Exit point for the then opcodes of this alternative. */
303      jump_list *quit;
304      /* Frame size of the current alternative. */
305      int framesize;
306    } then_trap_backtrack;
307    
308    #define MAX_RANGE_SIZE 6
309    
310  typedef struct compiler_common {  typedef struct compiler_common {
311      /* The sljit ceneric compiler. */
312    struct sljit_compiler *compiler;    struct sljit_compiler *compiler;
313    uschar *start;    /* First byte code. */
314    int localsize;    pcre_uchar *start;
315    int *localptrs;    /* Maps private data offset to each opcode. */
316    const uschar *fcc;    sljit_si *private_data_ptrs;
317    sljit_w lcc;    /* Tells whether the capturing bracket is optimized. */
318    int cbraptr;    pcre_uint8 *optimized_cbracket;
319      /* Tells whether the starting offset is a target of then. */
320      pcre_uint8 *then_offsets;
321      /* Current position where a THEN must jump. */
322      then_trap_backtrack *then_trap;
323      /* Starting offset of private data for capturing brackets. */
324      int cbra_ptr;
325      /* Output vector starting point. Must be divisible by 2. */
326      int ovector_start;
327      /* Last known position of the requested byte. */
328      int req_char_ptr;
329      /* Head of the last recursion. */
330      int recursive_head_ptr;
331      /* First inspected character for partial matching. */
332      int start_used_ptr;
333      /* Starting pointer for partial soft matches. */
334      int hit_start;
335      /* End pointer of the first line. */
336      int first_line_end;
337      /* Points to the marked string. */
338      int mark_ptr;
339      /* Recursive control verb management chain. */
340      int control_head_ptr;
341      /* Points to the last matched capture block index. */
342      int capture_last_ptr;
343      /* Points to the starting position of the current match. */
344      int start_ptr;
345    
346      /* Flipped and lower case tables. */
347      const pcre_uint8 *fcc;
348      sljit_sw lcc;
349      /* Mode can be PCRE_STUDY_JIT_COMPILE and others. */
350      int mode;
351      /* \K is found in the pattern. */
352      BOOL has_set_som;
353      /* (*SKIP:arg) is found in the pattern. */
354      BOOL has_skip_arg;
355      /* (*THEN) is found in the pattern. */
356      BOOL has_then;
357      /* Needs to know the start position anytime. */
358      BOOL needs_start_ptr;
359      /* Currently in recurse or negative assert. */
360      BOOL local_exit;
361      /* Currently in a positive assert. */
362      BOOL positive_assert;
363      /* Newline control. */
364    int nltype;    int nltype;
365    int newline;    int newline;
366    int bsr_nltype;    int bsr_nltype;
367      /* Dollar endonly. */
368    int endonly;    int endonly;
369    sljit_w ctypes;    /* Tables. */
370      sljit_sw ctypes;
371      int digits[2 + MAX_RANGE_SIZE];
372      /* Named capturing brackets. */
373    sljit_uw name_table;    sljit_uw name_table;
374    sljit_w name_count;    sljit_sw name_count;
375    sljit_w name_entry_size;    sljit_sw name_entry_size;
376    struct sljit_label *acceptlabel;  
377      /* Labels and jump lists. */
378      struct sljit_label *partialmatchlabel;
379      struct sljit_label *quit_label;
380      struct sljit_label *forced_quit_label;
381      struct sljit_label *accept_label;
382    stub_list *stubs;    stub_list *stubs;
383    recurse_entry *entries;    recurse_entry *entries;
384    recurse_entry *currententry;    recurse_entry *currententry;
385      jump_list *partialmatch;
386      jump_list *quit;
387      jump_list *positive_assert_quit;
388      jump_list *forced_quit;
389    jump_list *accept;    jump_list *accept;
390    jump_list *calllimit;    jump_list *calllimit;
391    jump_list *stackalloc;    jump_list *stackalloc;
# Line 296  typedef struct compiler_common { Line 396  typedef struct compiler_common {
396    jump_list *vspace;    jump_list *vspace;
397    jump_list *casefulcmp;    jump_list *casefulcmp;
398    jump_list *caselesscmp;    jump_list *caselesscmp;
399      jump_list *reset_match;
400    BOOL jscript_compat;    BOOL jscript_compat;
401  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
402    BOOL utf8;    BOOL utf;
403  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
404    BOOL useucp;    BOOL use_ucp;
405  #endif  #endif
406    jump_list *utf8readchar;  #ifndef COMPILE_PCRE32
407    jump_list *utf8readtype8;    jump_list *utfreadchar;
408  #endif  #endif
409    #ifdef COMPILE_PCRE8
410      jump_list *utfreadtype8;
411    #endif
412    #endif /* SUPPORT_UTF */
413  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
414    jump_list *getucd;    jump_list *getucd;
415  #endif  #endif
# Line 316  typedef struct compare_context { Line 421  typedef struct compare_context {
421    int length;    int length;
422    int sourcereg;    int sourcereg;
423  #if defined SLJIT_UNALIGNED && SLJIT_UNALIGNED  #if defined SLJIT_UNALIGNED && SLJIT_UNALIGNED
424    int byteptr;    int ucharptr;
425    union {    union {
426      int asint;      sljit_si asint;
427      short asshort;      sljit_uh asushort;
428    #if defined COMPILE_PCRE8
429      sljit_ub asbyte;      sljit_ub asbyte;
430      sljit_ub asbytes[4];      sljit_ub asuchars[4];
431    #elif defined COMPILE_PCRE16
432        sljit_uh asuchars[2];
433    #elif defined COMPILE_PCRE32
434        sljit_ui asuchars[1];
435    #endif
436    } c;    } c;
437    union {    union {
438      int asint;      sljit_si asint;
439      short asshort;      sljit_uh asushort;
440    #if defined COMPILE_PCRE8
441      sljit_ub asbyte;      sljit_ub asbyte;
442      sljit_ub asbytes[4];      sljit_ub asuchars[4];
443    #elif defined COMPILE_PCRE16
444        sljit_uh asuchars[2];
445    #elif defined COMPILE_PCRE32
446        sljit_ui asuchars[1];
447    #endif
448    } oc;    } oc;
449  #endif  #endif
450  } compare_context;  } compare_context;
451    
452  enum {  /* Undefine sljit macros. */
453    frame_end = 0,  #undef CMP
   frame_setstrbegin = -1  
 };  
454    
455  /* Used for accessing the elements of the stack. */  /* Used for accessing the elements of the stack. */
456  #define STACK(i)      ((-(i) - 1) * (int)sizeof(sljit_w))  #define STACK(i)      ((-(i) - 1) * (int)sizeof(sljit_sw))
457    
458  #define TMP1          SLJIT_TEMPORARY_REG1  #define TMP1          SLJIT_SCRATCH_REG1
459  #define TMP2          SLJIT_TEMPORARY_REG3  #define TMP2          SLJIT_SCRATCH_REG3
460  #define TMP3          SLJIT_TEMPORARY_EREG2  #define TMP3          SLJIT_TEMPORARY_EREG2
461  #define STR_PTR       SLJIT_GENERAL_REG1  #define STR_PTR       SLJIT_SAVED_REG1
462  #define STR_END       SLJIT_GENERAL_REG2  #define STR_END       SLJIT_SAVED_REG2
463  #define STACK_TOP     SLJIT_TEMPORARY_REG2  #define STACK_TOP     SLJIT_SCRATCH_REG2
464  #define STACK_LIMIT   SLJIT_GENERAL_REG3  #define STACK_LIMIT   SLJIT_SAVED_REG3
465  #define ARGUMENTS     SLJIT_GENERAL_EREG1  #define ARGUMENTS     SLJIT_SAVED_EREG1
466  #define CALL_COUNT    SLJIT_GENERAL_EREG2  #define CALL_COUNT    SLJIT_SAVED_EREG2
467  #define RETURN_ADDR   SLJIT_TEMPORARY_EREG1  #define RETURN_ADDR   SLJIT_TEMPORARY_EREG1
468    
469  /* Locals layout. */  /* Local space layout. */
470  /* These two locals can be used by the current opcode. */  /* These two locals can be used by the current opcode. */
471  #define LOCALS0          (0 * sizeof(sljit_w))  #define LOCALS0          (0 * sizeof(sljit_sw))
472  #define LOCALS1          (1 * sizeof(sljit_w))  #define LOCALS1          (1 * sizeof(sljit_sw))
473  /* Two local variables for possessive quantifiers (char1 cannot use them). */  /* Two local variables for possessive quantifiers (char1 cannot use them). */
474  #define POSSESSIVE0      (2 * sizeof(sljit_w))  #define POSSESSIVE0      (2 * sizeof(sljit_sw))
475  #define POSSESSIVE1      (3 * sizeof(sljit_w))  #define POSSESSIVE1      (3 * sizeof(sljit_sw))
 /* Head of the last recursion. */  
 #define RECURSIVE_HEAD   (4 * sizeof(sljit_w))  
476  /* Max limit of recursions. */  /* Max limit of recursions. */
477  #define CALL_LIMIT       (5 * sizeof(sljit_w))  #define CALL_LIMIT       (4 * sizeof(sljit_sw))
 /* 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))  
478  /* The output vector is stored on the stack, and contains pointers  /* The output vector is stored on the stack, and contains pointers
479  to characters. The vector data is divided into two groups: the first  to characters. The vector data is divided into two groups: the first
480  group contains the start / end character pointers, and the second is  group contains the start / end character pointers, and the second is
481  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. */
482  #define OVECTOR_START    (8 * sizeof(sljit_w))  #define OVECTOR_START    (common->ovector_start)
483  #define OVECTOR(i)       (OVECTOR_START + (i) * sizeof(sljit_w))  #define OVECTOR(i)       (OVECTOR_START + (i) * sizeof(sljit_sw))
484  #define OVECTOR_PRIV(i)  (common->cbraptr + (i) * sizeof(sljit_w))  #define OVECTOR_PRIV(i)  (common->cbra_ptr + (i) * sizeof(sljit_sw))
485  #define PRIV(cc)         (common->localptrs[(cc) - common->start])  #define PRIVATE_DATA(cc) (common->private_data_ptrs[(cc) - common->start])
486    
487    #if defined COMPILE_PCRE8
488    #define MOV_UCHAR  SLJIT_MOV_UB
489    #define MOVU_UCHAR SLJIT_MOVU_UB
490    #elif defined COMPILE_PCRE16
491    #define MOV_UCHAR  SLJIT_MOV_UH
492    #define MOVU_UCHAR SLJIT_MOVU_UH
493    #elif defined COMPILE_PCRE32
494    #define MOV_UCHAR  SLJIT_MOV_UI
495    #define MOVU_UCHAR SLJIT_MOVU_UI
496    #else
497    #error Unsupported compiling mode
498    #endif
499    
500  /* Shortcuts. */  /* Shortcuts. */
501  #define DEFINE_COMPILER \  #define DEFINE_COMPILER \
# Line 390  the start pointers when the end of the c Line 512  the start pointers when the end of the c
512    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))
513  #define JUMPHERE(jump) \  #define JUMPHERE(jump) \
514    sljit_set_label((jump), sljit_emit_label(compiler))    sljit_set_label((jump), sljit_emit_label(compiler))
515    #define SET_LABEL(jump, label) \
516      sljit_set_label((jump), (label))
517  #define CMP(type, src1, src1w, src2, src2w) \  #define CMP(type, src1, src1w, src2, src2w) \
518    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))
519  #define CMPTO(type, src1, src1w, src2, src2w, label) \  #define CMPTO(type, src1, src1w, src2, src2w, label) \
520    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))
521  #define COND_VALUE(op, dst, dstw, type) \  #define OP_FLAGS(op, dst, dstw, src, srcw, type) \
522    sljit_emit_cond_value(compiler, (op), (dst), (dstw), (type))    sljit_emit_op_flags(compiler, (op), (dst), (dstw), (src), (srcw), (type))
523    #define GET_LOCAL_BASE(dst, dstw, offset) \
524      sljit_get_local_base(compiler, (dst), (dstw), (offset))
525    
526  static uschar* bracketend(uschar* cc)  static pcre_uchar* bracketend(pcre_uchar* cc)
527  {  {
528  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));
529  do cc += GET(cc, 1); while (*cc == OP_ALT);  do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 408  return cc; Line 534  return cc;
534    
535  /* Functions whose might need modification for all new supported opcodes:  /* Functions whose might need modification for all new supported opcodes:
536   next_opcode   next_opcode
537   get_localspace   check_opcode_types
538   set_localptrs   set_private_data_ptrs
539   get_framesize   get_framesize
540   init_frame   init_frame
541   get_localsize   get_private_data_copy_length
542   copy_locals   copy_private_data
543   compile_hotpath   compile_matchingpath
544   compile_fallbackpath   compile_backtrackingpath
545  */  */
546    
547  static uschar *next_opcode(compiler_common *common, uschar *cc)  static pcre_uchar *next_opcode(compiler_common *common, pcre_uchar *cc)
548  {  {
549  SLJIT_UNUSED_ARG(common);  SLJIT_UNUSED_ARG(common);
550  switch(*cc)  switch(*cc)
# Line 436  switch(*cc) Line 562  switch(*cc)
562    case OP_WORDCHAR:    case OP_WORDCHAR:
563    case OP_ANY:    case OP_ANY:
564    case OP_ALLANY:    case OP_ALLANY:
565      case OP_NOTPROP:
566      case OP_PROP:
567    case OP_ANYNL:    case OP_ANYNL:
568    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
569    case OP_HSPACE:    case OP_HSPACE:
# Line 448  switch(*cc) Line 576  switch(*cc)
576    case OP_CIRCM:    case OP_CIRCM:
577    case OP_DOLL:    case OP_DOLL:
578    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:  
579    case OP_CRSTAR:    case OP_CRSTAR:
580    case OP_CRMINSTAR:    case OP_CRMINSTAR:
581    case OP_CRPLUS:    case OP_CRPLUS:
582    case OP_CRMINPLUS:    case OP_CRMINPLUS:
583    case OP_CRQUERY:    case OP_CRQUERY:
584    case OP_CRMINQUERY:    case OP_CRMINQUERY:
585      case OP_CRRANGE:
586      case OP_CRMINRANGE:
587      case OP_CLASS:
588      case OP_NCLASS:
589      case OP_REF:
590      case OP_REFI:
591      case OP_RECURSE:
592      case OP_CALLOUT:
593      case OP_ALT:
594      case OP_KET:
595      case OP_KETRMAX:
596      case OP_KETRMIN:
597      case OP_KETRPOS:
598      case OP_REVERSE:
599      case OP_ASSERT:
600      case OP_ASSERT_NOT:
601      case OP_ASSERTBACK:
602      case OP_ASSERTBACK_NOT:
603      case OP_ONCE:
604      case OP_ONCE_NC:
605      case OP_BRA:
606      case OP_BRAPOS:
607      case OP_CBRA:
608      case OP_CBRAPOS:
609      case OP_COND:
610      case OP_SBRA:
611      case OP_SBRAPOS:
612      case OP_SCBRA:
613      case OP_SCBRAPOS:
614      case OP_SCOND:
615      case OP_CREF:
616      case OP_NCREF:
617      case OP_RREF:
618      case OP_NRREF:
619    case OP_DEF:    case OP_DEF:
620    case OP_BRAZERO:    case OP_BRAZERO:
621    case OP_BRAMINZERO:    case OP_BRAMINZERO:
622    case OP_BRAPOSZERO:    case OP_BRAPOSZERO:
623      case OP_PRUNE:
624      case OP_SKIP:
625      case OP_THEN:
626      case OP_COMMIT:
627    case OP_FAIL:    case OP_FAIL:
628    case OP_ACCEPT:    case OP_ACCEPT:
629    case OP_ASSERT_ACCEPT:    case OP_ASSERT_ACCEPT:
630      case OP_CLOSE:
631    case OP_SKIPZERO:    case OP_SKIPZERO:
632    return cc + 1;    return cc + PRIV(OP_lengths)[*cc];
   
   case OP_ANYBYTE:  
 #ifdef SUPPORT_UTF8  
   if (common->utf8) return NULL;  
 #endif  
   return cc + 1;  
633    
634    case OP_CHAR:    case OP_CHAR:
635    case OP_CHARI:    case OP_CHARI:
636    case OP_NOT:    case OP_NOT:
637    case OP_NOTI:    case OP_NOTI:
   
638    case OP_STAR:    case OP_STAR:
639    case OP_MINSTAR:    case OP_MINSTAR:
640    case OP_PLUS:    case OP_PLUS:
641    case OP_MINPLUS:    case OP_MINPLUS:
642    case OP_QUERY:    case OP_QUERY:
643    case OP_MINQUERY:    case OP_MINQUERY:
644      case OP_UPTO:
645      case OP_MINUPTO:
646      case OP_EXACT:
647    case OP_POSSTAR:    case OP_POSSTAR:
648    case OP_POSPLUS:    case OP_POSPLUS:
649    case OP_POSQUERY:    case OP_POSQUERY:
650      case OP_POSUPTO:
651    case OP_STARI:    case OP_STARI:
652    case OP_MINSTARI:    case OP_MINSTARI:
653    case OP_PLUSI:    case OP_PLUSI:
654    case OP_MINPLUSI:    case OP_MINPLUSI:
655    case OP_QUERYI:    case OP_QUERYI:
656    case OP_MINQUERYI:    case OP_MINQUERYI:
657      case OP_UPTOI:
658      case OP_MINUPTOI:
659      case OP_EXACTI:
660    case OP_POSSTARI:    case OP_POSSTARI:
661    case OP_POSPLUSI:    case OP_POSPLUSI:
662    case OP_POSQUERYI:    case OP_POSQUERYI:
663      case OP_POSUPTOI:
664    case OP_NOTSTAR:    case OP_NOTSTAR:
665    case OP_NOTMINSTAR:    case OP_NOTMINSTAR:
666    case OP_NOTPLUS:    case OP_NOTPLUS:
667    case OP_NOTMINPLUS:    case OP_NOTMINPLUS:
668    case OP_NOTQUERY:    case OP_NOTQUERY:
669    case OP_NOTMINQUERY:    case OP_NOTMINQUERY:
670      case OP_NOTUPTO:
671      case OP_NOTMINUPTO:
672      case OP_NOTEXACT:
673    case OP_NOTPOSSTAR:    case OP_NOTPOSSTAR:
674    case OP_NOTPOSPLUS:    case OP_NOTPOSPLUS:
675    case OP_NOTPOSQUERY:    case OP_NOTPOSQUERY:
676      case OP_NOTPOSUPTO:
677    case OP_NOTSTARI:    case OP_NOTSTARI:
678    case OP_NOTMINSTARI:    case OP_NOTMINSTARI:
679    case OP_NOTPLUSI:    case OP_NOTPLUSI:
680    case OP_NOTMINPLUSI:    case OP_NOTMINPLUSI:
681    case OP_NOTQUERYI:    case OP_NOTQUERYI:
682    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:  
683    case OP_NOTUPTOI:    case OP_NOTUPTOI:
684    case OP_NOTMINUPTOI:    case OP_NOTMINUPTOI:
685    case OP_NOTEXACTI:    case OP_NOTEXACTI:
686      case OP_NOTPOSSTARI:
687      case OP_NOTPOSPLUSI:
688      case OP_NOTPOSQUERYI:
689    case OP_NOTPOSUPTOI:    case OP_NOTPOSUPTOI:
690    cc += 4;    cc += PRIV(OP_lengths)[*cc];
691  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
692    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]);
693  #endif  #endif
694    return cc;    return cc;
695    
696    case OP_NOTPROP:    /* Special cases. */
697    case OP_PROP:    case OP_TYPESTAR:
698      case OP_TYPEMINSTAR:
699      case OP_TYPEPLUS:
700      case OP_TYPEMINPLUS:
701      case OP_TYPEQUERY:
702      case OP_TYPEMINQUERY:
703    case OP_TYPEUPTO:    case OP_TYPEUPTO:
704    case OP_TYPEMINUPTO:    case OP_TYPEMINUPTO:
705    case OP_TYPEEXACT:    case OP_TYPEEXACT:
706      case OP_TYPEPOSSTAR:
707      case OP_TYPEPOSPLUS:
708      case OP_TYPEPOSQUERY:
709    case OP_TYPEPOSUPTO:    case OP_TYPEPOSUPTO:
710    case OP_REF:    return cc + PRIV(OP_lengths)[*cc] - 1;
   case OP_REFI:  
   case OP_CREF:  
   case OP_NCREF:  
   case OP_RREF:  
   case OP_NRREF:  
   case OP_CLOSE:  
   cc += 3;  
   return cc;  
   
   case OP_CRRANGE:  
   case OP_CRMINRANGE:  
   return cc + 5;  
711    
712    case OP_CLASS:    case OP_ANYBYTE:
713    case OP_NCLASS:  #ifdef SUPPORT_UTF
