/[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 741 by zherczeg, Mon Oct 31 09:31:46 2011 UTC revision 1599 by zherczeg, Wed Sep 2 08:50:31 2015 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2008 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, allocator_data) (PUBL(malloc))(size)
56    #define SLJIT_FREE(ptr, allocator_data) (PUBL(free))(ptr)
57  #define SLJIT_CONFIG_AUTO 1  #define SLJIT_CONFIG_AUTO 1
58  #define SLJIT_CONFIG_STATIC 1  #define SLJIT_CONFIG_STATIC 1
59  #define SLJIT_VERBOSE 0  #define SLJIT_VERBOSE 0
# Line 60  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 80  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 106  we have the following regular expression Line 120  we have the following regular expression
120    
121  The generated code will be the following:  The generated code will be the following:
122    
123   A hot path   A matching path
124   '(' hot path (pushing arguments to the stack)   '(' matching path (pushing arguments to the stack)
125   B hot path   B matching path
126   ')' hot path (pushing arguments to the stack)   ')' matching path (pushing arguments to the stack)
127   D hot path   D matching path
128   return with successful match   return with successful match
129    
130   D fallback path   D backtrack path
131   ')' fallback path (If we arrived from "C" jump to the fallback of "C")   ')' backtrack path (If we arrived from "C" jump to the backtrack of "C")
132   B fallback path   B backtrack path
133   C expected path   C expected path
134   jump to D hot path   jump to D matching path
135   C fallback path   C backtrack path
136   A fallback path   A backtrack path
137    
138   Notice, that the order of fallback code paths are the opposite of the fast   Notice, that the order of backtrack code paths are the opposite of the fast
139   code paths. In this way the topmost value on the stack is always belong   code paths. In this way the topmost value on the stack is always belong
140   to the current fallback code path. The fallback code path must check   to the current backtrack code path. The backtrack path must check
141   whether there is a next alternative. If so, it needs to jump back to   whether there is a next alternative. If so, it needs to jump back to
142   the hot path eventually. Otherwise it needs to clear out its own stack   the matching path eventually. Otherwise it needs to clear out its own stack
143   frame and continue the execution on the fallback code paths.   frame and continue the execution on the backtrack code paths.
144  */  */
145    
146  /*  /*
147  Saved stack frames:  Saved stack frames:
148    
149  Atomic blocks and asserts require reloading the values of local variables  Atomic blocks and asserts require reloading the values of private data
150  when the fallback mechanism performed. Because of OP_RECURSE, the locals  when the backtrack mechanism performed. Because of OP_RECURSE, the data
151  are not necessarly known in compile time, thus we need a dynamic restore  are not necessarly known in compile time, thus we need a dynamic restore
152  mechanism.  mechanism.
153    
154  The stack frames are stored in a chain list, and have the following format:  The stack frames are stored in a chain list, and have the following format:
155  ([ capturing bracket offset ][ start value ][ end value ])+ ... [ 0 ] [ previous head ]  ([ capturing bracket offset ][ start value ][ end value ])+ ... [ 0 ] [ previous head ]
156    
157  Thus we can restore the locals to a particular point in the stack.  Thus we can restore the private data to a particular point in the stack.
158  */  */
159    
160  typedef struct jit_arguments {  typedef struct jit_arguments {
161    /* Pointers first. */    /* Pointers first. */
162    struct sljit_stack *stack;    struct sljit_stack *stack;
163    PCRE_SPTR str;    const pcre_uchar *str;
164    PCRE_SPTR begin;    const pcre_uchar *begin;
165    PCRE_SPTR end;    const pcre_uchar *end;
166    int *offsets;    int *offsets;
167    uschar *ptr;    pcre_uchar *uchar_ptr;
168      pcre_uchar *mark_ptr;
169      void *callout_data;
170    /* Everything else after. */    /* Everything else after. */
171    int offsetcount;    pcre_uint32 limit_match;
172    int calllimit;    int real_offset_count;
173    uschar notbol;    int offset_count;
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;    void *read_only_data_heads[JIT_NUMBER_OF_COMPILE_MODES];
183      sljit_uw executable_sizes[JIT_NUMBER_OF_COMPILE_MODES];
184      PUBL(jit_callback) callback;
185    void *userdata;    void *userdata;
186  } executable_function;    pcre_uint32 top_bracket;
187      pcre_uint32 limit_match;
188    } executable_functions;
189    
190  typedef struct jump_list {  typedef struct jump_list {
191    struct sljit_jump *jump;    struct sljit_jump *jump;
192    struct jump_list *next;    struct jump_list *next;
193  } jump_list;  } jump_list;
194    
 enum stub_types { stack_alloc };  
   
195  typedef struct stub_list {  typedef struct stub_list {
   enum stub_types type;  
   int data;  
196    struct sljit_jump *start;    struct sljit_jump *start;
197    struct sljit_label *leave;    struct sljit_label *quit;
198    struct stub_list *next;    struct stub_list *next;