714    return cc + 33;    if (common->utf) return NULL;
715    #endif
716      return cc + 1;
717    
718  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
719    case OP_XCLASS:    case OP_XCLASS:
720    return cc + GET(cc, 1);    return cc + GET(cc, 1);
721  #endif  #endif
722    
723    case OP_RECURSE:    case OP_MARK:
724    case OP_ASSERT:    case OP_PRUNE_ARG:
725    case OP_ASSERT_NOT:    case OP_SKIP_ARG:
726    case OP_ASSERTBACK:    case OP_THEN_ARG:
727    case OP_ASSERTBACK_NOT:    return cc + 1 + 2 + cc[1];
   case OP_REVERSE:  
   case OP_ONCE:  
   case OP_ONCE_NC:  
   case OP_BRA:  
   case OP_BRAPOS:  
   case OP_COND:  
   case OP_SBRA:  
   case OP_SBRAPOS:  
   case OP_SCOND:  
   case OP_ALT:  
   case OP_KET:  
   case OP_KETRMAX:  
   case OP_KETRMIN:  
   case OP_KETRPOS:  
   return cc + 1 + LINK_SIZE;  
   
   case OP_CBRA:  
   case OP_CBRAPOS:  
   case OP_SCBRA:  
   case OP_SCBRAPOS:  
   return cc + 1 + LINK_SIZE + 2;  
728    
729    default:    default:
730      /* All opcodes are supported now! */
731      SLJIT_ASSERT_STOP();
732    return NULL;    return NULL;
733    }    }
734  }  }
735    
736  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)
737  {  {
738  int localspace = 0;  pcre_uchar *name;
739  uschar *alternative;  pcre_uchar *name2;
740    unsigned int cbra_index;
741    int i;
742    
743  /* 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. */
744  while (cc < ccend)  while (cc < ccend)
745    {    {
746    switch(*cc)    switch(*cc)
747      {      {
748      case OP_ASSERT:      case OP_SET_SOM:
749      case OP_ASSERT_NOT:      common->has_set_som = TRUE;
750      case OP_ASSERTBACK:      cc += 1;
751      case OP_ASSERTBACK_NOT:      break;
752      case OP_ONCE:  
753      case OP_ONCE_NC:      case OP_REF:
754      case OP_BRAPOS:      case OP_REFI:
755      case OP_SBRA:      common->optimized_cbracket[GET2(cc, 1)] = 0;
756      case OP_SBRAPOS:      cc += 1 + IMM2_SIZE;
     case OP_SCOND:  
     localspace += sizeof(sljit_w);  
     cc += 1 + LINK_SIZE;  
757      break;      break;
758    
759      case OP_CBRAPOS:      case OP_CBRAPOS:
760      case OP_SCBRAPOS:      case OP_SCBRAPOS:
761      localspace += sizeof(sljit_w);      common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] = 0;
762      cc += 1 + LINK_SIZE + 2;      cc += 1 + LINK_SIZE + IMM2_SIZE;
763      break;      break;
764    
765      case OP_COND:      case OP_COND:
766      /* Might be a hidden SCOND. */      case OP_SCOND:
767      alternative = cc + GET(cc, 1);      /* Only AUTO_CALLOUT can insert this opcode. We do
768      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)         not intend to support this case. */
769        localspace += sizeof(sljit_w);      if (cc[1 + LINK_SIZE] == OP_CALLOUT)
770          return FALSE;
771      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
772      break;      break;
773    
774      default:      case OP_CREF:
775      cc = next_opcode(common, cc);      i = GET2(cc, 1);
776      if (cc == NULL)      common->optimized_cbracket[i] = 0;
777        return -1;      cc += 1 + IMM2_SIZE;
778      break;      break;
     }  
   }  
 return localspace;  
 }  
779    
780  static void set_localptrs(compiler_common *common, int localptr, uschar *ccend)      case OP_NCREF:
781  {      cbra_index = GET2(cc, 1);
782  uschar *cc = common->start;      name = (pcre_uchar *)common->name_table;
783  uschar *alternative;      name2 = name;
784  while (cc < ccend)      for (i = 0; i < common->name_count; i++)
785    {        {
786    switch(*cc)        if (GET2(name, 0) == cbra_index) break;
787      {        name += common->name_entry_size;
788      case OP_ASSERT:        }
789      case OP_ASSERT_NOT:      SLJIT_ASSERT(i != common->name_count);
     case OP_ASSERTBACK:  
     case OP_ASSERTBACK_NOT:  
     case OP_ONCE:  
     case OP_ONCE_NC:  
     case OP_BRAPOS:  
     case OP_SBRA:  
     case OP_SBRAPOS:  
     case OP_SCOND:  
     common->localptrs[cc - common->start] = localptr;  
     localptr += sizeof(sljit_w);  
     cc += 1 + LINK_SIZE;  
     break;  
790    
791      case OP_CBRAPOS:      for (i = 0; i < common->name_count; i++)
792      case OP_SCBRAPOS:        {
793      common->localptrs[cc - common->start] = localptr;        if (STRCMP_UC_UC(name2 + IMM2_SIZE, name + IMM2_SIZE) == 0)
794      localptr += sizeof(sljit_w);          common->optimized_cbracket[GET2(name2, 0)] = 0;
795      cc += 1 + LINK_SIZE + 2;        name2 += common->name_entry_size;
796          }
797        cc += 1 + IMM2_SIZE;
798      break;      break;
799    
800      case OP_COND:      case OP_RECURSE:
801      /* Might be a hidden SCOND. */      /* Set its value only once. */
802      alternative = cc + GET(cc, 1);      if (common->recursive_head_ptr == 0)
     if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)  
803        {        {
804        common->localptrs[cc - common->start] = localptr;        common->recursive_head_ptr = common->ovector_start;
805        localptr += sizeof(sljit_w);        common->ovector_start += sizeof(sljit_sw);
806        }        }
807      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
808      break;      break;
809    
810      default:      case OP_CALLOUT:
811      cc = next_opcode(common, cc);      if (common->capture_last_ptr == 0)
812      SLJIT_ASSERT(cc != NULL);        {
813          common->capture_last_ptr = common->ovector_start;
814          common->ovector_start += sizeof(sljit_sw);
815          }
816        cc += 2 + 2 * LINK_SIZE;
817      break;      break;
     }  
   }  
 }  
818    
819  /* Returns with -1 if no need for frame. */      case OP_THEN_ARG:
820  static int get_framesize(compiler_common *common, uschar *cc, BOOL recursive)      common->has_then = TRUE;
821  {      common->control_head_ptr = 1;
822  uschar *ccend = bracketend(cc);      /* Fall through. */
 int length = 0;  
 BOOL possessive = FALSE;  
 BOOL setsom_found = FALSE;  
823    
824  if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))      case OP_PRUNE_ARG:
825    {      common->needs_start_ptr = TRUE;
826    length = 3;      /* Fall through. */
   possessive = TRUE;  
   }  
827    
828  cc = next_opcode(common, cc);      case OP_MARK:
829  SLJIT_ASSERT(cc != NULL);      if (common->mark_ptr == 0)
 while (cc < ccend)  
   switch(*cc)  
     {  
     case OP_SET_SOM:  
     case OP_RECURSE:  
     if (!setsom_found)  
830        {        {
831        length += 2;        common->mark_ptr = common->ovector_start;
832        setsom_found = TRUE;        common->ovector_start += sizeof(sljit_sw);
833        }        }
834      cc += (*cc == OP_SET_SOM) ? 1 : 1 + LINK_SIZE;      cc += 1 + 2 + cc[1];
835      break;      break;
836    
837      case OP_CBRA:      case OP_THEN:
838      case OP_CBRAPOS:      common->has_then = TRUE;
839      case OP_SCBRA:      common->control_head_ptr = 1;
840      case OP_SCBRAPOS:      /* Fall through. */
841      length += 3;  
842      cc += 1 + LINK_SIZE + 2;      case OP_PRUNE:
843        case OP_SKIP:
844        common->needs_start_ptr = TRUE;
845        cc += 1;
846        break;
847    
848        case OP_SKIP_ARG:
849        common->control_head_ptr = 1;
850        common->has_skip_arg = TRUE;
851        cc += 1 + 2 + cc[1];
852      break;      break;
853    
854      default:      default:
855      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
856      SLJIT_ASSERT(cc != NULL);      if (cc == NULL)
857          return FALSE;
858      break;      break;
859      }      }
860      }
861  /* Possessive quantifiers can use a special case. */  return TRUE;
 if (SLJIT_UNLIKELY(possessive) && length == 3)  
   return -1;  
   
 if (length > 0)  
   return length + 1;  
 return -1;  
862  }  }
863    
864  static void init_frame(compiler_common *common, uschar *cc, int stackpos, int stacktop, BOOL recursive)  static int get_class_iterator_size(pcre_uchar *cc)
865  {  {
866  DEFINE_COMPILER;  switch(*cc)
867  uschar *ccend = bracketend(cc);    {
868  BOOL setsom_found = FALSE;    case OP_CRSTAR:
869  int offset;    case OP_CRPLUS:
870      return 2;
 /* >= 1 + shortest item size (2) */  
 SLJIT_ASSERT(stackpos >= stacktop + 2);  
871    
872  stackpos = STACK(stackpos);    case OP_CRMINSTAR:
873  if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))    case OP_CRMINPLUS:
874      case OP_CRQUERY:
875      case OP_CRMINQUERY:
876      return 1;
877    
878      case OP_CRRANGE:
879      case OP_CRMINRANGE:
880      if (GET2(cc, 1) == GET2(cc, 1 + IMM2_SIZE))
881        return 0;
882      return 2;
883    
884      default:
885      return 0;
886      }
887    }
888    
889    static BOOL detect_repeat(compiler_common *common, pcre_uchar *begin)
890    {
891    pcre_uchar *end = bracketend(begin);
892    pcre_uchar *next;
893    pcre_uchar *next_end;
894    pcre_uchar *max_end;
895    pcre_uchar type;
896    sljit_sw length = end - begin;
897    int min, max, i;
898    
899    /* Detect fixed iterations first. */
900    if (end[-(1 + LINK_SIZE)] != OP_KET)
901      return FALSE;
902    
903    /* Already detected repeat. */
904    if (common->private_data_ptrs[end - common->start - LINK_SIZE] != 0)
905      return TRUE;
906    
907    next = end;
908    min = 1;
909    while (1)
910      {
911      if (*next != *begin)
912        break;
913      next_end = bracketend(next);
914      if (next_end - next != length || memcmp(begin, next, IN_UCHARS(length)) != 0)
915        break;
916      next = next_end;
917      min++;
918      }
919    
920    if (min == 2)
921      return FALSE;
922    
923    max = 0;
924    max_end = next;
925    if (*next == OP_BRAZERO || *next == OP_BRAMINZERO)
926      {
927      type = *next;
928      while (1)
929        {
930        if (next[0] != type || next[1] != OP_BRA || next[2 + LINK_SIZE] != *begin)
931          break;
932        next_end = bracketend(next + 2 + LINK_SIZE);
933        if (next_end - next != (length + 2 + LINK_SIZE) || memcmp(begin, next + 2 + LINK_SIZE, IN_UCHARS(length)) != 0)
934          break;
935        next = next_end;
936        max++;
937        }
938    
939      if (next[0] == type && next[1] == *begin && max >= 1)
940        {
941        next_end = bracketend(next + 1);
942        if (next_end - next == (length + 1) && memcmp(begin, next + 1, IN_UCHARS(length)) == 0)
943          {
944          for (i = 0; i < max; i++, next_end += 1 + LINK_SIZE)
945            if (*next_end != OP_KET)
946              break;
947    
948          if (i == max)
949            {
950            common->private_data_ptrs[max_end - common->start - LINK_SIZE] = next_end - max_end;
951            common->private_data_ptrs[max_end - common->start - LINK_SIZE + 1] = (type == OP_BRAZERO) ? OP_UPTO : OP_MINUPTO;
952            /* +2 the original and the last. */
953            common->private_data_ptrs[max_end - common->start - LINK_SIZE + 2] = max + 2;
954            if (min == 1)
955              return TRUE;
956            min--;
957            max_end -= (1 + LINK_SIZE) + GET(max_end, -LINK_SIZE);
958            }
959          }
960        }
961      }
962    
963    if (min >= 3)
964      {
965      common->private_data_ptrs[end - common->start - LINK_SIZE] = max_end - end;
966      common->private_data_ptrs[end - common->start - LINK_SIZE + 1] = OP_EXACT;
967      common->private_data_ptrs[end - common->start - LINK_SIZE + 2] = min;
968      return TRUE;
969      }
970    
971    return FALSE;
972    }
973    
974    #define CASE_ITERATOR_PRIVATE_DATA_1 \
975        case OP_MINSTAR: \
976        case OP_MINPLUS: \
977        case OP_QUERY: \
978        case OP_MINQUERY: \
979        case OP_MINSTARI: \
980        case OP_MINPLUSI: \
981        case OP_QUERYI: \
982        case OP_MINQUERYI: \
983        case OP_NOTMINSTAR: \
984        case OP_NOTMINPLUS: \
985        case OP_NOTQUERY: \
986        case OP_NOTMINQUERY: \
987        case OP_NOTMINSTARI: \
988        case OP_NOTMINPLUSI: \
989        case OP_NOTQUERYI: \
990        case OP_NOTMINQUERYI:
991    
992    #define CASE_ITERATOR_PRIVATE_DATA_2A \
993        case OP_STAR: \
994        case OP_PLUS: \
995        case OP_STARI: \
996        case OP_PLUSI: \
997        case OP_NOTSTAR: \
998        case OP_NOTPLUS: \
999        case OP_NOTSTARI: \
1000        case OP_NOTPLUSI:
1001    
1002    #define CASE_ITERATOR_PRIVATE_DATA_2B \
1003        case OP_UPTO: \
1004        case OP_MINUPTO: \
1005        case OP_UPTOI: \
1006        case OP_MINUPTOI: \
1007        case OP_NOTUPTO: \
1008        case OP_NOTMINUPTO: \
1009        case OP_NOTUPTOI: \
1010        case OP_NOTMINUPTOI:
1011    
1012    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_1 \
1013        case OP_TYPEMINSTAR: \
1014        case OP_TYPEMINPLUS: \
1015        case OP_TYPEQUERY: \
1016        case OP_TYPEMINQUERY:
1017    
1018    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2A \
1019        case OP_TYPESTAR: \
1020        case OP_TYPEPLUS:
1021    
1022    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2B \
1023        case OP_TYPEUPTO: \
1024        case OP_TYPEMINUPTO:
1025    
1026    static void set_private_data_ptrs(compiler_common *common, int *private_data_start, pcre_uchar *ccend)
1027    {
1028    pcre_uchar *cc = common->start;
1029    pcre_uchar *alternative;
1030    pcre_uchar *end = NULL;
1031    int private_data_ptr = *private_data_start;
1032    int space, size, bracketlen;
1033    
1034    while (cc < ccend)
1035      {
1036      space = 0;
1037      size = 0;
1038      bracketlen = 0;
1039      if (private_data_ptr > SLJIT_MAX_LOCAL_SIZE)
1040        return;
1041    
1042      if (*cc == OP_ONCE || *cc == OP_ONCE_NC || *cc == OP_BRA || *cc == OP_CBRA || *cc == OP_COND)
1043        if (detect_repeat(common, cc))
1044          {
1045          /* These brackets are converted to repeats, so no global
1046          based single character repeat is allowed. */
1047          if (cc >= end)
1048            end = bracketend(cc);
1049          }
1050    
1051      switch(*cc)
1052        {
1053        case OP_KET:
1054        if (common->private_data_ptrs[cc + 1 - common->start] != 0)
1055          {
1056          common->private_data_ptrs[cc - common->start] = private_data_ptr;
1057          private_data_ptr += sizeof(sljit_sw);
1058          cc += common->private_data_ptrs[cc + 1 - common->start];
1059          }
1060        cc += 1 + LINK_SIZE;
1061        break;
1062    
1063        case OP_ASSERT:
1064        case OP_ASSERT_NOT:
1065        case OP_ASSERTBACK:
1066        case OP_ASSERTBACK_NOT:
1067        case OP_ONCE:
1068        case OP_ONCE_NC:
1069        case OP_BRAPOS:
1070        case OP_SBRA:
1071        case OP_SBRAPOS:
1072        case OP_SCOND:
1073        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1074        private_data_ptr += sizeof(sljit_sw);
1075        bracketlen = 1 + LINK_SIZE;
1076        break;
1077    
1078        case OP_CBRAPOS:
1079        case OP_SCBRAPOS:
1080        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1081        private_data_ptr += sizeof(sljit_sw);
1082        bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
1083        break;
1084    
1085        case OP_COND:
1086        /* Might be a hidden SCOND. */
1087        alternative = cc + GET(cc, 1);
1088        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1089          {
1090          common->private_data_ptrs[cc - common->start] = private_data_ptr;
1091          private_data_ptr += sizeof(sljit_sw);
1092          }
1093        bracketlen = 1 + LINK_SIZE;
1094        break;
1095    
1096        case OP_BRA:
1097        bracketlen = 1 + LINK_SIZE;
1098        break;
1099    
1100        case OP_CBRA:
1101        case OP_SCBRA:
1102        bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
1103        break;
1104    
1105        CASE_ITERATOR_PRIVATE_DATA_1
1106        space = 1;
1107        size = -2;
1108        break;
1109    
1110        CASE_ITERATOR_PRIVATE_DATA_2A
1111        space = 2;
1112        size = -2;
1113        break;
1114    
1115        CASE_ITERATOR_PRIVATE_DATA_2B
1116        space = 2;
1117        size = -(2 + IMM2_SIZE);
1118        break;
1119    
1120        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1121        space = 1;
1122        size = 1;
1123        break;
1124    
1125        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1126        if (cc[1] != OP_ANYNL && cc[1] != OP_EXTUNI)
1127          space = 2;
1128        size = 1;
1129        break;
1130    
1131        CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1132        if (cc[1 + IMM2_SIZE] != OP_ANYNL && cc[1 + IMM2_SIZE] != OP_EXTUNI)
1133          space = 2;
1134        size = 1 + IMM2_SIZE;
1135        break;
1136    
1137        case OP_CLASS:
1138        case OP_NCLASS:
1139        size += 1 + 32 / sizeof(pcre_uchar);
1140        space = get_class_iterator_size(cc + size);
1141        break;
1142    
1143    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1144        case OP_XCLASS:
1145        size = GET(cc, 1);
1146        space = get_class_iterator_size(cc + size);
1147        break;
1148    #endif
1149    
1150        default:
1151        cc = next_opcode(common, cc);
1152        SLJIT_ASSERT(cc != NULL);
1153        break;
1154        }
1155    
1156      /* Character iterators, which are not inside a repeated bracket,
1157         gets a private slot instead of allocating it on the stack. */
1158      if (space > 0 && cc >= end)
1159        {
1160        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1161        private_data_ptr += sizeof(sljit_sw) * space;
1162        }
1163    
1164      if (size != 0)
1165        {
1166        if (size < 0)