199  } stub_list;  } stub_list;
200    
201    typedef struct label_addr_list {
202      struct sljit_label *label;
203      sljit_uw *update_addr;
204      struct label_addr_list *next;
205    } label_addr_list;
206    
207    enum frame_types {
208      no_frame = -1,
209      no_stack = -2
210    };
211    
212    enum control_types {
213      type_mark = 0,
214      type_then_trap = 1
215    };
216    
217  typedef int (SLJIT_CALL *jit_function)(jit_arguments *args);  typedef int (SLJIT_CALL *jit_function)(jit_arguments *args);
218    
219  /* The following structure is the key data type for the recursive  /* The following structure is the key data type for the recursive
220  code generator. It is allocated by compile_hotpath, and contains  code generator. It is allocated by compile_matchingpath, and contains
221  the aguments for compile_fallbackpath. Must be the first member  the arguments for compile_backtrackingpath. Must be the first member
222  of its descendants. */  of its descendants. */
223  typedef struct fallback_common {  typedef struct backtrack_common {
224    /* Concatenation stack. */    /* Concatenation stack. */
225    struct fallback_common *prev;    struct backtrack_common *prev;
226    jump_list *nextfallbacks;    jump_list *nextbacktracks;
227    /* Internal stack (for component operators). */    /* Internal stack (for component operators). */
228    struct fallback_common *top;    struct backtrack_common *top;
229    jump_list *topfallbacks;    jump_list *topbacktracks;
230    /* Opcode pointer. */    /* Opcode pointer. */
231    uschar *cc;    pcre_uchar *cc;
232  } fallback_common;  } backtrack_common;
233    
234  typedef struct assert_fallback {  typedef struct assert_backtrack {
235    fallback_common common;    backtrack_common common;
236    jump_list *condfailed;    jump_list *condfailed;
237    /* Less than 0 (-1) if a frame is not needed. */    /* Less than 0 if a frame is not needed. */
238    int framesize;    int framesize;
239    /* Points to our private memory word on the stack. */    /* Points to our private memory word on the stack. */
240    int localptr;    int private_data_ptr;
241    /* For iterators. */    /* For iterators. */
242    struct sljit_label *hotpath;    struct sljit_label *matchingpath;
243  } assert_fallback;  } assert_backtrack;
244    
245  typedef struct bracket_fallback {  typedef struct bracket_backtrack {
246    fallback_common common;    backtrack_common common;
247    /* Where to coninue if an alternative is successfully matched. */    /* Where to coninue if an alternative is successfully matched. */
248    struct sljit_label *althotpath;    struct sljit_label *alternative_matchingpath;
249    /* For rmin and rmax iterators. */    /* For rmin and rmax iterators. */
250    struct sljit_label *recursivehotpath;    struct sljit_label *recursive_matchingpath;
251    /* For greedy ? operator. */    /* For greedy ? operator. */
252    struct sljit_label *zerohotpath;    struct sljit_label *zero_matchingpath;
253    /* Contains the branches of a failed condition. */    /* Contains the branches of a failed condition. */
254    union {    union {
255      /* Both for OP_COND, OP_SCOND. */      /* Both for OP_COND, OP_SCOND. */
256      jump_list *condfailed;      jump_list *condfailed;
257      assert_fallback *assert;      assert_backtrack *assert;
258      /* For OP_ONCE. -1 if not needed. */      /* For OP_ONCE. Less than 0 if not needed. */
259      int framesize;      int framesize;
260    } u;    } u;
261    /* Points to our private memory word on the stack. */    /* Points to our private memory word on the stack. */
262    int localptr;    int private_data_ptr;
263  } bracket_fallback;  } bracket_backtrack;
264    
265  typedef struct bracketpos_fallback {  typedef struct bracketpos_backtrack {
266    fallback_common common;    backtrack_common common;
267    /* Points to our private memory word on the stack. */    /* Points to our private memory word on the stack. */
268    int localptr;    int private_data_ptr;
269    /* Reverting stack is needed. */    /* Reverting stack is needed. */
270    int framesize;    int framesize;
271    /* Allocated stack size. */    /* Allocated stack size. */
272    int stacksize;    int stacksize;
273  } bracketpos_fallback;  } bracketpos_backtrack;
274    
275  typedef struct braminzero_fallback {  typedef struct braminzero_backtrack {
276    fallback_common common;    backtrack_common common;
277    struct sljit_label *hotpath;    struct sljit_label *matchingpath;
278  } braminzero_fallback;  } braminzero_backtrack;
279    
280  typedef struct iterator_fallback {  typedef struct iterator_backtrack {
281    fallback_common common;    backtrack_common common;
282    /* Next iteration. */    /* Next iteration. */
283    struct sljit_label *hotpath;    struct sljit_label *matchingpath;
284  } iterator_fallback;  } iterator_backtrack;
285    
286  typedef struct recurse_entry {  typedef struct recurse_entry {
287    struct recurse_entry *next;    struct recurse_entry *next;
# Line 257  typedef struct recurse_entry { Line 290  typedef struct recurse_entry {
290    /* Collects the calls until the function is not created. */    /* Collects the calls until the function is not created. */