1167          {
1168          cc += -size;
1169    #ifdef SUPPORT_UTF
1170          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1171    #endif
1172          }
1173        else
1174          cc += size;
1175        }
1176    
1177      if (bracketlen > 0)
1178        {
1179        if (cc >= end)
1180          {
1181          end = bracketend(cc);
1182          if (end[-1 - LINK_SIZE] == OP_KET)
1183            end = NULL;
1184          }
1185        cc += bracketlen;
1186        }
1187      }
1188    *private_data_start = private_data_ptr;
1189    }
1190    
1191    /* Returns with a frame_types (always < 0) if no need for frame. */
1192    static int get_framesize(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, BOOL recursive, BOOL* needs_control_head)
1193    {
1194    int length = 0;
1195    int possessive = 0;
1196    BOOL stack_restore = FALSE;
1197    BOOL setsom_found = recursive;
1198    BOOL setmark_found = recursive;
1199    /* The last capture is a local variable even for recursions. */
1200    BOOL capture_last_found = FALSE;
1201    
1202    #if defined DEBUG_FORCE_CONTROL_HEAD && DEBUG_FORCE_CONTROL_HEAD
1203    SLJIT_ASSERT(common->control_head_ptr != 0);
1204    *needs_control_head = TRUE;
1205    #else
1206    *needs_control_head = FALSE;
1207    #endif
1208    
1209    if (ccend == NULL)
1210      {
1211      ccend = bracketend(cc) - (1 + LINK_SIZE);
1212      if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))
1213        {
1214        possessive = length = (common->capture_last_ptr != 0) ? 5 : 3;
1215        /* This is correct regardless of common->capture_last_ptr. */
1216        capture_last_found = TRUE;
1217        }
1218    cc = next_opcode(common, cc);    cc = next_opcode(common, cc);
1219      }
1220    
1221  SLJIT_ASSERT(cc != NULL);  SLJIT_ASSERT(cc != NULL);
1222  while (cc < ccend)  while (cc < ccend)
1223    switch(*cc)    switch(*cc)
1224      {      {
1225      case OP_SET_SOM:      case OP_SET_SOM:
1226        SLJIT_ASSERT(common->has_set_som);
1227        stack_restore = TRUE;
1228        if (!setsom_found)
1229          {
1230          length += 2;
1231          setsom_found = TRUE;
1232          }
1233        cc += 1;
1234        break;
1235    
1236        case OP_MARK:
1237        case OP_PRUNE_ARG:
1238        case OP_THEN_ARG:
1239        SLJIT_ASSERT(common->mark_ptr != 0);
1240        stack_restore = TRUE;
1241        if (!setmark_found)
1242          {
1243          length += 2;
1244          setmark_found = TRUE;
1245          }
1246        if (common->control_head_ptr != 0)
1247          *needs_control_head = TRUE;
1248        cc += 1 + 2 + cc[1];
1249        break;
1250    
1251      case OP_RECURSE:      case OP_RECURSE:
1252        stack_restore = TRUE;
1253        if (common->has_set_som && !setsom_found)
1254          {
1255          length += 2;
1256          setsom_found = TRUE;
1257          }
1258        if (common->mark_ptr != 0 && !setmark_found)
1259          {
1260          length += 2;
1261          setmark_found = TRUE;
1262          }
1263        if (common->capture_last_ptr != 0 && !capture_last_found)
1264          {
1265          length += 2;
1266          capture_last_found = TRUE;
1267          }
1268        cc += 1 + LINK_SIZE;
1269        break;
1270    
1271        case OP_CBRA:
1272        case OP_CBRAPOS:
1273        case OP_SCBRA:
1274        case OP_SCBRAPOS:
1275        stack_restore = TRUE;
1276        if (common->capture_last_ptr != 0 && !capture_last_found)
1277          {
1278          length += 2;
1279          capture_last_found = TRUE;
1280          }
1281        length += 3;
1282        cc += 1 + LINK_SIZE + IMM2_SIZE;
1283        break;
1284    
1285        default:
1286        stack_restore = TRUE;
1287        /* Fall through. */
1288    
1289        case OP_NOT_WORD_BOUNDARY:
1290        case OP_WORD_BOUNDARY:
1291        case OP_NOT_DIGIT:
1292        case OP_DIGIT:
1293        case OP_NOT_WHITESPACE:
1294        case OP_WHITESPACE:
1295        case OP_NOT_WORDCHAR:
1296        case OP_WORDCHAR:
1297        case OP_ANY:
1298        case OP_ALLANY:
1299        case OP_ANYBYTE:
1300        case OP_NOTPROP:
1301        case OP_PROP:
1302        case OP_ANYNL:
1303        case OP_NOT_HSPACE:
1304        case OP_HSPACE:
1305        case OP_NOT_VSPACE:
1306        case OP_VSPACE:
1307        case OP_EXTUNI:
1308        case OP_EODN:
1309        case OP_EOD:
1310        case OP_CIRC:
1311        case OP_CIRCM:
1312        case OP_DOLL:
1313        case OP_DOLLM:
1314        case OP_CHAR:
1315        case OP_CHARI:
1316        case OP_NOT:
1317        case OP_NOTI:
1318    
1319        case OP_EXACT:
1320        case OP_POSSTAR:
1321        case OP_POSPLUS:
1322        case OP_POSQUERY:
1323        case OP_POSUPTO:
1324    
1325        case OP_EXACTI:
1326        case OP_POSSTARI:
1327        case OP_POSPLUSI:
1328        case OP_POSQUERYI:
1329        case OP_POSUPTOI:
1330    
1331        case OP_NOTEXACT:
1332        case OP_NOTPOSSTAR:
1333        case OP_NOTPOSPLUS:
1334        case OP_NOTPOSQUERY:
1335        case OP_NOTPOSUPTO:
1336    
1337        case OP_NOTEXACTI:
1338        case OP_NOTPOSSTARI:
1339        case OP_NOTPOSPLUSI:
1340        case OP_NOTPOSQUERYI:
1341        case OP_NOTPOSUPTOI:
1342    
1343        case OP_TYPEEXACT:
1344        case OP_TYPEPOSSTAR:
1345        case OP_TYPEPOSPLUS:
1346        case OP_TYPEPOSQUERY:
1347        case OP_TYPEPOSUPTO:
1348    
1349        case OP_CLASS:
1350        case OP_NCLASS:
1351        case OP_XCLASS:
1352    
1353        cc = next_opcode(common, cc);
1354        SLJIT_ASSERT(cc != NULL);
1355        break;
1356        }
1357    
1358    /* Possessive quantifiers can use a special case. */
1359    if (SLJIT_UNLIKELY(possessive == length))
1360      return stack_restore ? no_frame : no_stack;
1361    
1362    if (length > 0)
1363      return length + 1;
1364    return stack_restore ? no_frame : no_stack;
1365    }
1366    
1367    static void init_frame(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, int stackpos, int stacktop, BOOL recursive)
1368    {
1369    DEFINE_COMPILER;
1370    BOOL setsom_found = recursive;
1371    BOOL setmark_found = recursive;
1372    /* The last capture is a local variable even for recursions. */
1373    BOOL capture_last_found = FALSE;
1374    int offset;
1375    
1376    /* >= 1 + shortest item size (2) */
1377    SLJIT_UNUSED_ARG(stacktop);
1378    SLJIT_ASSERT(stackpos >= stacktop + 2);
1379    
1380    stackpos = STACK(stackpos);
1381    if (ccend == NULL)
1382      {
1383      ccend = bracketend(cc) - (1 + LINK_SIZE);
1384      if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))
1385        cc = next_opcode(common, cc);
1386      }
1387    
1388    SLJIT_ASSERT(cc != NULL);
1389    while (cc < ccend)
1390      switch(*cc)
1391        {
1392        case OP_SET_SOM:
1393        SLJIT_ASSERT(common->has_set_som);
1394      if (!setsom_found)      if (!setsom_found)
1395        {        {
1396        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));
1397        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setstrbegin);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1398        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1399        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1400        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1401        setsom_found = TRUE;        setsom_found = TRUE;
1402        }        }
1403      cc += (*cc == OP_SET_SOM) ? 1 : 1 + LINK_SIZE;      cc += 1;
1404        break;
1405    
1406        case OP_MARK:
1407        case OP_PRUNE_ARG:
1408        case OP_THEN_ARG:
1409        SLJIT_ASSERT(common->mark_ptr != 0);
1410        if (!setmark_found)
1411          {
1412          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);
1413          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
1414          stackpos += (int)sizeof(sljit_sw);
1415          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1416          stackpos += (int)sizeof(sljit_sw);
1417          setmark_found = TRUE;
1418          }
1419        cc += 1 + 2 + cc[1];
1420        break;
1421    
1422        case OP_RECURSE:
1423        if (common->has_set_som && !setsom_found)
1424          {
1425          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));
1426          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1427          stackpos += (int)sizeof(sljit_sw);
1428          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1429          stackpos += (int)sizeof(sljit_sw);
1430          setsom_found = TRUE;
1431          }
1432        if (common->mark_ptr != 0 && !setmark_found)
1433          {
1434          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);
1435          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
1436          stackpos += (int)sizeof(sljit_sw);
1437          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1438          stackpos += (int)sizeof(sljit_sw);
1439          setmark_found = TRUE;
1440          }
1441        if (common->capture_last_ptr != 0 && !capture_last_found)
1442          {
1443          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->capture_last_ptr);
1444          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1445          stackpos += (int)sizeof(sljit_sw);
1446          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1447          stackpos += (int)sizeof(sljit_sw);
1448          capture_last_found = TRUE;
1449          }
1450        cc += 1 + LINK_SIZE;
1451      break;      break;
1452    
1453      case OP_CBRA:      case OP_CBRA:
1454      case OP_CBRAPOS:      case OP_CBRAPOS:
1455      case OP_SCBRA:      case OP_SCBRA:
1456      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1457        if (common->capture_last_ptr != 0 && !capture_last_found)
1458          {
1459          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->capture_last_ptr);
1460          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1461          stackpos += (int)sizeof(sljit_sw);
1462          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1463          stackpos += (int)sizeof(sljit_sw);
1464          capture_last_found = TRUE;
1465          }
1466      offset = (GET2(cc, 1 + LINK_SIZE)) << 1;      offset = (GET2(cc, 1 + LINK_SIZE)) << 1;
1467      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));
1468      stackpos += (int)sizeof(sljit_w);      stackpos += (int)sizeof(sljit_sw);
1469      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset));      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset));
1470      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset + 1));      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset + 1));
1471      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1472      stackpos += (int)sizeof(sljit_w);      stackpos += (int)sizeof(sljit_sw);
1473      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);
1474      stackpos += (int)sizeof(sljit_w);      stackpos += (int)sizeof(sljit_sw);
1475    
1476      cc += 1 + LINK_SIZE + 2;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1477      break;      break;
1478    
1479      default:      default:
# Line 811  while (cc < ccend) Line 1482  while (cc < ccend)
1482      break;      break;
1483      }      }
1484    
1485  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_end);  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, 0);
1486  SLJIT_ASSERT(stackpos == STACK(stacktop));  SLJIT_ASSERT(stackpos == STACK(stacktop));
1487  }  }
1488    
1489  static SLJIT_INLINE int get_localsize(compiler_common *common, 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)
1490  {  {
1491  int localsize = 2;  int private_data_length = needs_control_head ? 3 : 2;
1492  uschar *alternative;  int size;
1493  /* Calculate the sum of the local variables. */  pcre_uchar *alternative;
1494    /* Calculate the sum of the private machine words. */
1495  while (cc < ccend)  while (cc < ccend)
1496    {    {
1497      size = 0;
1498    switch(*cc)    switch(*cc)
1499      {      {
1500        case OP_KET:
1501        if (PRIVATE_DATA(cc) != 0)
1502          private_data_length++;
1503        cc += 1 + LINK_SIZE;
1504        break;
1505    
1506      case OP_ASSERT:      case OP_ASSERT:
1507      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1508      case OP_ASSERTBACK:      case OP_ASSERTBACK:
# Line 834  while (cc < ccend) Line 1513  while (cc < ccend)
1513      case OP_SBRA:      case OP_SBRA:
1514      case OP_SBRAPOS:      case OP_SBRAPOS:
1515      case OP_SCOND:      case OP_SCOND:
1516      localsize++;      private_data_length++;
1517      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1518      break;      break;
1519    
1520      case OP_CBRA:      case OP_CBRA:
1521      case OP_SCBRA:      case OP_SCBRA:
1522      localsize++;      if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1523      cc += 1 + LINK_SIZE + 2;        private_data_length++;
1524        cc += 1 + LINK_SIZE + IMM2_SIZE;
1525      break;      break;
1526    
1527      case OP_CBRAPOS:      case OP_CBRAPOS:
1528      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1529      localsize += 2;      private_data_length += 2;
1530      cc += 1 + LINK_SIZE + 2;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1531      break;      break;
1532    
1533      case OP_COND:      case OP_COND:
1534      /* Might be a hidden SCOND. */      /* Might be a hidden SCOND. */
1535      alternative = cc + GET(cc, 1);      alternative = cc + GET(cc, 1);
1536      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1537        localsize++;        private_data_length++;
1538      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1539      break;      break;
1540    
1541        CASE_ITERATOR_PRIVATE_DATA_1
1542        if (PRIVATE_DATA(cc))
1543          private_data_length++;
1544        cc += 2;
1545    #ifdef SUPPORT_UTF
1546        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1547    #endif
1548        break;
1549    
1550        CASE_ITERATOR_PRIVATE_DATA_2A
1551        if (PRIVATE_DATA(cc))
1552          private_data_length += 2;
1553        cc += 2;
1554    #ifdef SUPPORT_UTF
1555        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1556    #endif
1557        break;
1558    
1559        CASE_ITERATOR_PRIVATE_DATA_2B
1560        if (PRIVATE_DATA(cc))
1561          private_data_length += 2;
1562        cc += 2 + IMM2_SIZE;
1563    #ifdef SUPPORT_UTF
1564        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1565    #endif
1566        break;
1567    
1568        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1569        if (PRIVATE_DATA(cc))
1570          private_data_length++;
1571        cc += 1;
1572        break;
1573    
1574        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1575        if (PRIVATE_DATA(cc))
1576          private_data_length += 2;
1577        cc += 1;
1578        break;
1579    
1580        CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1581        if (PRIVATE_DATA(cc))
1582          private_data_length += 2;
1583        cc += 1 + IMM2_SIZE;
1584        break;
1585    
1586        case OP_CLASS:
1587        case OP_NCLASS:
1588    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1589        case OP_XCLASS:
1590        size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
1591    #else
1592        size = 1 + 32 / (int)sizeof(pcre_uchar);
1593    #endif
1594        if (PRIVATE_DATA(cc))
1595          private_data_length += get_class_iterator_size(cc + size);
1596        cc += size;
1597        break;
1598    
1599      default:      default:
1600      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
1601      SLJIT_ASSERT(cc != NULL);      SLJIT_ASSERT(cc != NULL);
# Line 865  while (cc < ccend) Line 1603  while (cc < ccend)
1603      }      }
1604    }    }
1605  SLJIT_ASSERT(cc == ccend);  SLJIT_ASSERT(cc == ccend);
1606  return localsize;  return private_data_length;
1607  }  }
1608    
1609  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,
1610    BOOL save, int stackptr, int stacktop)    BOOL save, int stackptr, int stacktop, BOOL needs_control_head)
1611  {  {
1612  DEFINE_COMPILER;  DEFINE_COMPILER;
1613  int srcw[2];  int srcw[2];
1614  int count;  int count, size;
1615  BOOL tmp1next = TRUE;  BOOL tmp1next = TRUE;
1616  BOOL tmp1empty = TRUE;  BOOL tmp1empty = TRUE;
1617  BOOL tmp2empty = TRUE;  BOOL tmp2empty = TRUE;
1618  uschar *alternative;  pcre_uchar *alternative;
1619  enum {  enum {
1620    start,    start,
1621    loop,    loop,
# Line 890  stacktop = STACK(stacktop - 1); Line 1628  stacktop = STACK(stacktop - 1);
1628    
1629  if (!save)  if (!save)
1630    {    {
1631    stackptr += sizeof(sljit_w);    stackptr += (needs_control_head ? 2 : 1) * sizeof(sljit_sw);
1632    if (stackptr < stacktop)    if (stackptr < stacktop)
1633      {      {
1634      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1635      stackptr += sizeof(sljit_w);      stackptr += sizeof(sljit_sw);
1636      tmp1empty = FALSE;      tmp1empty = FALSE;
1637      }      }
1638    if (stackptr < stacktop)    if (stackptr < stacktop)
1639      {      {
1640      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1641      stackptr += sizeof(sljit_w);      stackptr += sizeof(sljit_sw);
1642      tmp2empty = FALSE;      tmp2empty = FALSE;
1643      }      }
1644    /* The tmp1next must be TRUE in either way. */    /* The tmp1next must be TRUE in either way. */
1645    }    }
1646    
1647  while (status != end)  do
1648    {    {
1649    count = 0;    count = 0;
1650    switch(status)    switch(status)
1651      {      {