291    jump_list *calls;    jump_list *calls;
292    /* Points to the starting opcode. */    /* Points to the starting opcode. */
293    int start;    sljit_sw start;
294  } recurse_entry;  } recurse_entry;
295    
296  typedef struct recurse_fallback {  typedef struct recurse_backtrack {
297    fallback_common common;    backtrack_common common;
298  } recurse_fallback;    BOOL inlined_pattern;
299    } recurse_backtrack;
300    
301    #define OP_THEN_TRAP OP_TABLE_LENGTH
302    
303    typedef struct then_trap_backtrack {
304      backtrack_common common;
305      /* If then_trap is not NULL, this structure contains the real
306      then_trap for the backtracking path. */
307      struct then_trap_backtrack *then_trap;
308      /* Points to the starting opcode. */
309      sljit_sw start;
310      /* Exit point for the then opcodes of this alternative. */
311      jump_list *quit;
312      /* Frame size of the current alternative. */
313      int framesize;
314    } then_trap_backtrack;
315    
316    #define MAX_RANGE_SIZE 4
317    
318  typedef struct compiler_common {  typedef struct compiler_common {
319      /* The sljit ceneric compiler. */
320    struct sljit_compiler *compiler;    struct sljit_compiler *compiler;
321    uschar *start;    /* First byte code. */
322    int localsize;    pcre_uchar *start;
323    int *localptrs;    /* Maps private data offset to each opcode. */
324    const uschar *fcc;    sljit_si *private_data_ptrs;
325    sljit_w lcc;    /* Chain list of read-only data ptrs. */
326    int cbraptr;    void *read_only_data_head;
327      /* Tells whether the capturing bracket is optimized. */
328      pcre_uint8 *optimized_cbracket;
329      /* Tells whether the starting offset is a target of then. */
330      pcre_uint8 *then_offsets;
331      /* Current position where a THEN must jump. */
332      then_trap_backtrack *then_trap;
333      /* Starting offset of private data for capturing brackets. */
334      int cbra_ptr;
335      /* Output vector starting point. Must be divisible by 2. */
336      int ovector_start;
337      /* Last known position of the requested byte. */
338      int req_char_ptr;
339      /* Head of the last recursion. */
340      int recursive_head_ptr;
341      /* First inspected character for partial matching. */
342      int start_used_ptr;
343      /* Starting pointer for partial soft matches. */
344      int hit_start;
345      /* End pointer of the first line. */
346      int first_line_end;
347      /* Points to the marked string. */
348      int mark_ptr;
349      /* Recursive control verb management chain. */
350      int control_head_ptr;
351      /* Points to the last matched capture block index. */
352      int capture_last_ptr;
353      /* Points to the starting position of the current match. */
354      int start_ptr;
355    
356      /* Flipped and lower case tables. */
357      const pcre_uint8 *fcc;
358      sljit_sw lcc;
359      /* Mode can be PCRE_STUDY_JIT_COMPILE and others. */
360      int mode;
361      /* TRUE, when minlength is greater than 0. */
362      BOOL might_be_empty;
363      /* \K is found in the pattern. */
364      BOOL has_set_som;
365      /* (*SKIP:arg) is found in the pattern. */
366      BOOL has_skip_arg;
367      /* (*THEN) is found in the pattern. */
368      BOOL has_then;
369      /* Needs to know the start position anytime. */
370      BOOL needs_start_ptr;
371      /* Currently in recurse or negative assert. */
372      BOOL local_exit;
373      /* Currently in a positive assert. */
374      BOOL positive_assert;
375      /* Newline control. */
376    int nltype;    int nltype;
377      pcre_uint32 nlmax;
378      pcre_uint32 nlmin;
379    int newline;    int newline;
380    int bsr_nltype;    int bsr_nltype;
381      pcre_uint32 bsr_nlmax;
382      pcre_uint32 bsr_nlmin;
383      /* Dollar endonly. */
384    int endonly;    int endonly;
385    sljit_w ctypes;    /* Tables. */
386    sljit_uw name_table;    sljit_sw ctypes;
387    sljit_w name_count;    /* Named capturing brackets. */
388    sljit_w name_entry_size;    pcre_uchar *name_table;
389    struct sljit_label *acceptlabel;    sljit_sw name_count;
390      sljit_sw name_entry_size;
391    
392      /* Labels and jump lists. */
393      struct sljit_label *partialmatchlabel;
394      struct sljit_label *quit_label;
395      struct sljit_label *forced_quit_label;
396      struct sljit_label *accept_label;
397      struct sljit_label *ff_newline_shortcut;
398    stub_list *stubs;    stub_list *stubs;
399      label_addr_list *label_addrs;
400    recurse_entry *entries;    recurse_entry *entries;
401    recurse_entry *currententry;    recurse_entry *currententry;
402      jump_list *partialmatch;
403      jump_list *quit;
404      jump_list *positive_assert_quit;
405      jump_list *forced_quit;
406    jump_list *accept;    jump_list *accept;
407    jump_list *calllimit;    jump_list *calllimit;
408    jump_list *stackalloc;    jump_list *stackalloc;
# Line 294  typedef struct compiler_common { Line 413  typedef struct compiler_common {
413    jump_list *vspace;    jump_list *vspace;
414    jump_list *casefulcmp;    jump_list *casefulcmp;