1652      case start:      case start:
1653      SLJIT_ASSERT(save);      SLJIT_ASSERT(save && common->recursive_head_ptr != 0);
1654      count = 1;      count = 1;
1655      srcw[0] = RECURSIVE_HEAD;      srcw[0] = common->recursive_head_ptr;
1656        if (needs_control_head)
1657          {
1658          SLJIT_ASSERT(common->control_head_ptr != 0);
1659          count = 2;
1660          srcw[1] = common->control_head_ptr;
1661          }
1662      status = loop;      status = loop;
1663      break;      break;
1664    
# Line 927  while (status != end) Line 1671  while (status != end)
1671    
1672      switch(*cc)      switch(*cc)
1673        {        {
1674          case OP_KET:
1675          if (PRIVATE_DATA(cc) != 0)
1676            {
1677            count = 1;
1678            srcw[0] = PRIVATE_DATA(cc);
1679            }
1680          cc += 1 + LINK_SIZE;
1681          break;
1682    
1683        case OP_ASSERT:        case OP_ASSERT:
1684        case OP_ASSERT_NOT:        case OP_ASSERT_NOT:
1685        case OP_ASSERTBACK:        case OP_ASSERTBACK:
# Line 938  while (status != end) Line 1691  while (status != end)
1691        case OP_SBRAPOS:        case OP_SBRAPOS:
1692        case OP_SCOND:        case OP_SCOND:
1693        count = 1;        count = 1;
1694        srcw[0] = PRIV(cc);        srcw[0] = PRIVATE_DATA(cc);
1695        SLJIT_ASSERT(srcw[0] != 0);        SLJIT_ASSERT(srcw[0] != 0);
1696        cc += 1 + LINK_SIZE;        cc += 1 + LINK_SIZE;
1697        break;        break;
1698    
1699        case OP_CBRA:        case OP_CBRA:
1700        case OP_SCBRA:        case OP_SCBRA:
1701        count = 1;        if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1702        srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));          {
1703        cc += 1 + LINK_SIZE + 2;          count = 1;
1704            srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1705            }
1706          cc += 1 + LINK_SIZE + IMM2_SIZE;
1707        break;        break;
1708    
1709        case OP_CBRAPOS:        case OP_CBRAPOS:
1710        case OP_SCBRAPOS:        case OP_SCBRAPOS:
1711        count = 2;        count = 2;
1712          srcw[0] = PRIVATE_DATA(cc);
1713        srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));        srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1714        srcw[0] = PRIV(cc);        SLJIT_ASSERT(srcw[0] != 0 && srcw[1] != 0);
1715        SLJIT_ASSERT(srcw[0] != 0);        cc += 1 + LINK_SIZE + IMM2_SIZE;
       cc += 1 + LINK_SIZE + 2;  
1716        break;        break;
1717    
1718        case OP_COND:        case OP_COND:
# Line 965  while (status != end) Line 1721  while (status != end)
1721        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1722          {          {
1723          count = 1;          count = 1;
1724          srcw[0] = PRIV(cc);          srcw[0] = PRIVATE_DATA(cc);
1725          SLJIT_ASSERT(srcw[0] != 0);          SLJIT_ASSERT(srcw[0] != 0);
1726          }          }
1727        cc += 1 + LINK_SIZE;        cc += 1 + LINK_SIZE;
1728        break;        break;
1729    
1730          CASE_ITERATOR_PRIVATE_DATA_1
1731          if (PRIVATE_DATA(cc))
1732            {
1733            count = 1;
1734            srcw[0] = PRIVATE_DATA(cc);
1735            }
1736          cc += 2;
1737    #ifdef SUPPORT_UTF
1738          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1739    #endif
1740          break;
1741    
1742          CASE_ITERATOR_PRIVATE_DATA_2A
1743          if (PRIVATE_DATA(cc))
1744            {
1745            count = 2;
1746            srcw[0] = PRIVATE_DATA(cc);
1747            srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1748            }
1749          cc += 2;
1750    #ifdef SUPPORT_UTF
1751          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1752    #endif
1753          break;
1754    
1755          CASE_ITERATOR_PRIVATE_DATA_2B
1756          if (PRIVATE_DATA(cc))
1757            {
1758            count = 2;
1759            srcw[0] = PRIVATE_DATA(cc);
1760            srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1761            }
1762          cc += 2 + IMM2_SIZE;
1763    #ifdef SUPPORT_UTF
1764          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1765    #endif
1766          break;
1767    
1768          CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1769          if (PRIVATE_DATA(cc))
1770            {
1771            count = 1;
1772            srcw[0] = PRIVATE_DATA(cc);
1773            }
1774          cc += 1;
1775          break;
1776    
1777          CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1778          if (PRIVATE_DATA(cc))
1779            {
1780            count = 2;
1781            srcw[0] = PRIVATE_DATA(cc);
1782            srcw[1] = srcw[0] + sizeof(sljit_sw);
1783            }
1784          cc += 1;
1785          break;
1786    
1787          CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1788          if (PRIVATE_DATA(cc))
1789            {
1790            count = 2;
1791            srcw[0] = PRIVATE_DATA(cc);
1792            srcw[1] = srcw[0] + sizeof(sljit_sw);
1793            }
1794          cc += 1 + IMM2_SIZE;
1795          break;
1796    
1797          case OP_CLASS:
1798          case OP_NCLASS:
1799    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1800          case OP_XCLASS:
1801          size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
1802    #else
1803          size = 1 + 32 / (int)sizeof(pcre_uchar);
1804    #endif
1805          if (PRIVATE_DATA(cc))
1806            switch(get_class_iterator_size(cc + size))
1807              {
1808              case 1:
1809              count = 1;
1810              srcw[0] = PRIVATE_DATA(cc);
1811              break;
1812    
1813              case 2:
1814              count = 2;
1815              srcw[0] = PRIVATE_DATA(cc);
1816              srcw[1] = srcw[0] + sizeof(sljit_sw);
1817              break;
1818    
1819              default:
1820              SLJIT_ASSERT_STOP();
1821              break;
1822              }
1823          cc += size;
1824          break;
1825    
1826        default:        default:
1827        cc = next_opcode(common, cc);        cc = next_opcode(common, cc);
1828        SLJIT_ASSERT(cc != NULL);        SLJIT_ASSERT(cc != NULL);
# Line 993  while (status != end) Line 1845  while (status != end)
1845          if (!tmp1empty)          if (!tmp1empty)
1846            {            {
1847            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1848            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1849            }            }
1850          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);
1851          tmp1empty = FALSE;          tmp1empty = FALSE;
# Line 1004  while (status != end) Line 1856  while (status != end)
1856          if (!tmp2empty)          if (!tmp2empty)
1857            {            {
1858            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1859            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1860            }            }
1861          OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);          OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);
1862          tmp2empty = FALSE;          tmp2empty = FALSE;
# Line 1021  while (status != end) Line 1873  while (status != end)
1873          if (!tmp1empty)          if (!tmp1empty)
1874            {            {
1875            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1876            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1877            }            }
1878          tmp1next = FALSE;          tmp1next = FALSE;
1879          }          }
# Line 1033  while (status != end) Line 1885  while (status != end)
1885          if (!tmp2empty)          if (!tmp2empty)
1886            {            {
1887            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1888            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1889            }            }
1890          tmp1next = TRUE;          tmp1next = TRUE;
1891          }          }
1892        }        }
1893      }      }
1894    }    }
1895    while (status != end);
1896    
1897  if (save)  if (save)
1898    {    {
# Line 1048  if (save) Line 1901  if (save)
1901      if (!tmp1empty)      if (!tmp1empty)
1902        {        {
1903        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1904        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1905        }        }
1906      if (!tmp2empty)      if (!tmp2empty)
1907        {        {
1908        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1909        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1910        }        }
1911      }      }
1912    else    else
# Line 1061  if (save) Line 1914  if (save)
1914      if (!tmp2empty)      if (!tmp2empty)
1915        {        {
1916        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1917        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1918        }        }
1919      if (!tmp1empty)      if (!tmp1empty)
1920        {        {
1921        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1922        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1923        }        }
1924      }      }
1925    }    }
1926  SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));  SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));
1927  }  }
1928    
1929  static SLJIT_INLINE BOOL ispowerof2(unsigned int value)  static SLJIT_INLINE pcre_uchar *set_then_offsets(compiler_common *common, pcre_uchar *cc, pcre_uint8 *current_offset)
1930    {
1931    pcre_uchar *end = bracketend(cc);
1932    BOOL has_alternatives = cc[GET(cc, 1)] == OP_ALT;
1933    
1934    /* Assert captures then. */
1935    if (*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT)
1936      current_offset = NULL;
1937    /* Conditional block does not. */
1938    if (*cc == OP_COND || *cc == OP_SCOND)
1939      has_alternatives = FALSE;
1940    
1941    cc = next_opcode(common, cc);
1942    if (has_alternatives)
1943      current_offset = common->then_offsets + (cc - common->start);
1944    
1945    while (cc < end)
1946      {
1947      if ((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND))
1948        cc = set_then_offsets(common, cc, current_offset);
1949      else
1950        {
1951        if (*cc == OP_ALT && has_alternatives)
1952          current_offset = common->then_offsets + (cc + 1 + LINK_SIZE - common->start);
1953        if (*cc >= OP_THEN && *cc <= OP_THEN_ARG && current_offset != NULL)
1954          *current_offset = 1;
1955        cc = next_opcode(common, cc);
1956        }
1957      }
1958    
1959    return end;
1960    }
1961    
1962    #undef CASE_ITERATOR_PRIVATE_DATA_1
1963    #undef CASE_ITERATOR_PRIVATE_DATA_2A
1964    #undef CASE_ITERATOR_PRIVATE_DATA_2B
1965    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1966    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1967    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1968    
1969    static SLJIT_INLINE BOOL is_powerof2(unsigned int value)
1970  {  {
1971  return (value & (value - 1)) == 0;  return (value & (value - 1)) == 0;
1972  }  }
# Line 1083  static SLJIT_INLINE void set_jumps(jump_ Line 1976  static SLJIT_INLINE void set_jumps(jump_
1976  while (list)  while (list)
1977    {    {
1978    /* sljit_set_label is clever enough to do nothing    /* sljit_set_label is clever enough to do nothing
1979    if either the jump or the label is NULL */    if either the jump or the label is NULL. */
1980    sljit_set_label(list->jump, label);    SET_LABEL(list->jump, label);
1981    list = list->next;    list = list->next;
1982    }    }
1983  }  }
# Line 1100  if (list_item) Line 1993  if (list_item)
1993    }    }
1994  }  }
1995    
1996  static void add_stub(compiler_common *common, enum stub_types type, int data, struct sljit_jump *start)  static void add_stub(compiler_common *common, struct sljit_jump *start)
1997  {  {
1998  DEFINE_COMPILER;  DEFINE_COMPILER;
1999  stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));  stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));
2000    
2001  if (list_item)  if (list_item)
2002    {    {
   list_item->type = type;  
   list_item->data = data;  
2003    list_item->start = start;    list_item->start = start;
2004    list_item->leave = LABEL();    list_item->quit = LABEL();
2005    list_item->next = common->stubs;    list_item->next = common->stubs;
2006    common->stubs = list_item;    common->stubs = list_item;
2007    }    }
# Line 1124  stub_list* list_item = common->stubs; Line 2015  stub_list* list_item = common->stubs;
2015  while (list_item)  while (list_item)
2016    {    {
2017    JUMPHERE(list_item->start);    JUMPHERE(list_item->start);
2018    switch(list_item->type)    add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));
2019      {    JUMPTO(SLJIT_JUMP, list_item->quit);
     case stack_alloc:  
     add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));  
     break;  
     }  
   JUMPTO(SLJIT_JUMP, list_item->leave);  
2020    list_item = list_item->next;    list_item = list_item->next;
2021    }    }
2022  common->stubs = NULL;  common->stubs = NULL;
# Line 1149  static SLJIT_INLINE void allocate_stack( Line 2035  static SLJIT_INLINE void allocate_stack(
2035  /* May destroy all locals and registers except TMP2. */  /* May destroy all locals and registers except TMP2. */
2036  DEFINE_COMPILER;  DEFINE_COMPILER;
2037    
2038  OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_w));  OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_sw));
2039  #ifdef DESTROY_REGISTERS  #ifdef DESTROY_REGISTERS
2040  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);
2041  OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);  OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);
# Line 1157  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0); Line 2043  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);
2043  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);
2044  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);
2045  #endif  #endif
2046  add_stub(common, stack_alloc, 0, CMP(SLJIT_C_GREATER, STACK_TOP, 0, STACK_LIMIT, 0));  add_stub(common, CMP(SLJIT_C_GREATER, STACK_TOP, 0, STACK_LIMIT, 0));
2047  }  }
2048    
2049  static SLJIT_INLINE void free_stack(compiler_common *common, int size)  static SLJIT_INLINE void free_stack(compiler_common *common, int size)
2050  {  {
2051  DEFINE_COMPILER;  DEFINE_COMPILER;
2052  OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_w));  OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_sw));
2053    }
2054    
2055    static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)
2056    {
2057    DEFINE_COMPILER;
2058    struct sljit_label *loop;
2059    int i;
2060    
2061    /* At this point we can freely use all temporary registers. */
2062    SLJIT_ASSERT(length > 1);
2063    /* TMP1 returns with begin - 1. */
2064    OP2(SLJIT_SUB, SLJIT_SCRATCH_REG1, 0, SLJIT_MEM1(SLJIT_SAVED_REG1), SLJIT_OFFSETOF(jit_arguments, begin), SLJIT_IMM, IN_UCHARS(1));
2065    if (length < 8)
2066      {
2067      for (i = 1; i < length; i++)
2068        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), SLJIT_SCRATCH_REG1, 0);
2069      }
2070    else
2071      {
2072      GET_LOCAL_BASE(SLJIT_SCRATCH_REG2, 0, OVECTOR_START);
2073      OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, length - 1);
2074      loop = LABEL();
2075      OP1(SLJIT_MOVU, SLJIT_MEM1(SLJIT_SCRATCH_REG2), sizeof(sljit_sw), SLJIT_SCRATCH_REG1, 0);
2076      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 1);
2077      JUMPTO(SLJIT_C_NOT_ZERO, loop);
2078      }
2079  }  }
2080    
2081  static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)  static SLJIT_INLINE void do_reset_match(compiler_common *common, int length)
2082  {  {
2083  DEFINE_COMPILER;  DEFINE_COMPILER;
2084  struct sljit_label *loop;  struct sljit_label *loop;
2085  int i;  int i;
2086  /* At this point we can freely use all temporary registers. */  
2087  /* TMP1 returns with begin - 1. */  SLJIT_ASSERT(length > 1);
2088  OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG1, 0, SLJIT_MEM1(SLJIT_GENERAL_REG1), SLJIT_OFFSETOF(jit_arguments, begin), SLJIT_IMM, 1);  /* OVECTOR(1) contains the "string begin - 1" constant. */
2089    if (length > 2)
2090      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));
2091  if (length < 8)  if (length < 8)
2092    {    {
2093    for (i = 0; i < length; i++)    for (i = 2; i < length; i++)
2094      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), SLJIT_TEMPORARY_REG1, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), TMP1, 0);
2095    }    }
2096  else  else
2097    {    {
2098    OP2(SLJIT_ADD, SLJIT_TEMPORARY_REG2, 0, SLJIT_LOCALS_REG, 0, SLJIT_IMM, OVECTOR_START - sizeof(sljit_w));    GET_LOCAL_BASE(TMP2, 0, OVECTOR_START + sizeof(sljit_sw));
2099    OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, length);    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_IMM, length - 2);
2100    loop = LABEL();    loop = LABEL();
2101    OP1(SLJIT_MOVU, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), sizeof(sljit_w), SLJIT_TEMPORARY_REG1, 0);    OP1(SLJIT_MOVU, SLJIT_MEM1(TMP2), sizeof(sljit_sw), TMP1, 0);
2102    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_TEMPORARY_REG3, 0, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, 1);    OP2(SLJIT_SUB | SLJIT_SET_E, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, 1);
2103    JUMPTO(SLJIT_C_NOT_ZERO, loop);    JUMPTO(SLJIT_C_NOT_ZERO, loop);
2104    }    }
2105    
2106    OP1(SLJIT_MOV, STACK_TOP, 0, ARGUMENTS, 0);
2107    if (common->mark_ptr != 0)
2108      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr, SLJIT_IMM, 0);
2109    if (common->control_head_ptr != 0)
2110      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->control_head_ptr, SLJIT_IMM, 0);
2111    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(jit_arguments, stack));
2112    OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_ptr);
2113    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(struct sljit_stack, base));
2114    }
2115    
2116    static sljit_sw SLJIT_CALL do_search_mark(sljit_sw *current, const pcre_uchar *skip_arg)
2117    {
2118    while (current != NULL)
2119      {
2120      switch (current[-2])
2121        {
2122        case type_then_trap:
2123        break;
2124    
2125        case type_mark:
2126        if (STRCMP_UC_UC(skip_arg, (pcre_uchar *)current[-3]) == 0)
2127          return current[-4];
2128        break;
2129    
2130        default:
2131        SLJIT_ASSERT_STOP();
2132        break;
2133        }
2134      current = (sljit_sw*)current[-1];
2135      }
2136    return -1;
2137  }  }
2138    
2139  static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)  static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)
2140  {  {
2141  DEFINE_COMPILER;  DEFINE_COMPILER;
2142  struct sljit_label *loop;  struct sljit_label *loop;
2143  struct sljit_jump *earlyexit;  struct sljit_jump *early_quit;
2144    
2145  /* At this point we can freely use all registers. */  /* At this point we can freely use all registers. */
2146  OP1(SLJIT_MOV, SLJIT_GENERAL_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));  OP1(SLJIT_MOV, SLJIT_SAVED_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));
2147  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1), STR_PTR, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1), STR_PTR, 0);
2148    
2149  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, ARGUMENTS, 0);  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG1, 0, ARGUMENTS, 0);
2150  OP1(SLJIT_MOV_SI, SLJIT_TEMPORARY_REG2, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, offsetcount));  if (common->mark_ptr != 0)
2151  OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);
2152  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, begin));  OP1(SLJIT_MOV_SI, SLJIT_SCRATCH_REG2, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, offset_count));
2153  OP2(SLJIT_ADD, SLJIT_GENERAL_REG1, 0, SLJIT_LOCALS_REG, 0, SLJIT_IMM, OVECTOR_START);  if (common->mark_ptr != 0)
2154      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, mark_ptr), SLJIT_SCRATCH_REG3, 0);
2155    OP2(SLJIT_SUB, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));
2156    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG1, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, begin));
2157    GET_LOCAL_BASE(SLJIT_SAVED_REG1, 0, OVECTOR_START);
2158  /* Unlikely, but possible */  /* Unlikely, but possible */
2159  earlyexit = CMP(SLJIT_C_EQUAL, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, 0);  early_quit = CMP(SLJIT_C_EQUAL, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, 0);
2160  loop = LABEL();  loop = LABEL();
2161  OP2(SLJIT_SUB, SLJIT_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);
2162  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));
2163  /* Copy the integer value to the output buffer */  /* Copy the integer value to the output buffer */
2164  OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_TEMPORARY_REG3), sizeof(int), SLJIT_GENERAL_REG2, 0);  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2165  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);
2166    #endif
2167    OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG3), sizeof(int), SLJIT_SAVED_REG2, 0);
2168    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_SCRATCH_REG2, 0, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, 1);
2169  JUMPTO(SLJIT_C_NOT_ZERO, loop);  JUMPTO(SLJIT_C_NOT_ZERO, loop);
2170  JUMPHERE(earlyexit);  JUMPHERE(early_quit);
2171    
2172  /* Calculate the return value, which is the maximum ovector value. */  /* Calculate the return value, which is the maximum ovector value. */
2173  if (topbracket > 1)  if (topbracket > 1)
2174    {    {
2175    OP2(SLJIT_ADD, SLJIT_TEMPORARY_REG1, 0, SLJIT_LOCALS_REG, 0, SLJIT_IMM, OVECTOR_START + topbracket * 2 * sizeof(sljit_w));    GET_LOCAL_BASE(SLJIT_SCRATCH_REG1, 0, OVECTOR_START + topbracket * 2 * sizeof(sljit_sw));
2176    OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, topbracket + 1);    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, topbracket + 1);
2177    
2178    /* OVECTOR(0) is never equal to SLJIT_GENERAL_REG3. */    /* OVECTOR(0) is never equal to SLJIT_SAVED_REG3. */
2179    loop = LABEL();    loop = LABEL();
2180    OP1(SLJIT_MOVU, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), -(2 * (sljit_w)sizeof(sljit_w)));    OP1(SLJIT_MOVU, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), -(2 * (sljit_sw)sizeof(sljit_sw)));
2181    OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG2, 0, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, 1);    OP2(SLJIT_SUB, SLJIT_SCRATCH_REG2, 0, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, 1);
2182    CMPTO(SLJIT_C_EQUAL, SLJIT_TEMPORARY_REG3, 0, SLJIT_GENERAL_REG3, 0, loop);    CMPTO(SLJIT_C_EQUAL, SLJIT_SCRATCH_REG3, 0, SLJIT_SAVED_REG3, 0, loop);
2183    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_TEMPORARY_REG2, 0);    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_SCRATCH_REG2, 0);
2184    }    }
2185  else  else
2186    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, 1);    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, 1);
2187  }  }
2188    
2189  static SLJIT_INLINE BOOL char_has_othercase(compiler_common *common, uschar* cc)  static SLJIT_INLINE void return_with_partial_match(compiler_common *common, struct sljit_label *quit)
2190    {
2191    DEFINE_COMPILER;
2192    struct sljit_jump *jump;
2193    
2194    SLJIT_COMPILE_ASSERT(STR_END == SLJIT_SAVED_REG2, str_end_must_be_saved_reg2);
2195    SLJIT_ASSERT(common->start_used_ptr != 0 && common->start_ptr != 0
2196      && (common->mode == JIT_PARTIAL_SOFT_COMPILE ? common->hit_start != 0 : common->hit_start == 0));
2197    
2198    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, ARGUMENTS, 0);
2199    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, PCRE_ERROR_PARTIAL);
2200    OP1(SLJIT_MOV_SI, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, real_offset_count));
2201    CMPTO(SLJIT_C_SIG_LESS, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 2, quit);
2202    
2203    /* Store match begin and end. */
2204    OP1(SLJIT_MOV, SLJIT_SAVED_REG1, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, begin));
2205    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, offsets));
2206    
2207    jump = CMP(SLJIT_C_SIG_LESS, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 3);
2208    OP2(SLJIT_SUB, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mode == JIT_PARTIAL_HARD_COMPILE ? common->start_ptr : (common->hit_start + (int)sizeof(sljit_sw)), SLJIT_SAVED_REG1, 0);
2209    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2210    OP2(SLJIT_ASHR, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, UCHAR_SHIFT);
2211    #endif
2212    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG2), 2 * sizeof(int), SLJIT_SCRATCH_REG3, 0);
2213    JUMPHERE(jump);
2214    
2215    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mode == JIT_PARTIAL_HARD_COMPILE ? common->start_used_ptr : common->hit_start);
2216    OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, STR_END, 0, SLJIT_SAVED_REG1, 0);
2217    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2218    OP2(SLJIT_ASHR, SLJIT_SAVED_REG2, 0, SLJIT_SAVED_REG2, 0, SLJIT_IMM, UCHAR_SHIFT);
2219    #endif
2220    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG2), sizeof(int), SLJIT_SAVED_REG2, 0);
2221    
2222    OP2(SLJIT_SUB, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_SAVED_REG1, 0);
2223    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2224    OP2(SLJIT_ASHR, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, UCHAR_SHIFT);
2225    #endif
2226    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG2), 0, SLJIT_SCRATCH_REG3, 0);
2227    
2228    JUMPTO(SLJIT_JUMP, quit);
2229    }
2230    
2231    static SLJIT_INLINE void check_start_used_ptr(compiler_common *common)
2232    {
2233    /* May destroy TMP1. */
2234    DEFINE_COMPILER;
2235    struct sljit_jump *jump;
2236    
2237    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2238      {
2239      /* The value of -1 must be kept for start_used_ptr! */
2240      OP2(SLJIT_ADD, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, SLJIT_IMM, 1);
2241      /* Jumps if start_used_ptr < STR_PTR, or start_used_ptr == -1. Although overwriting
2242      is not necessary if start_used_ptr == STR_PTR, it does not hurt as well. */
2243      jump = CMP(SLJIT_C_LESS_EQUAL, TMP1, 0, STR_PTR, 0);
2244      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);
2245      JUMPHERE(jump);
2246      }
2247    else if (common->mode == JIT_PARTIAL_HARD_COMPILE)
2248      {
2249      jump = CMP(SLJIT_C_LESS_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);
2250      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);
2251      JUMPHERE(jump);
2252      }
2253    }
2254    
2255    static SLJIT_INLINE BOOL char_has_othercase(compiler_common *common, pcre_uchar* cc)
2256  {  {
2257  /* Detects if the character has an othercase. */  /* Detects if the character has an othercase. */
2258  unsigned int c;  unsigned int c;
2259    
2260  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2261  if (common->utf8)  if (common->utf)
2262    {    {
2263    GETCHAR(c, cc);    GETCHAR(c, cc);
2264    if (c > 127)    if (c > 127)
# Line 1250  if (common->utf8) Line 2269  if (common->utf8)
2269      return FALSE;      return FALSE;
2270  #endif  #endif
2271      }      }
2272    #ifndef COMPILE_PCRE8
2273      return common->fcc[c] != c;
2274    #endif
2275    }    }
2276  else  else
2277  #endif  #endif
2278    c = *cc;    c = *cc;
2279  return common->fcc[c] != c;  return MAX_255(c) ? common->fcc[c] != c : FALSE;
2280  }  }
2281    
2282  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)
2283  {  {
2284  /* Returns with the othercase. */  /* Returns with the othercase. */
2285  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2286  if (common->utf8 && c > 127)  if (common->utf && c > 127)
2287    {    {
2288  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2289    return UCD_OTHERCASE(c);    return UCD_OTHERCASE(c);
# Line 1270  if (common->utf8 && c > 127) Line 2292  if (common->utf8 && c > 127)
2292  #endif  #endif
2293    }    }
2294  #endif  #endif
2295  return common->fcc[c];  return TABLE_GET(c, common->fcc, c);
2296  }  }
2297    
2298  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)
2299  {  {
2300  /* Detects if the character and its othercase has only 1 bit difference. */  /* Detects if the character and its othercase has only 1 bit difference. */
2301  unsigned int c, oc, bit;  unsigned int c, oc, bit;
2302  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2303  int n;  int n;
2304  #endif  #endif
2305    
2306  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2307  if (common->utf8)  if (common->utf)
2308    {    {
2309    GETCHAR(c, cc);    GETCHAR(c, cc);
2310    if (c <= 127)    if (c <= 127)
# Line 1299  if (common->utf8) Line 2321  if (common->utf8)
2321  else  else
2322    {    {
2323    c = *cc;    c = *cc;
2324    oc = common->fcc[c];    oc = TABLE_GET(c, common->fcc, c);
2325    }    }
2326  #else  #else
2327  c = *cc;  c = *cc;
2328  oc = common->fcc[c];  oc = TABLE_GET(c, common->fcc, c);
2329  #endif  #endif
2330    
2331  SLJIT_ASSERT(c != oc);  SLJIT_ASSERT(c != oc);
# Line 1314  if (c <= 127 && bit == 0x20) Line 2336  if (c <= 127 && bit == 0x20)
2336    return (0 << 8) | 0x20;    return (0 << 8) | 0x20;
2337    
2338  /* Since c != oc, they must have at least 1 bit difference. */  /* Since c != oc, they must have at least 1 bit difference. */
2339  if (!ispowerof2(bit))  if (!is_powerof2(bit))
2340    return 0;    return 0;
2341    
2342  #ifdef SUPPORT_UTF8  #if defined COMPILE_PCRE8
2343  if (common->utf8 && c > 127)  
2344    #ifdef SUPPORT_UTF
2345    if (common->utf && c > 127)
2346    {    {
2347    n = _pcre_utf8_table4[*cc & 0x3f];    n = GET_EXTRALEN(*cc);
2348    while ((bit & 0x3f) == 0)    while ((bit & 0x3f) == 0)
2349      {      {
2350      n--;      n--;
# Line 1328  if (common->utf8 && c > 127) Line 2352  if (common->utf8 && c > 127)
2352      }      }
2353    return (n << 8) | bit;    return (n << 8) | bit;
2354    }    }
2355  #endif  #endif /* SUPPORT_UTF */
2356  return (0 << 8) | bit;  return (0 << 8) | bit;
2357    
2358    #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2359    
2360    #ifdef SUPPORT_UTF
2361    if (common->utf && c > 65535)
2362      {
2363      if (bit >= (1 << 10))
2364        bit >>= 10;
2365      else
2366        return (bit < 256) ? ((2 << 8) | bit) : ((3 << 8) | (bit >> 8));
2367      }
2368    #endif /* SUPPORT_UTF */
2369    return (bit < 256) ? ((0 << 8) | bit) : ((1 << 8) | (bit >> 8));
2370    
2371    #endif /* COMPILE_PCRE[8|16|32] */
2372    }
2373    
2374    static void check_partial(compiler_common *common, BOOL force)
2375    {
2376    /* Checks whether a partial matching is occured. Does not modify registers. */
2377    DEFINE_COMPILER;
2378    struct sljit_jump *jump = NULL;
2379    
2380    SLJIT_ASSERT(!force || common->mode != JIT_COMPILE);
2381    
2382    if (common->mode == JIT_COMPILE)
2383      return;
2384    
2385    if (!force)
2386      jump = CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);
2387    else if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2388      jump = CMP(SLJIT_C_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, SLJIT_IMM, -1);
2389    
2390    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2391      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2392    else
2393      {
2394      if (common->partialmatchlabel != NULL)
2395        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2396      else
2397        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2398      }
2399    
2400    if (jump != NULL)
2401      JUMPHERE(jump);
2402    }
2403    
2404    static void check_str_end(compiler_common *common, jump_list **end_reached)
2405    {
2406    /* Does not affect registers. Usually used in a tight spot. */
2407    DEFINE_COMPILER;
2408    struct sljit_jump *jump;
2409    
2410    if (common->mode == JIT_COMPILE)
2411      {
2412      add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2413      return;
2414      }
2415    
2416    jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);
2417    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2418      {
2419      add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));
2420      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2421      add_jump(compiler, end_reached, JUMP(SLJIT_JUMP));
2422      }
2423    else
2424      {
2425      add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));
2426      if (common->partialmatchlabel != NULL)
2427        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2428      else
2429        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2430      }
2431    JUMPHERE(jump);
2432  }  }
2433    
2434  static SLJIT_INLINE void check_input_end(compiler_common *common, jump_list **fallbacks)  static void detect_partial_match(compiler_common *common, jump_list **backtracks)
2435  {  {
2436  DEFINE_COMPILER;  DEFINE_COMPILER;
2437  add_jump(compiler, fallbacks, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));  struct sljit_jump *jump;
2438    
2439    if (common->mode == JIT_COMPILE)
2440      {
2441      add_jump(compiler, backtracks, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2442      return;
2443      }
2444    
2445    /* Partial matching mode. */
2446    jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);
2447    add_jump(compiler, backtracks, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));
2448    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2449      {
2450      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2451      add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));
2452      }
2453    else
2454      {
2455      if (common->partialmatchlabel != NULL)
2456        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2457      else
2458        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2459      }
2460    JUMPHERE(jump);
2461  }  }
2462    
2463  static void read_char(compiler_common *common)  static void read_char(compiler_common *common)
# Line 1343  static void read_char(compiler_common *c Line 2465  static void read_char(compiler_common *c
2465  /* Reads the character into TMP1, updates STR_PTR.  /* Reads the character into TMP1, updates STR_PTR.
2466  Does not check STR_END. TMP2 Destroyed. */  Does not check STR_END. TMP2 Destroyed. */
2467  DEFINE_COMPILER;  DEFINE_COMPILER;
2468  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2469  struct sljit_jump *jump;  struct sljit_jump *jump;
2470  #endif  #endif
2471    
2472  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2473  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2474  if (common->utf8)  if (common->utf)
2475    {    {
2476    #if defined COMPILE_PCRE8
2477    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2478    add_jump(compiler, &common->utf8readchar, JUMP(SLJIT_FAST_CALL));  #elif defined COMPILE_PCRE16