415    jump_list *caselesscmp;    jump_list *caselesscmp;
416      jump_list *reset_match;
417    BOOL jscript_compat;    BOOL jscript_compat;
418  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
419    BOOL utf8;    BOOL utf;
420  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
421    BOOL useucp;    BOOL use_ucp;
422  #endif  #endif
423    jump_list *utf8readchar;  #ifdef COMPILE_PCRE8
424    jump_list *utf8readtype8;    jump_list *utfreadchar;
425      jump_list *utfreadchar16;
426      jump_list *utfreadtype8;
427  #endif  #endif
428    #endif /* SUPPORT_UTF */
429  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
430    jump_list *getucd;    jump_list *getucd;
431  #endif  #endif
# Line 314  typedef struct compare_context { Line 437  typedef struct compare_context {
437    int length;    int length;
438    int sourcereg;    int sourcereg;
439  #if defined SLJIT_UNALIGNED && SLJIT_UNALIGNED  #if defined SLJIT_UNALIGNED && SLJIT_UNALIGNED
440    int byteptr;    int ucharptr;
441    union {    union {
442      int asint;      sljit_si asint;
443      short asshort;      sljit_uh asushort;
444    #if defined COMPILE_PCRE8
445      sljit_ub asbyte;      sljit_ub asbyte;
446      sljit_ub asbytes[4];      sljit_ub asuchars[4];
447    #elif defined COMPILE_PCRE16
448        sljit_uh asuchars[2];
449    #elif defined COMPILE_PCRE32
450        sljit_ui asuchars[1];
451    #endif
452    } c;    } c;
453    union {    union {
454      int asint;      sljit_si asint;
455      short asshort;      sljit_uh asushort;
456    #if defined COMPILE_PCRE8
457      sljit_ub asbyte;      sljit_ub asbyte;
458      sljit_ub asbytes[4];      sljit_ub asuchars[4];
459    #elif defined COMPILE_PCRE16
460        sljit_uh asuchars[2];
461    #elif defined COMPILE_PCRE32
462        sljit_ui asuchars[1];
463    #endif
464    } oc;    } oc;
465  #endif  #endif
466  } compare_context;  } compare_context;
467    
468  enum {  /* Undefine sljit macros. */
469    frame_end = 0,  #undef CMP
   frame_setstrbegin = -1  
 };  
470    
471  /* Used for accessing the elements of the stack. */  /* Used for accessing the elements of the stack. */
472  #define STACK(i)      ((-(i) - 1) * (int)sizeof(sljit_w))  #define STACK(i)      ((-(i) - 1) * (int)sizeof(sljit_sw))
473    
474  #define TMP1          SLJIT_TEMPORARY_REG1  #define TMP1          SLJIT_R0
475  #define TMP2          SLJIT_TEMPORARY_REG3  #define TMP2          SLJIT_R2
476  #define TMP3          SLJIT_TEMPORARY_EREG2  #define TMP3          SLJIT_R3
477  #define STR_PTR       SLJIT_GENERAL_REG1  #define STR_PTR       SLJIT_S0
478  #define STR_END       SLJIT_GENERAL_REG2  #define STR_END       SLJIT_S1
479  #define STACK_TOP     SLJIT_TEMPORARY_REG2  #define STACK_TOP     SLJIT_R1
480  #define STACK_LIMIT   SLJIT_GENERAL_REG3  #define STACK_LIMIT   SLJIT_S2
481  #define ARGUMENTS     SLJIT_GENERAL_EREG1  #define COUNT_MATCH   SLJIT_S3
482  #define CALL_COUNT    SLJIT_GENERAL_EREG2  #define ARGUMENTS     SLJIT_S4
483  #define RETURN_ADDR   SLJIT_TEMPORARY_EREG1  #define RETURN_ADDR   SLJIT_R4
484    
485  /* Locals layout. */  /* Local space layout. */
486  /* These two locals can be used by the current opcode. */  /* These two locals can be used by the current opcode. */
487  #define LOCALS0          (0 * sizeof(sljit_w))  #define LOCALS0          (0 * sizeof(sljit_sw))
488  #define LOCALS1          (1 * sizeof(sljit_w))  #define LOCALS1          (1 * sizeof(sljit_sw))
489  /* Two local variables for possessive quantifiers (char1 cannot use them). */  /* Two local variables for possessive quantifiers (char1 cannot use them). */
490  #define POSSESSIVE0      (2 * sizeof(sljit_w))  #define POSSESSIVE0      (2 * sizeof(sljit_sw))
491  #define POSSESSIVE1      (3 * sizeof(sljit_w))  #define POSSESSIVE1      (3 * sizeof(sljit_sw))
 /* Head of the last recursion. */  
 #define RECURSIVE_HEAD   (4 * sizeof(sljit_w))  
492  /* Max limit of recursions. */  /* Max limit of recursions. */
493  #define CALL_LIMIT       (5 * sizeof(sljit_w))  #define LIMIT_MATCH      (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))  
494  /* The output vector is stored on the stack, and contains pointers  /* The output vector is stored on the stack, and contains pointers
495  to characters. The vector data is divided into two groups: the first  to characters. The vector data is divided into two groups: the first
496  group contains the start / end character pointers, and the second is  group contains the start / end character pointers, and the second is
497  the start pointers when the end of the capturing group has not yet reached. */  the start pointers when the end of the capturing group has not yet reached. */
498  #define OVECTOR_START    (8 * sizeof(sljit_w))  #define OVECTOR_START    (common->ovector_start)
499  #define OVECTOR(i)       (OVECTOR_START + (i) * sizeof(sljit_w))  #define OVECTOR(i)       (OVECTOR_START + (i) * (sljit_sw)sizeof(sljit_sw))
500  #define OVECTOR_PRIV(i)  (common->cbraptr + (i) * sizeof(sljit_w))  #define OVECTOR_PRIV(i)  (common->cbra_ptr + (i) * (sljit_sw)sizeof(sljit_sw))
501  #define PRIV(cc)         (common->localptrs[(cc) - common->start])  #define PRIVATE_DATA(cc) (common->private_data_ptrs[(cc) - common->start])