2479      jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);
2480    #endif /* COMPILE_PCRE[8|16] */
2481      add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2482    JUMPHERE(jump);    JUMPHERE(jump);
2483    }    }
2484  #endif  #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2485  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));
2486  }  }
2487    
2488  static void peek_char(compiler_common *common)  static void peek_char(compiler_common *common)
# Line 1364  static void peek_char(compiler_common *c Line 2490  static void peek_char(compiler_common *c
2490  /* Reads the character into TMP1, keeps STR_PTR.  /* Reads the character into TMP1, keeps STR_PTR.
2491  Does not check STR_END. TMP2 Destroyed. */  Does not check STR_END. TMP2 Destroyed. */
2492  DEFINE_COMPILER;  DEFINE_COMPILER;
2493  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2494  struct sljit_jump *jump;  struct sljit_jump *jump;
2495  #endif  #endif
2496    
2497  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2498  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2499  if (common->utf8)  if (common->utf)
2500    {    {
2501    #if defined COMPILE_PCRE8
2502    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2503    add_jump(compiler, &common->utf8readchar, JUMP(SLJIT_FAST_CALL));  #elif defined COMPILE_PCRE16
2504      jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);
2505    #endif /* COMPILE_PCRE[8|16] */
2506      add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2507    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2508    JUMPHERE(jump);    JUMPHERE(jump);
2509    }    }
2510  #endif  #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2511  }  }
2512    
2513  static void read_char8_type(compiler_common *common)  static void read_char8_type(compiler_common *common)
2514  {  {
2515  /* 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. */
2516  DEFINE_COMPILER;  DEFINE_COMPILER;
2517  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2518  struct sljit_jump *jump;  struct sljit_jump *jump;
2519  #endif  #endif
2520    
2521  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2522  if (common->utf8)  if (common->utf)
2523    {    {
2524    OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2525    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));
2526    #if defined COMPILE_PCRE8
2527    /* This can be an extra read in some situations, but hopefully    /* This can be an extra read in some situations, but hopefully
2528    it is a clever early read in most cases. */    it is needed in most cases. */
2529    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2530    jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);    jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2531    add_jump(compiler, &common->utf8readtype8, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->utfreadtype8, JUMP(SLJIT_FAST_CALL));
2532      JUMPHERE(jump);
2533    #elif defined COMPILE_PCRE16
2534      OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2535      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);
2536      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2537    JUMPHERE(jump);    JUMPHERE(jump);
2538      /* Skip low surrogate if necessary. */
2539      OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0xfc00);
2540      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0xd800);
2541      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
2542      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 1);
2543      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2544    #elif defined COMPILE_PCRE32
2545      OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2546      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);
2547      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2548      JUMPHERE(jump);
2549    #endif /* COMPILE_PCRE[8|16|32] */
2550    return;    return;
2551    }    }
2552    #endif /* SUPPORT_UTF */
2553    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2554    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2555    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2556    /* The ctypes array contains only 256 values. */
2557    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2558    jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);
2559    #endif
2560    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2561    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2562    JUMPHERE(jump);
2563  #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);  
2564  }  }
2565    
2566  static void skip_char_back(compiler_common *common)  static void skip_char_back(compiler_common *common)
2567  {  {
2568  /* 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. */
2569  DEFINE_COMPILER;  DEFINE_COMPILER;
2570  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2571    #if defined COMPILE_PCRE8
2572  struct sljit_label *label;  struct sljit_label *label;
2573    
2574  if (common->utf8)  if (common->utf)
2575    {    {
2576    label = LABEL();    label = LABEL();
2577    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
2578    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));
2579    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xc0);    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xc0);
2580    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, 0x80, label);    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, 0x80, label);
2581    return;    return;
2582    }    }
2583  #endif  #elif defined COMPILE_PCRE16
2584  OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);  if (common->utf)
2585      {
2586      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
2587      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2588      /* Skip low surrogate if necessary. */
2589      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
2590      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xdc00);
2591      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
2592      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
2593      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2594      return;
2595      }
2596    #endif /* COMPILE_PCRE[8|16] */
2597    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2598    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2599  }  }
2600    
2601  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)
2602  {  {
2603  /* 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. */
2604  DEFINE_COMPILER;  DEFINE_COMPILER;
# Line 1435  DEFINE_COMPILER; Line 2606  DEFINE_COMPILER;
2606  if (nltype == NLTYPE_ANY)  if (nltype == NLTYPE_ANY)
2607    {    {
2608    add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));
2609    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));
2610    }    }
2611  else if (nltype == NLTYPE_ANYCRLF)  else if (nltype == NLTYPE_ANYCRLF)
2612    {    {
2613    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);
2614    COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
2615    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);
2616    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);
2617    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));
2618    }    }
2619  else  else
2620    {    {
2621    SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline <= 255);    SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline < 256);
2622    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));
2623    }    }
2624  }  }
2625    
2626  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2627  static void do_utf8readchar(compiler_common *common)  
2628    #if defined COMPILE_PCRE8
2629    static void do_utfreadchar(compiler_common *common)
2630  {  {
2631  /* Fast decoding an utf8 character. TMP1 contains the first byte  /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2632  of the character (>= 0xc0). Return char value in TMP1, length - 1 in TMP2. */  of the character (>= 0xc0). Return char value in TMP1, length - 1 in TMP2. */
2633  DEFINE_COMPILER;  DEFINE_COMPILER;
2634  struct sljit_jump *jump;  struct sljit_jump *jump;
2635    
2636  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2637  /* Searching for the first zero. */  /* Searching for the first zero. */
2638  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);
2639  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_C_NOT_ZERO);
2640  /* 2 byte sequence */  /* Two byte sequence. */
2641  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 1);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2642  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));
2643  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x1f);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x1f);
2644  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2645  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2646  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2647  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 1);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(1));
2648  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2649  JUMPHERE(jump);  JUMPHERE(jump);
2650    
2651  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x10);  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x10);
2652  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_C_NOT_ZERO);
2653  /* 3 byte sequence */  /* Three byte sequence. */
2654  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 1);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2655  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x0f);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x0f);
2656  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 12);  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 12);
2657  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2658  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2659  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2660  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 2);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2661  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 2);  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2662  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2663  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2664  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 2);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(2));
2665  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2666  JUMPHERE(jump);  JUMPHERE(jump);
2667    
2668  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x08);  /* Four byte sequence. */
2669  jump = JUMP(SLJIT_C_NOT_ZERO);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
 /* 4 byte sequence */  
 OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 1);  
2670  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x07);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x07);
2671  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 18);  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 18);
2672  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2673  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 12);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 12);
2674  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2675  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 2);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
 OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  
 OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
 OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 3);  
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 3);  
 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, 3);  
 sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  
 JUMPHERE(jump);  
   
 /* 5 byte sequence */  
 OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 1);  
 OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x03);  
 OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 24);  
 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);  
 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), 3);  
2676  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2677  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2678  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2679  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 4);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(3));
2680  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));
2681  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2682  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2683  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 4);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(3));
2684  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2685  }  }
2686    
2687  static void do_utf8readtype8(compiler_common *common)  static void do_utfreadtype8(compiler_common *common)
2688  {  {
2689  /* Fast decoding an utf8 character type. TMP2 contains the first byte  /* Fast decoding a UTF-8 character type. TMP2 contains the first byte
2690  of the character (>= 0xc0) and TMP1 is destroyed. Return value in TMP1. */  of the character (>= 0xc0). Return value in TMP1. */
2691  DEFINE_COMPILER;  DEFINE_COMPILER;
2692  struct sljit_jump *jump;  struct sljit_jump *jump;
2693  struct sljit_jump *compare;  struct sljit_jump *compare;
2694    
2695  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2696    
2697  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);
2698  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_C_NOT_ZERO);
2699  /* 2 byte sequence */  /* Two byte sequence. */
2700  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2701  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));
2702  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);
2703  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2704  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
# Line 1565  sljit_emit_fast_return(compiler, RETURN_ Line 2713  sljit_emit_fast_return(compiler, RETURN_
2713  JUMPHERE(jump);  JUMPHERE(jump);
2714    
2715  /* We only have types for characters less than 256. */  /* We only have types for characters less than 256. */
2716  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), (sljit_w)_pcre_utf8_char_sizes - 0xc0);  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2717  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2718  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2719  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2720  }  }
2721    
2722  #endif  #elif defined COMPILE_PCRE16
2723    
2724    static void do_utfreadchar(compiler_common *common)
2725    {
2726    /* Fast decoding a UTF-16 character. TMP1 contains the first 16 bit char
2727    of the character (>= 0xd800). Return char value in TMP1, length - 1 in TMP2. */
2728    DEFINE_COMPILER;
2729    struct sljit_jump *jump;
2730    
2731    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2732    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xdc00);
2733    /* Do nothing, only return. */
2734    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2735    
2736    JUMPHERE(jump);
2737    /* Combine two 16 bit characters. */
2738    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2739    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2740    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2741    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 10);
2742    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3ff);
2743    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2744    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(1));
2745    OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2746    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2747    }
2748    
2749    #endif /* COMPILE_PCRE[8|16] */
2750    
2751    #endif /* SUPPORT_UTF */
2752    
2753  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2754    
# Line 1587  DEFINE_COMPILER; Line 2764  DEFINE_COMPILER;
2764    
2765  SLJIT_ASSERT(UCD_BLOCK_SIZE == 128 && sizeof(ucd_record) == 8);  SLJIT_ASSERT(UCD_BLOCK_SIZE == 128 && sizeof(ucd_record) == 8);
2766    
2767  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2768  OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);  OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
2769  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));
2770  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);
2771  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
2772  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, TMP2, 0);
2773  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, (sljit_w)_pcre_ucd_stage2);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_stage2));
2774  OP1(SLJIT_MOV_UH, TMP2, 0, SLJIT_MEM2(TMP2, TMP1), 1);  OP1(SLJIT_MOV_UH, TMP2, 0, SLJIT_MEM2(TMP2, TMP1), 1);
2775  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));
2776  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM2(TMP1, TMP2), 3);  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM2(TMP1, TMP2), 3);
2777  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2778  }  }
# Line 1609  struct sljit_label *newlinelabel = NULL; Line 2786  struct sljit_label *newlinelabel = NULL;
2786  struct sljit_jump *start;  struct sljit_jump *start;
2787  struct sljit_jump *end = NULL;  struct sljit_jump *end = NULL;
2788  struct sljit_jump *nl = NULL;  struct sljit_jump *nl = NULL;
2789  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2790  struct sljit_jump *singlebyte;  struct sljit_jump *singlechar;
2791  #endif  #endif
2792  jump_list *newline = NULL;  jump_list *newline = NULL;
2793  BOOL newlinecheck = FALSE;  BOOL newlinecheck = FALSE;
2794  BOOL readbyte = FALSE;  BOOL readuchar = FALSE;
2795    
2796    if (!(hascrorlf || firstline) && (common->nltype == NLTYPE_ANY ||
2797        common->nltype == NLTYPE_ANYCRLF || common->newline > 255))
2798      newlinecheck = TRUE;
2799    
2800    if (firstline)
2801      {
2802      /* Search for the end of the first line. */
2803      SLJIT_ASSERT(common->first_line_end != 0);
2804      OP1(SLJIT_MOV, TMP3, 0, STR_PTR, 0);
2805    
2806      if (common->nltype == NLTYPE_FIXED && common->newline > 255)
2807        {
2808        mainloop = LABEL();
2809        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2810        end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
2811        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(-1));
2812        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2813        CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, mainloop);
2814        CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, mainloop);
2815        JUMPHERE(end);
2816        OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2817        }
2818      else
2819        {
2820        end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
2821        mainloop = LABEL();
2822        /* Continual stores does not cause data dependency. */
2823        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);
2824        read_char(common);
2825        check_newlinechar(common, common->nltype, &newline, TRUE);
2826        CMPTO(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0, mainloop);
2827        JUMPHERE(end);
2828        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);
2829        set_jumps(newline, LABEL());
2830        }
2831    
2832      OP1(SLJIT_MOV, STR_PTR, 0, TMP3, 0);
2833      }
2834    
2835    start = JUMP(SLJIT_JUMP);
2836    
2837    if (newlinecheck)
2838      {
2839      newlinelabel = LABEL();
2840      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2841      end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
2842      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2843      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, common->newline & 0xff);
2844      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
2845    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2846      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, UCHAR_SHIFT);
2847    #endif
2848      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2849      nl = JUMP(SLJIT_JUMP);
2850      }
2851    
2852    mainloop = LABEL();
2853    
2854    /* Increasing the STR_PTR here requires one less jump in the most common case. */
2855    #ifdef SUPPORT_UTF
2856    if (common->utf) readuchar = TRUE;
2857    #endif
2858    if (newlinecheck) readuchar = TRUE;
2859    
2860    if (readuchar)
2861      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2862    
2863    if (newlinecheck)
2864      CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, newlinelabel);
2865    
2866    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2867    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2868    #if defined COMPILE_PCRE8
2869    if (common->utf)
2870      {
2871      singlechar = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2872      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2873      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2874      JUMPHERE(singlechar);
2875      }
2876    #elif defined COMPILE_PCRE16
2877    if (common->utf)
2878      {
2879      singlechar = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);
2880      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
2881      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2882      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
2883      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
2884      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2885      JUMPHERE(singlechar);
2886      }
2887    #endif /* COMPILE_PCRE[8|16] */
2888    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2889    JUMPHERE(start);
2890    
2891    if (newlinecheck)
2892      {
2893      JUMPHERE(end);
2894      JUMPHERE(nl);
2895      }
2896    
2897    return mainloop;
2898    }
2899    
2900    #define MAX_N_CHARS 3
2901    
2902    static SLJIT_INLINE BOOL fast_forward_first_n_chars(compiler_common *common, BOOL firstline)
2903    {
2904    DEFINE_COMPILER;
2905    struct sljit_label *start;
2906    struct sljit_jump *quit;
2907    pcre_uint32 chars[MAX_N_CHARS * 2];
2908    pcre_uchar *cc = common->start + 1 + LINK_SIZE;
2909    int location = 0;
2910    pcre_int32 len, c, bit, caseless;
2911    int must_stop;
2912    
2913    /* We do not support alternatives now. */
2914    if (*(common->start + GET(common->start, 1)) == OP_ALT)
2915      return FALSE;
2916    
2917    while (TRUE)
2918      {
2919      caseless = 0;
2920      must_stop = 1;
2921      switch(*cc)
2922        {
2923        case OP_CHAR:
2924        must_stop = 0;
2925        cc++;
2926        break;
2927    
2928        case OP_CHARI:
2929        caseless = 1;
2930        must_stop = 0;
2931        cc++;
2932        break;
2933    
2934        case OP_SOD:
2935        case OP_SOM:
2936        case OP_SET_SOM:
2937        case OP_NOT_WORD_BOUNDARY:
2938        case OP_WORD_BOUNDARY:
2939        case OP_EODN:
2940        case OP_EOD:
2941        case OP_CIRC:
2942        case OP_CIRCM:
2943        case OP_DOLL:
2944        case OP_DOLLM:
2945        /* Zero width assertions. */
2946        cc++;
2947        continue;
2948    
2949        case OP_PLUS:
2950        case OP_MINPLUS:
2951        case OP_POSPLUS:
2952        cc++;
2953        break;
2954    
2955        case OP_EXACT:
2956        cc += 1 + IMM2_SIZE;
2957        break;
2958    
2959        case OP_PLUSI:
2960        case OP_MINPLUSI:
2961        case OP_POSPLUSI:
2962        caseless = 1;
2963        cc++;
2964        break;
2965    
2966  if (!(hascrorlf || firstline) && (common->nltype == NLTYPE_ANY ||      case OP_EXACTI:
2967      common->nltype == NLTYPE_ANYCRLF || common->newline > 255))      caseless = 1;
2968    newlinecheck = TRUE;      cc += 1 + IMM2_SIZE;
2969        break;
2970    
2971  if (firstline)      default:
2972    {      must_stop = 2;
2973    /* Search for the end of the first line. */      break;
2974    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, STR_PTR, 0);      }
   OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), FIRSTLINE_END, STR_END, 0);  
2975    
2976    if (common->nltype == NLTYPE_FIXED && common->newline > 255)    if (must_stop == 2)
2977      {        break;
2978      mainloop = LABEL();  
2979      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);    len = 1;
2980      end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);  #ifdef SUPPORT_UTF
2981      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), -1);    if (common->utf && HAS_EXTRALEN(cc[0])) len += GET_EXTRALEN(cc[0]);
2982      OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);  #endif
2983      CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, mainloop);  
2984      CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, mainloop);    if (caseless && char_has_othercase(common, cc))
2985      OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_LOCALS_REG), FIRSTLINE_END, STR_PTR, 0, SLJIT_IMM, 1);      {
2986        caseless = char_get_othercase_bit(common, cc);
2987        if (caseless == 0)
2988          return FALSE;
2989    #ifdef COMPILE_PCRE8
2990        caseless = ((caseless & 0xff) << 8) | (len - (caseless >> 8));
2991    #else
2992        if ((caseless & 0x100) != 0)
2993          caseless = ((caseless & 0xff) << 16) | (len - (caseless >> 9));
2994        else
2995          caseless = ((caseless & 0xff) << 8) | (len - (caseless >> 9));
2996    #endif
2997      }      }
2998    else    else
2999        caseless = 0;
3000    
3001      while (len > 0 && location < MAX_N_CHARS * 2)
3002      {      {
3003      end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);      c = *cc;
3004      mainloop = LABEL();      bit = 0;
3005      /* Continual stores does not cause data dependency. */      if (len == (caseless & 0xff))
3006      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), FIRSTLINE_END, STR_PTR, 0);        {
3007      read_char(common);        bit = caseless >> 8;
3008      check_newlinechar(common, common->nltype, &newline, TRUE);        c |= bit;
3009      CMPTO(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0, mainloop);        }
3010      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), FIRSTLINE_END, STR_PTR, 0);  
3011      set_jumps(newline, LABEL());      chars[location] = c;
3012        chars[location + 1] = bit;
3013    
3014        len--;
3015        location += 2;
3016        cc++;
3017      }      }
3018    
3019    JUMPHERE(end);    if (location >= MAX_N_CHARS * 2 || must_stop != 0)
3020    OP1(SLJIT_MOV, STR_PTR, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0);      break;
3021    }    }
3022    
3023  start = JUMP(SLJIT_JUMP);  /* At least two characters are required. */
3024    if (location < 2 * 2)
3025        return FALSE;
3026    
3027  if (newlinecheck)  if (firstline)
3028    {    {
3029    newlinelabel = LABEL();    SLJIT_ASSERT(common->first_line_end != 0);
3030    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);    OP1(SLJIT_MOV, TMP3, 0, STR_END, 0);
3031    end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);    OP2(SLJIT_SUB, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, SLJIT_IMM, IN_UCHARS((location >> 1) - 1));
   OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  
   OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, common->newline & 0xff);  
   COND_VALUE(SLJIT_MOV, TMP1, 0, SLJIT_C_EQUAL);  
   OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);  
   nl = JUMP(SLJIT_JUMP);  
3032    }    }
3033    else
3034      OP2(SLJIT_SUB, STR_END, 0, STR_END, 0, SLJIT_IMM, IN_UCHARS((location >> 1) - 1));
3035    
3036  mainloop = LABEL();  start = LABEL();
3037    quit = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
 /* Increasing the STR_PTR here requires one less jump in the most common case. */  
 #ifdef SUPPORT_UTF8  
 if (common->utf8) readbyte = TRUE;  
 #endif  
 if (newlinecheck) readbyte = TRUE;  
   
 if (readbyte)  
   OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  
   
 if (newlinecheck)  
   CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, newlinelabel);  
3038    
3039  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
3040  #ifdef SUPPORT_UTF8  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
3041  if (common->utf8)  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3042    {  if (chars[1] != 0)
3043    singlebyte = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, chars[1]);
3044    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_w)_pcre_utf8_char_sizes - 0xc0);  CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, chars[0], start);
3045    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);  if (location > 2 * 2)
3046    JUMPHERE(singlebyte);    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
3047    if (chars[3] != 0)
3048      OP2(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_IMM, chars[3]);
3049    CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, chars[2], start);
3050    if (location > 2 * 2)
3051      {
3052      if (chars[5] != 0)
3053        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, chars[5]);
3054      CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, chars[4], start);
3055    }    }
3056  #endif  OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
 JUMPHERE(start);  
3057    
3058  if (newlinecheck)  JUMPHERE(quit);
   {  
   JUMPHERE(end);  
   JUMPHERE(nl);  
   }  
3059    
3060  return mainloop;  if (firstline)
3061      OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);
3062    else
3063      OP2(SLJIT_ADD, STR_END, 0, STR_END, 0, SLJIT_IMM, IN_UCHARS((location >> 1) - 1));
3064    return TRUE;
3065  }  }
3066    
3067  static SLJIT_INLINE void fast_forward_first_byte(compiler_common *common, pcre_uint16 firstbyte, BOOL firstline)  #undef MAX_N_CHARS
3068    
3069    static SLJIT_INLINE void fast_forward_first_char(compiler_common *common, pcre_uchar first_char, BOOL caseless, BOOL firstline)
3070  {  {
3071  DEFINE_COMPILER;  DEFINE_COMPILER;
3072  struct sljit_label *start;  struct sljit_label *start;
3073  struct sljit_jump *leave;  struct sljit_jump *quit;
3074  struct sljit_jump *found;  struct sljit_jump *found;
3075  pcre_uint16 oc, bit;  pcre_uchar oc, bit;
3076    
3077  if (firstline)  if (firstline)
3078    {    {
3079    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), POSSESSIVE0, STR_END, 0);    SLJIT_ASSERT(common->first_line_end != 0);
3080    OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), FIRSTLINE_END);    OP1(SLJIT_MOV, TMP3, 0, STR_END, 0);
3081      OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end);
3082    }    }
3083    
3084  start = LABEL();  start = LABEL();
3085  leave = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);  quit = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3086  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3087    
3088  if ((firstbyte & REQ_CASELESS) == 0)  oc = first_char;
3089    found = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, firstbyte & 0xff);  if (caseless)
3090      {
3091      oc = TABLE_GET(first_char, common->fcc, first_char);
3092    #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8)
3093      if (first_char > 127 && common->utf)
3094        oc = UCD_OTHERCASE(first_char);
3095    #endif
3096      }
3097    if (first_char == oc)
3098      found = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, first_char);
3099  else  else
3100    {    {
3101    firstbyte &= 0xff;    bit = first_char ^ oc;
3102    oc = common->fcc[firstbyte];    if (is_powerof2(bit))
   bit = firstbyte ^ oc;  
   if (ispowerof2(bit))  
3103      {      {
3104      OP2(SLJIT_OR, TMP2, 0, TMP1, 0, SLJIT_IMM, bit);      OP2(SLJIT_OR, TMP2, 0, TMP1, 0, SLJIT_IMM, bit);
3105      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, firstbyte | bit);      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, first_char | bit);
3106      }      }
3107    else    else
3108      {      {
3109      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);
3110      COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_EQUAL);      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
3111      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);
3112      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);
3113      found = JUMP(SLJIT_C_NOT_ZERO);      found = JUMP(SLJIT_C_NOT_ZERO);
3114      }      }
3115    }    }
3116    
3117  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));
 #ifdef SUPPORT_UTF8  
 if (common->utf8)  
   {  
   CMPTO(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0, start);  
   OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_w)_pcre_utf8_char_sizes - 0xc0);  
   OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);  
   }  
 #endif  
3118  JUMPTO(SLJIT_JUMP, start);  JUMPTO(SLJIT_JUMP, start);
3119  JUMPHERE(found);  JUMPHERE(found);
3120  JUMPHERE(leave);  JUMPHERE(quit);
3121    
3122  if (firstline)  if (firstline)
3123    OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), POSSESSIVE0);    OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);
3124  }  }
3125    
3126  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 1766  DEFINE_COMPILER; Line 3129  DEFINE_COMPILER;
3129  struct sljit_label *loop;  struct sljit_label *loop;
3130  struct sljit_jump *lastchar;  struct sljit_jump *lastchar;
3131  struct sljit_jump *firstchar;  struct sljit_jump *firstchar;
3132  struct sljit_jump *leave;  struct sljit_jump *quit;
3133  struct sljit_jump *foundcr = NULL;  struct sljit_jump *foundcr = NULL;
3134  struct sljit_jump *notfoundnl;  struct sljit_jump *notfoundnl;
3135  jump_list *newline = NULL;  jump_list *newline = NULL;
3136    
3137  if (firstline)  if (firstline)
3138    {    {
3139    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), POSSESSIVE0, STR_END, 0);    SLJIT_ASSERT(common->first_line_end != 0);
3140    OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), FIRSTLINE_END);    OP1(SLJIT_MOV, TMP3, 0, STR_END, 0);
3141      OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end);
3142    }    }
3143    
3144  if (common->nltype == NLTYPE_FIXED && common->newline > 255)  if (common->nltype == NLTYPE_FIXED && common->newline > 255)
# Line 1785  if (common->nltype == NLTYPE_FIXED && co Line 3149  if (common->nltype == NLTYPE_FIXED && co
3149    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));
3150    firstchar = CMP(SLJIT_C_LESS_EQUAL, STR_PTR, 0, TMP2, 0);    firstchar = CMP(SLJIT_C_LESS_EQUAL, STR_PTR, 0, TMP2, 0);
3151    
3152    OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 2);    OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, IN_UCHARS(2));
3153    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);
3154    COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_GREATER_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_GREATER_EQUAL);
3155    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
3156      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCHAR_SHIFT);
3157    #endif
3158    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
3159    
3160    loop = LABEL();    loop = LABEL();
3161    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));