502    
503    #if defined COMPILE_PCRE8
504    #define MOV_UCHAR  SLJIT_MOV_UB
505    #define MOVU_UCHAR SLJIT_MOVU_UB
506    #elif defined COMPILE_PCRE16
507    #define MOV_UCHAR  SLJIT_MOV_UH
508    #define MOVU_UCHAR SLJIT_MOVU_UH
509    #elif defined COMPILE_PCRE32
510    #define MOV_UCHAR  SLJIT_MOV_UI
511    #define MOVU_UCHAR SLJIT_MOVU_UI
512    #else
513    #error Unsupported compiling mode
514    #endif
515    
516  /* Shortcuts. */  /* Shortcuts. */
517  #define DEFINE_COMPILER \  #define DEFINE_COMPILER \
# Line 388  the start pointers when the end of the c Line 528  the start pointers when the end of the c
528    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))
529  #define JUMPHERE(jump) \  #define JUMPHERE(jump) \
530    sljit_set_label((jump), sljit_emit_label(compiler))    sljit_set_label((jump), sljit_emit_label(compiler))
531    #define SET_LABEL(jump, label) \
532      sljit_set_label((jump), (label))
533  #define CMP(type, src1, src1w, src2, src2w) \  #define CMP(type, src1, src1w, src2, src2w) \
534    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))
535  #define CMPTO(type, src1, src1w, src2, src2w, label) \  #define CMPTO(type, src1, src1w, src2, src2w, label) \
536    sljit_set_label(sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w)), (label))    sljit_set_label(sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w)), (label))
537  #define COND_VALUE(op, dst, dstw, type) \  #define OP_FLAGS(op, dst, dstw, src, srcw, type) \
538    sljit_emit_cond_value(compiler, (op), (dst), (dstw), (type))    sljit_emit_op_flags(compiler, (op), (dst), (dstw), (src), (srcw), (type))
539    #define GET_LOCAL_BASE(dst, dstw, offset) \
540      sljit_get_local_base(compiler, (dst), (dstw), (offset))
541    
542    #define READ_CHAR_MAX 0x7fffffff
543    
544  static uschar* bracketend(uschar* cc)  static pcre_uchar *bracketend(pcre_uchar *cc)
545  {  {
546  SLJIT_ASSERT((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND));  SLJIT_ASSERT((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND));
547  do cc += GET(cc, 1); while (*cc == OP_ALT);  do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 404  cc += 1 + LINK_SIZE; Line 550  cc += 1 + LINK_SIZE;
550  return cc;  return cc;
551  }  }
552    
553    static int no_alternatives(pcre_uchar *cc)
554    {
555    int count = 0;
556    SLJIT_ASSERT((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND));
557    do
558      {
559      cc += GET(cc, 1);
560      count++;
561      }
562    while (*cc == OP_ALT);
563    SLJIT_ASSERT(*cc >= OP_KET && *cc <= OP_KETRPOS);
564    return count;
565    }
566    
567    static int ones_in_half_byte[16] = {
568      /* 0 */ 0, 1, 1, 2, /* 4 */ 1, 2, 2, 3,
569      /* 8 */ 1, 2, 2, 3, /* 12 */ 2, 3, 3, 4
570    };
571    
572  /* Functions whose might need modification for all new supported opcodes:  /* Functions whose might need modification for all new supported opcodes:
573   next_opcode   next_opcode
574   get_localspace   check_opcode_types
575   set_localptrs   set_private_data_ptrs
576   get_framesize   get_framesize
577   init_frame   init_frame
578   get_localsize   get_private_data_copy_length
579   copy_locals   copy_private_data
580   compile_hotpath   compile_matchingpath
581   compile_fallbackpath   compile_backtrackingpath
582  */  */
583    
584  static uschar *next_opcode(compiler_common *common, uschar *cc)  static pcre_uchar *next_opcode(compiler_common *common, pcre_uchar *cc)
585  {  {
586  SLJIT_UNUSED_ARG(common);  SLJIT_UNUSED_ARG(common);
587  switch(*cc)  switch(*cc)
# Line 434  switch(*cc) Line 599  switch(*cc)
599    case OP_WORDCHAR:    case OP_WORDCHAR:
600    case OP_ANY:    case OP_ANY:
601    case OP_ALLANY:    case OP_ALLANY:
602      case OP_NOTPROP:
603      case OP_PROP:
604    case OP_ANYNL:    case OP_ANYNL:
605    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
606    case OP_HSPACE:    case OP_HSPACE:
# Line 446  switch(*cc) Line 613  switch(*cc)
613    case OP_CIRCM:    case OP_CIRCM:
614    case OP_DOLL:    case OP_DOLL:
615    case OP_DOLLM:    case OP_DOLLM:
   case OP_TYPESTAR:  
   case OP_TYPEMINSTAR:  
   case OP_TYPEPLUS:  
   case OP_TYPEMINPLUS:  
   case OP_TYPEQUERY:  
   case OP_TYPEMINQUERY:  
   case OP_TYPEPOSSTAR:  
   case OP_TYPEPOSPLUS:  
   case OP_TYPEPOSQUERY:  
616    case OP_CRSTAR:    case OP_CRSTAR:
617    case OP_CRMINSTAR:    case OP_CRMINSTAR:
618    case OP_CRPLUS:    case OP_CRPLUS:
619    case OP_CRMINPLUS:    case OP_CRMINPLUS:
620    case OP_CRQUERY:    case OP_CRQUERY:
621    case OP_CRMINQUERY:    case OP_CRMINQUERY:
622      case OP_CRRANGE:
623      case OP_CRMINRANGE:
624      case OP_CRPOSSTAR:
625      case OP_CRPOSPLUS:
626      case OP_CRPOSQUERY:
627      case OP_CRPOSRANGE:
628      case OP_CLASS:
629      case OP_NCLASS:
630      case OP_REF:
631      case OP_REFI:
632      case OP_DNREF:
633      case OP_DNREFI:
634      case OP_RECURSE:
635      case OP_CALLOUT:
636      case OP_ALT:
637      case OP_KET:
638      case OP_KETRMAX:
639      case OP_KETRMIN:
640      case OP_KETRPOS:
641      case OP_REVERSE:
642      case OP_ASSERT:
643      case OP_ASSERT_NOT:
644      case OP_ASSERTBACK:
645      case OP_ASSERTBACK_NOT:
646      case OP_ONCE:
647      case OP_ONCE_NC:
648      case OP_BRA:
649      case OP_BRAPOS:
650      case OP_CBRA:
651      case OP_CBRAPOS:
652      case OP_COND:
653      case OP_SBRA:
654      case OP_SBRAPOS:
655      case OP_SCBRA:
656      case OP_SCBRAPOS:
657      case OP_SCOND:
658      case OP_CREF:
659      case OP_DNCREF:
660      case OP_RREF:
661      case OP_DNRREF:
662    case OP_DEF:    case OP_DEF:
663    case OP_BRAZERO:    case OP_BRAZERO:
664    case OP_BRAMINZERO:    case OP_BRAMINZERO:
665    case OP_BRAPOSZERO:    case OP_BRAPOSZERO:
666      case OP_PRUNE:
667      case OP_SKIP:
668      case OP_THEN:
669      case OP_COMMIT:
670    case OP_FAIL:    case OP_FAIL:
671    case OP_ACCEPT:    case OP_ACCEPT:
672    case OP_ASSERT_ACCEPT:    case OP_ASSERT_ACCEPT:
673      case OP_CLOSE:
674    case OP_SKIPZERO:    case OP_SKIPZERO:
675    return cc + 1;    return cc + PRIV(OP_lengths)[*cc];
   
   case OP_ANYBYTE:  
 #ifdef SUPPORT_UTF8  
   if (common->utf8) return NULL;  
 #endif  
   return cc + 1;  
676    
677    case OP_CHAR:    case OP_CHAR:
678    case OP_CHARI:    case OP_CHARI:
679    case OP_NOT:    case OP_NOT:
680    case OP_NOTI:    case OP_NOTI:
   
681    case OP_STAR:    case OP_STAR:
682    case OP_MINSTAR:    case OP_MINSTAR:
683    case OP_PLUS:    case OP_PLUS:
684    case OP_MINPLUS:    case OP_MINPLUS:
685    case OP_QUERY:    case OP_QUERY:
686    case OP_MINQUERY:    case OP_MINQUERY:
687      case OP_UPTO:
688      case OP_MINUPTO:
689      case OP_EXACT:
690    case OP_POSSTAR:    case OP_POSSTAR:
691    case OP_POSPLUS:    case OP_POSPLUS:
692    case OP_POSQUERY:    case OP_POSQUERY:
693      case OP_POSUPTO:
694    case OP_STARI:    case OP_STARI:
695    case OP_MINSTARI:    case OP_MINSTARI:
696    case OP_PLUSI:    case OP_PLUSI:
697    case OP_MINPLUSI:    case OP_MINPLUSI:
698    case OP_QUERYI:    case OP_QUERYI:
699    case OP_MINQUERYI:    case OP_MINQUERYI:
700      case OP_UPTOI:
701      case OP_MINUPTOI:
702      case OP_EXACTI:
703    case OP_POSSTARI:    case OP_POSSTARI:
704    case OP_POSPLUSI:    case OP_POSPLUSI:
705    case OP_POSQUERYI:    case OP_POSQUERYI:
706      case OP_POSUPTOI:
707    case OP_NOTSTAR:    case OP_NOTSTAR:
708    case OP_NOTMINSTAR:    case OP_NOTMINSTAR:
709    case OP_NOTPLUS:    case OP_NOTPLUS:
710    case OP_NOTMINPLUS:    case OP_NOTMINPLUS:
711    case OP_NOTQUERY:    case OP_NOTQUERY:
712    case OP_NOTMINQUERY:    case OP_NOTMINQUERY:
713      case OP_NOTUPTO:
714      case OP_NOTMINUPTO:
715      case OP_NOTEXACT:
716    case OP_NOTPOSSTAR:    case OP_NOTPOSSTAR:
717    case OP_NOTPOSPLUS:    case OP_NOTPOSPLUS:
718    case OP_NOTPOSQUERY:    case OP_NOTPOSQUERY:
719      case OP_NOTPOSUPTO:
720    case OP_NOTSTARI:    case OP_NOTSTARI:
721    case OP_NOTMINSTARI:    case OP_NOTMINSTARI:
722    case OP_NOTPLUSI:    case OP_NOTPLUSI:
723    case OP_NOTMINPLUSI:    case OP_NOTMINPLUSI:
724    case OP_NOTQUERYI:    case OP_NOTQUERYI:
725    case OP_NOTMINQUERYI:    case OP_NOTMINQUERYI:
   case OP_NOTPOSSTARI:  
   case OP_NOTPOSPLUSI:  
   case OP_NOTPOSQUERYI:  
   cc += 2;  
 #ifdef SUPPORT_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:  
726    case OP_NOTUPTOI:    case OP_NOTUPTOI:
727    case OP_NOTMINUPTOI:    case OP_NOTMINUPTOI:
728    case OP_NOTEXACTI:    case OP_NOTEXACTI:
729      case OP_NOTPOSSTARI:
730      case OP_NOTPOSPLUSI:
731      case OP_NOTPOSQUERYI:
732    case OP_NOTPOSUPTOI:    case OP_NOTPOSUPTOI:
733    cc += 4;    cc += PRIV(OP_lengths)[*cc];
734  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
735    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]);
736  #endif  #endif
737    return cc;    return cc;
738    
739    case OP_NOTPROP:    /* Special cases. */
740    case OP_PROP:    case OP_TYPESTAR:
741      case OP_TYPEMINSTAR:
742      case OP_TYPEPLUS:
743      case OP_TYPEMINPLUS:
744      case OP_TYPEQUERY:
745      case OP_TYPEMINQUERY:
746    case OP_TYPEUPTO:    case OP_TYPEUPTO:
747    case OP_TYPEMINUPTO:    case OP_TYPEMINUPTO:
748    case OP_TYPEEXACT:    case OP_TYPEEXACT:
749      case OP_TYPEPOSSTAR:
750      case OP_TYPEPOSPLUS:
751      case OP_TYPEPOSQUERY:
752    case OP_TYPEPOSUPTO:    case OP_TYPEPOSUPTO:
753    case OP_REF:    return cc + PRIV(OP_lengths)[*cc] - 1;
   case OP_REFI:  
   case OP_CREF:  
   case OP_NCREF:  
   case OP_RREF:  
   case OP_NRREF:  
   case OP_CLOSE:  
   cc += 3;  
   return cc;  
754    
755    case OP_CRRANGE:    case OP_ANYBYTE:
756    case OP_CRMINRANGE:  #ifdef SUPPORT_UTF
757    return cc + 5;    if (common->utf) return NULL;
758    #endif
759    case OP_CLASS:    return cc + 1;
   case OP_NCLASS:  
   return cc + 33;  
760    
761  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
762    case OP_XCLASS:    case OP_XCLASS:
763    return cc + GET(cc, 1);    return cc + GET(cc, 1);
764  #endif  #endif
765    
766    case OP_RECURSE:    case OP_MARK:
767    case OP_ASSERT:    case OP_PRUNE_ARG:
768    case OP_ASSERT_NOT:    case OP_SKIP_ARG:
769    case OP_ASSERTBACK:    case OP_THEN_ARG:
770    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;  
771    
772    default:    default:
773      /* All opcodes are supported now! */
774      SLJIT_ASSERT_STOP();
775    return NULL;    return NULL;
776    }    }
777  }  }
778    
779  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)
780  {  {
781  int localspace = 0;  int count;
782  uschar *alternative;  pcre_uchar *slot;
783    
784  /* 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. */
785  while (cc < ccend)  while (cc < ccend)
786    {    {
787    switch(*cc)    switch(*cc)
788      {      {
789      case OP_ASSERT:      case OP_SET_SOM:
790      case OP_ASSERT_NOT:      common->has_set_som = TRUE;
791      case OP_ASSERTBACK:      common->might_be_empty = TRUE;
792      case OP_ASSERTBACK_NOT:      cc += 1;
793      case OP_ONCE:      break;
794      case OP_ONCE_NC:  
795      case OP_BRAPOS:      case OP_REF:
796      case OP_SBRA:      case OP_REFI:
797      case OP_SBRAPOS:      common->optimized_cbracket[GET2(cc, 1)] = 0;
798      case OP_SCOND:      cc += 1 + IMM2_SIZE;
     localspace += sizeof(sljit_w);  
     cc += 1 + LINK_SIZE;  
799      break;      break;
800    
801      case OP_CBRAPOS:      case OP_CBRAPOS:
802      case OP_SCBRAPOS:      case OP_SCBRAPOS:
803      localspace += sizeof(sljit_w);      common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] = 0;
804      cc += 1 + LINK_SIZE + 2;      cc += 1 + LINK_SIZE + IMM2_SIZE;
805      break;      break;
806    
807      case OP_COND:      case OP_COND:
     /* Might be a hidden SCOND. */  
     alternative = cc + GET(cc, 1);  
     if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)  
       localspace += sizeof(sljit_w);  
     cc += 1 + LINK_SIZE;  
     break;  
   
     default:  
     cc = next_opcode(common, cc);  
     if (cc == NULL)  
       return -1;  
     break;  
     }  
   }  
 return localspace;  
 }  
   
 static void set_localptrs(compiler_common *common, int localptr, uschar *ccend)  
 {  
 uschar *cc = common->start;  
 uschar *alternative;  
 while (cc < ccend)  
   {  
   switch(*cc)  
     {  
     case OP_ASSERT:  
     case OP_ASSERT_NOT:  
     case OP_ASSERTBACK:  
     case OP_ASSERTBACK_NOT:  
     case OP_ONCE:  
     case OP_ONCE_NC:  
     case OP_BRAPOS:  
     case OP_SBRA:  
     case OP_SBRAPOS:  
808      case OP_SCOND:      case OP_SCOND:
809      common->localptrs[cc - common->start] = localptr;      /* Only AUTO_CALLOUT can insert this opcode. We do
810      localptr += sizeof(sljit_w);         not intend to support this case. */
811        if (cc[1 + LINK_SIZE] == OP_CALLOUT)
812          return FALSE;
813      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
814      break;      break;
815    
816      case OP_CBRAPOS:      case OP_CREF:
817      case OP_SCBRAPOS:      common->optimized_cbracket[GET2(cc, 1)] = 0;
818      common->localptrs[cc - common->start] = localptr;      cc += 1 + IMM2_SIZE;
819      localptr += sizeof(sljit_w);      break;
820      cc += 1 + LINK_SIZE + 2;  
821        case OP_DNREF:
822        case OP_DNREFI:
823        case OP_DNCREF:
824        count = GET2(cc, 1 + IMM2_SIZE);
825        slot = common->name_table + GET2(cc, 1) * common->name_entry_size;
826        while (count-- > 0)
827          {
828          common->optimized_cbracket[GET2(slot, 0)] = 0;
829          slot += common->name_entry_size;
830          }
831        cc += 1 + 2 * IMM2_SIZE;
832      break;      break;
833    
834      case OP_COND:      case OP_RECURSE:
835      /* Might be a hidden SCOND. */      /* Set its value only once. */
836      alternative = cc + GET(cc, 1);      if (common->recursive_head_ptr == 0)
     if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)  
837        {        {
838        common->localptrs[cc - common->start] = localptr;        common->recursive_head_ptr = common->ovector_start;
839        localptr += sizeof(sljit_w);        common->ovector_start += sizeof(sljit_sw);
840        }        }
841      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
842      break;      break;
843    
844      default:      case OP_CALLOUT:
845      cc = next_opcode(common, cc);      if (common->capture_last_ptr == 0)
846      SLJIT_ASSERT(cc != NULL);        {
847          common->capture_last_ptr = common->ovector_start;
848          common->ovector_start += sizeof(sljit_sw);
849          }
850        cc += 2 + 2 * LINK_SIZE;
851      break;      break;
     }  
   }  
 }  
852    
853  /* Returns with -1 if no need for frame. */      case OP_THEN_ARG:
854  static int get_framesize(compiler_common *common, uschar *cc, BOOL recursive)      common->has_then = TRUE;
855  {      common->control_head_ptr = 1;
856  uschar *ccend = bracketend(cc);      /* Fall through. */
 int length = 0;  
 BOOL possessive = FALSE;  
 BOOL setsom_found = FALSE;  
857    
858  if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))      case OP_PRUNE_ARG:
859    {      common->needs_start_ptr = TRUE;
860    length = 3;      /* Fall through. */
   possessive = TRUE;  
   }  
861    
862  cc = next_opcode(common, cc);      case OP_MARK:
863  SLJIT_ASSERT(cc != NULL);      if (common->mark_ptr == 0)
 while (cc < ccend)  
   switch(*cc)  
     {  
     case OP_SET_SOM:  
     case OP_RECURSE:  
     if (!setsom_found)  
864        {        {
865        length += 2;        common->mark_ptr = common->ovector_start;