3162    leave = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);    quit = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3163    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), -2);    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(-2));
3164    OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), -1);    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(-1));
3165    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);
3166    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);
3167    
3168    JUMPHERE(leave);    JUMPHERE(quit);
3169    JUMPHERE(firstchar);    JUMPHERE(firstchar);
3170    JUMPHERE(lastchar);    JUMPHERE(lastchar);
3171    
# Line 1822  set_jumps(newline, loop); Line 3189  set_jumps(newline, loop);
3189    
3190  if (common->nltype == NLTYPE_ANY || common->nltype == NLTYPE_ANYCRLF)  if (common->nltype == NLTYPE_ANY || common->nltype == NLTYPE_ANYCRLF)
3191    {    {
3192    leave = JUMP(SLJIT_JUMP);    quit = JUMP(SLJIT_JUMP);
3193    JUMPHERE(foundcr);    JUMPHERE(foundcr);
3194    notfoundnl = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);    notfoundnl = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3195    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3196    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);
3197    COND_VALUE(SLJIT_MOV, TMP1, 0, SLJIT_C_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
3198    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
3199      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, UCHAR_SHIFT);
3200    #endif
3201    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3202    JUMPHERE(notfoundnl);    JUMPHERE(notfoundnl);
3203    JUMPHERE(leave);    JUMPHERE(quit);
3204    }    }
3205  JUMPHERE(lastchar);  JUMPHERE(lastchar);
3206  JUMPHERE(firstchar);  JUMPHERE(firstchar);
3207    
3208  if (firstline)  if (firstline)
3209    OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), POSSESSIVE0);    OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);
3210  }  }
3211    
3212    static BOOL check_class_ranges(compiler_common *common, const pcre_uint8 *bits, BOOL nclass, jump_list **backtracks);
3213    
3214  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)
3215  {  {
3216  DEFINE_COMPILER;  DEFINE_COMPILER;
3217  struct sljit_label *start;  struct sljit_label *start;
3218  struct sljit_jump *leave;  struct sljit_jump *quit;
3219  struct sljit_jump *found;  struct sljit_jump *found = NULL;
3220    jump_list *matches = NULL;
3221    pcre_uint8 inverted_start_bits[32];
3222    int i;
3223    #ifndef COMPILE_PCRE8
3224    struct sljit_jump *jump;
3225    #endif
3226    
3227    for (i = 0; i < 32; ++i)
3228      inverted_start_bits[i] = ~(((pcre_uint8*)start_bits)[i]);
3229    
3230  if (firstline)  if (firstline)
3231    {    {
3232    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), POSSESSIVE0, STR_END, 0);    SLJIT_ASSERT(common->first_line_end != 0);
3233    OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), FIRSTLINE_END);    OP1(SLJIT_MOV, RETURN_ADDR, 0, STR_END, 0);
3234      OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end);
3235    }    }
3236    
3237  start = LABEL();  start = LABEL();
3238  leave = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);  quit = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3239  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3240  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3241  if (common->utf8)  if (common->utf)
3242    OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);    OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);
3243  #endif  #endif
 OP2(SLJIT_AND, TMP2, 0, TMP1, 0, SLJIT_IMM, 0x7);  
 OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 3);  
 OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), start_bits);  
 OP2(SLJIT_SHL, TMP2, 0, SLJIT_IMM, 1, TMP2, 0);  
 OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);  
 found = JUMP(SLJIT_C_NOT_ZERO);  
3244    
3245  #ifdef SUPPORT_UTF8  if (!check_class_ranges(common, inverted_start_bits, (inverted_start_bits[31] & 0x80) != 0, &matches))
3246  if (common->utf8)    {
3247    #ifndef COMPILE_PCRE8
3248      jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 255);
3249      OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 255);
3250      JUMPHERE(jump);
3251    #endif
3252      OP2(SLJIT_AND, TMP2, 0, TMP1, 0, SLJIT_IMM, 0x7);
3253      OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 3);
3254      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), start_bits);
3255      OP2(SLJIT_SHL, TMP2, 0, SLJIT_IMM, 1, TMP2, 0);
3256      OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);
3257      found = JUMP(SLJIT_C_NOT_ZERO);
3258      }
3259    
3260    #ifdef SUPPORT_UTF
3261    if (common->utf)
3262    OP1(SLJIT_MOV, TMP1, 0, TMP3, 0);    OP1(SLJIT_MOV, TMP1, 0, TMP3, 0);
3263  #endif  #endif
3264  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));
3265  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3266  if (common->utf8)  #if defined COMPILE_PCRE8
3267    if (common->utf)
3268    {    {
3269    CMPTO(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0, start);    CMPTO(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0, start);
3270    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_w)_pcre_utf8_char_sizes - 0xc0);    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
3271    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3272    }    }
3273  #endif  #elif defined COMPILE_PCRE16
3274    if (common->utf)
3275      {
3276      CMPTO(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800, start);
3277      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
3278      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);
3279      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
3280      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
3281      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3282      }
3283    #endif /* COMPILE_PCRE[8|16] */
3284    #endif /* SUPPORT_UTF */
3285  JUMPTO(SLJIT_JUMP, start);  JUMPTO(SLJIT_JUMP, start);
3286  JUMPHERE(found);  if (found != NULL)
3287  JUMPHERE(leave);    JUMPHERE(found);
3288    if (matches != NULL)
3289      set_jumps(matches, LABEL());
3290    JUMPHERE(quit);
3291    
3292  if (firstline)  if (firstline)
3293    OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), POSSESSIVE0);    OP1(SLJIT_MOV, STR_END, 0, RETURN_ADDR, 0);
3294  }  }
3295    
3296  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)
3297  {  {
3298  DEFINE_COMPILER;  DEFINE_COMPILER;
3299  struct sljit_label *loop;  struct sljit_label *loop;
# Line 1896  struct sljit_jump *alreadyfound; Line 3302  struct sljit_jump *alreadyfound;
3302  struct sljit_jump *found;  struct sljit_jump *found;
3303  struct sljit_jump *foundoc = NULL;  struct sljit_jump *foundoc = NULL;
3304  struct sljit_jump *notfound;  struct sljit_jump *notfound;
3305  pcre_uint16 oc, bit;  pcre_uint32 oc, bit;
3306    
3307  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), REQ_BYTE_PTR);  SLJIT_ASSERT(common->req_char_ptr != 0);
3308    OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->req_char_ptr);
3309  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);
3310  toolong = CMP(SLJIT_C_LESS, TMP1, 0, STR_END, 0);  toolong = CMP(SLJIT_C_LESS, TMP1, 0, STR_END, 0);
3311  alreadyfound = CMP(SLJIT_C_LESS, STR_PTR, 0, TMP2, 0);  alreadyfound = CMP(SLJIT_C_LESS, STR_PTR, 0, TMP2, 0);
3312    
3313  if (has_firstbyte)  if (has_firstchar)
3314    OP2(SLJIT_ADD, TMP1, 0, STR_PTR, 0, SLJIT_IMM, 1);    OP2(SLJIT_ADD, TMP1, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3315  else  else
3316    OP1(SLJIT_MOV, TMP1, 0, STR_PTR, 0);    OP1(SLJIT_MOV, TMP1, 0, STR_PTR, 0);
3317    
3318  loop = LABEL();  loop = LABEL();
3319  notfound = CMP(SLJIT_C_GREATER_EQUAL, TMP1, 0, STR_END, 0);  notfound = CMP(SLJIT_C_GREATER_EQUAL, TMP1, 0, STR_END, 0);
3320    
3321  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP1), 0);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(TMP1), 0);
3322  if ((reqbyte & REQ_CASELESS) == 0)  oc = req_char;
3323    found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, reqbyte & 0xff);  if (caseless)
3324      {
3325      oc = TABLE_GET(req_char, common->fcc, req_char);
3326    #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8)
3327      if (req_char > 127 && common->utf)
3328        oc = UCD_OTHERCASE(req_char);
3329    #endif
3330      }
3331    if (req_char == oc)
3332      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, req_char);
3333  else  else
3334    {    {
3335    reqbyte &= 0xff;    bit = req_char ^ oc;
3336    oc = common->fcc[reqbyte];    if (is_powerof2(bit))
   bit = reqbyte ^ oc;  
   if (ispowerof2(bit))  
3337      {      {
3338      OP2(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_IMM, bit);      OP2(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_IMM, bit);
3339      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, reqbyte | bit);      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, req_char | bit);
3340      }      }
3341    else    else
3342      {      {
3343      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, reqbyte);      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, req_char);
3344      foundoc = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, oc);      foundoc = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, oc);
3345      }      }
3346    }    }
3347  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, IN_UCHARS(1));
3348  JUMPTO(SLJIT_JUMP, loop);  JUMPTO(SLJIT_JUMP, loop);
3349    
3350  JUMPHERE(found);  JUMPHERE(found);
3351  if (foundoc)  if (foundoc)
3352    JUMPHERE(foundoc);    JUMPHERE(foundoc);
3353  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);
3354  JUMPHERE(alreadyfound);  JUMPHERE(alreadyfound);
3355  JUMPHERE(toolong);  JUMPHERE(toolong);
3356  return notfound;  return notfound;
# Line 1948  DEFINE_COMPILER; Line 3362  DEFINE_COMPILER;
3362  struct sljit_jump *jump;  struct sljit_jump *jump;
3363  struct sljit_label *mainloop;  struct sljit_label *mainloop;
3364    
3365  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
3366  OP1(SLJIT_MOV, TMP1, 0, STACK_TOP, 0);  OP1(SLJIT_MOV, TMP1, 0, STACK_TOP, 0);
3367    GET_LOCAL_BASE(TMP3, 0, 0);
3368    
3369  /* Drop frames until we reach STACK_TOP. */  /* Drop frames until we reach STACK_TOP. */
3370  mainloop = LABEL();  mainloop = LABEL();
3371  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(TMP1), 0);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(TMP1), 0);
3372  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);
3373  OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_LOCALS_REG, 0);  jump = JUMP(SLJIT_C_SIG_LESS_EQUAL);
3374  OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), 0, SLJIT_MEM1(TMP1), sizeof(sljit_w));  
3375  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);
3376  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));
3377    OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), sizeof(sljit_sw), SLJIT_MEM1(TMP1), 2 * sizeof(sljit_sw));
3378    OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 3 * sizeof(sljit_sw));
3379  JUMPTO(SLJIT_JUMP, mainloop);  JUMPTO(SLJIT_JUMP, mainloop);
3380    
3381  JUMPHERE(jump);  JUMPHERE(jump);
3382  jump = CMP(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, frame_end);  jump = JUMP(SLJIT_C_SIG_LESS);
3383  /* End of dropping frames. */  /* End of dropping frames. */
3384  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3385    
3386  JUMPHERE(jump);  JUMPHERE(jump);
3387  jump = CMP(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, frame_setstrbegin);  OP1(SLJIT_NEG, TMP2, 0, TMP2, 0);
3388  /* Set string begin. */  OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, TMP3, 0);
3389  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(TMP1), sizeof(sljit_w));  OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), 0, SLJIT_MEM1(TMP1), sizeof(sljit_sw));
3390  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));
 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));  
3391  JUMPTO(SLJIT_JUMP, mainloop);  JUMPTO(SLJIT_JUMP, mainloop);
3392  }  }
3393    
3394  static void check_wordboundary(compiler_common *common)  static void check_wordboundary(compiler_common *common)
3395  {  {
3396  DEFINE_COMPILER;  DEFINE_COMPILER;
3397  struct sljit_jump *beginend;  struct sljit_jump *skipread;
3398  #ifdef SUPPORT_UTF8  jump_list *skipread_list = NULL;
3399    #if !(defined COMPILE_PCRE8) || defined SUPPORT_UTF
3400  struct sljit_jump *jump;  struct sljit_jump *jump;
3401  #endif  #endif
3402    
3403  SLJIT_COMPILE_ASSERT(ctype_word == 0x10, ctype_word_must_be_16);  SLJIT_COMPILE_ASSERT(ctype_word == 0x10, ctype_word_must_be_16);
3404    
3405  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);
3406  /* Get type of the previous char, and put it to LOCALS1. */  /* Get type of the previous char, and put it to LOCALS1. */
3407  OP1(SLJIT_MOV, TMP1, 0, ARGUMENTS, 0);  OP1(SLJIT_MOV, TMP1, 0, ARGUMENTS, 0);
3408  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));
3409  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, SLJIT_IMM, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, SLJIT_IMM, 0);
3410  beginend = CMP(SLJIT_C_LESS_EQUAL, STR_PTR, 0, TMP1, 0);  skipread = CMP(SLJIT_C_LESS_EQUAL, STR_PTR, 0, TMP1, 0);
3411  skip_char_back(common);  skip_char_back(common);
3412    check_start_used_ptr(common);
3413  read_char(common);  read_char(common);
3414    
3415  /* Testing char type. */  /* Testing char type. */
3416  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3417  if (common->useucp)  if (common->use_ucp)
3418    {    {
3419    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 1);    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 1);
3420    jump = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_UNDERSCORE);    jump = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_UNDERSCORE);
3421    add_jump(compiler, &common->getucd, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->getucd, JUMP(SLJIT_FAST_CALL));
3422    OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Ll);    OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Ll);
3423    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);
3424    COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_LESS_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_LESS_EQUAL);
3425    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);
3426    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);
3427    COND_VALUE(SLJIT_OR, TMP2, 0, SLJIT_C_LESS_EQUAL);    OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_C_LESS_EQUAL);
3428    JUMPHERE(jump);    JUMPHERE(jump);
3429    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP2, 0);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP2, 0);
3430    }    }
3431  else  else
3432  #endif  #endif
3433    {    {
3434  #ifdef SUPPORT_UTF8  #ifndef COMPILE_PCRE8
3435      jump = CMP(SLJIT_C_GREATER, TMP1, 0, SLJIT_IMM, 255);
3436    #elif defined SUPPORT_UTF
3437    /* Here LOCALS1 has already been zeroed. */    /* Here LOCALS1 has already been zeroed. */
3438    jump = NULL;    jump = NULL;
3439    if (common->utf8)    if (common->utf)
3440      jump = CMP(SLJIT_C_GREATER, TMP1, 0, SLJIT_IMM, 255);      jump = CMP(SLJIT_C_GREATER, TMP1, 0, SLJIT_IMM, 255);
3441  #endif  #endif /* COMPILE_PCRE8 */
3442    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), common->ctypes);    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), common->ctypes);
3443    OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 4 /* ctype_word */);    OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 4 /* ctype_word */);
3444    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
3445    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);
3446  #ifdef SUPPORT_UTF8  #ifndef COMPILE_PCRE8
3447      JUMPHERE(jump);
3448    #elif defined SUPPORT_UTF
3449    if (jump != NULL)    if (jump != NULL)
3450      JUMPHERE(jump);      JUMPHERE(jump);
3451  #endif  #endif /* COMPILE_PCRE8 */
3452    }    }
3453  JUMPHERE(beginend);  JUMPHERE(skipread);
3454    
3455  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 0);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 0);
3456  beginend = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);  check_str_end(common, &skipread_list);
3457  peek_char(common);  peek_char(common);
3458    
3459  /* Testing char type. This is a code duplication. */  /* Testing char type. This is a code duplication. */
3460  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3461  if (common->useucp)  if (common->use_ucp)
3462    {    {
3463    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 1);    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 1);
3464    jump = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_UNDERSCORE);    jump = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_UNDERSCORE);
3465    add_jump(compiler, &common->getucd, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->getucd, JUMP(SLJIT_FAST_CALL));
3466    OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Ll);    OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Ll);
3467    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);
3468    COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_LESS_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_LESS_EQUAL);
3469    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);
3470    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);
3471    COND_VALUE(SLJIT_OR, TMP2, 0, SLJIT_C_LESS_EQUAL);    OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_C_LESS_EQUAL);
3472    JUMPHERE(jump);    JUMPHERE(jump);
3473    }    }
3474  else  else