866        setsom_found = TRUE;        common->ovector_start += sizeof(sljit_sw);
867        }        }
868      cc += (*cc == OP_SET_SOM) ? 1 : 1 + LINK_SIZE;      cc += 1 + 2 + cc[1];
869      break;      break;
870    
871      case OP_CBRA:      case OP_THEN:
872      case OP_CBRAPOS:      common->has_then = TRUE;
873      case OP_SCBRA:      common->control_head_ptr = 1;
874      case OP_SCBRAPOS:      /* Fall through. */
875      length += 3;  
876      cc += 1 + LINK_SIZE + 2;      case OP_PRUNE:
877        case OP_SKIP:
878        common->needs_start_ptr = TRUE;
879        cc += 1;
880        break;
881    
882        case OP_SKIP_ARG:
883        common->control_head_ptr = 1;
884        common->has_skip_arg = TRUE;
885        cc += 1 + 2 + cc[1];
886      break;      break;
887    
888      default:      default:
889      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
890      SLJIT_ASSERT(cc != NULL);      if (cc == NULL)
891          return FALSE;
892      break;      break;
893      }      }
894      }
895  /* 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;  
896  }  }
897    
898  static void init_frame(compiler_common *common, uschar *cc, int stackpos, int stacktop, BOOL recursive)  static int get_class_iterator_size(pcre_uchar *cc)
899  {  {
900  DEFINE_COMPILER;  switch(*cc)
901  uschar *ccend = bracketend(cc);    {
902  BOOL setsom_found = FALSE;    case OP_CRSTAR:
903  int offset;    case OP_CRPLUS:
904      return 2;
905    
906  /* >= 1 + shortest item size (2) */    case OP_CRMINSTAR:
907  SLJIT_ASSERT(stackpos >= stacktop + 2);    case OP_CRMINPLUS:
908      case OP_CRQUERY:
909      case OP_CRMINQUERY:
910      return 1;
911    
912  stackpos = STACK(stackpos);    case OP_CRRANGE:
913  if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))    case OP_CRMINRANGE:
914    cc = next_opcode(common, cc);    if (GET2(cc, 1) == GET2(cc, 1 + IMM2_SIZE))
915  SLJIT_ASSERT(cc != NULL);      return 0;
916  while (cc < ccend)    return 2;
917    switch(*cc)  
918      default:
919      return 0;
920      }
921    }
922    
923    static BOOL detect_repeat(compiler_common *common, pcre_uchar *begin)
924    {
925    pcre_uchar *end = bracketend(begin);
926    pcre_uchar *next;
927    pcre_uchar *next_end;
928    pcre_uchar *max_end;
929    pcre_uchar type;
930    sljit_sw length = end - begin;
931    int min, max, i;
932    
933    /* Detect fixed iterations first. */
934    if (end[-(1 + LINK_SIZE)] != OP_KET)
935      return FALSE;
936    
937    /* Already detected repeat. */
938    if (common->private_data_ptrs[end - common->start - LINK_SIZE] != 0)
939      return TRUE;
940    
941    next = end;
942    min = 1;
943    while (1)
944      {
945      if (*next != *begin)
946        break;
947      next_end = bracketend(next);
948      if (next_end - next != length || memcmp(begin, next, IN_UCHARS(length)) != 0)
949        break;
950      next = next_end;
951      min++;
952      }
953    
954    if (min == 2)
955      return FALSE;
956    
957    max = 0;
958    max_end = next;
959    if (*next == OP_BRAZERO || *next == OP_BRAMINZERO)
960      {
961      type = *next;
962      while (1)
963      {      {
964      case OP_SET_SOM:      if (next[0] != type || next[1] != OP_BRA || next[2 + LINK_SIZE] != *begin)
965      case OP_RECURSE:        break;
966      if (!setsom_found)      next_end = bracketend(next + 2 + LINK_SIZE);
967        if (next_end - next != (length + 2 + LINK_SIZE) || memcmp(begin, next + 2 + LINK_SIZE, IN_UCHARS(length)) != 0)
968          break;
969        next = next_end;
970        max++;
971        }
972    
973      if (next[0] == type && next[1] == *begin && max >= 1)
974        {
975        next_end = bracketend(next + 1);
976        if (next_end - next == (length + 1) && memcmp(begin, next + 1, IN_UCHARS(length)) == 0)
977        {        {
978        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));        for (i = 0; i < max; i++, next_end += 1 + LINK_SIZE)
979        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setstrbegin);          if (*next_end != OP_KET)
980        stackpos += (int)sizeof(sljit_w);            break;
981        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);  
982        stackpos += (int)sizeof(sljit_w);        if (i == max)
983        setsom_found = TRUE;          {
984            common->private_data_ptrs[max_end - common->start - LINK_SIZE] = next_end - max_end;
985            common->private_data_ptrs[max_end - common->start - LINK_SIZE + 1] = (type == OP_BRAZERO) ? OP_UPTO : OP_MINUPTO;
986            /* +2 the original and the last. */
987            common->private_data_ptrs[max_end - common->start - LINK_SIZE + 2] = max + 2;
988            if (min == 1)
989              return TRUE;
990            min--;
991            max_end -= (1 + LINK_SIZE) + GET(max_end, -LINK_SIZE);
992            }
993        }        }
994      cc += (*cc == OP_SET_SOM) ? 1 : 1 + LINK_SIZE;      }
995      break;    }
996    
997      case OP_CBRA:  if (min >= 3)
998      case OP_CBRAPOS:    {
999      case OP_SCBRA:    common->private_data_ptrs[end - common->start - LINK_SIZE] = max_end - end;
1000      case OP_SCBRAPOS:    common->private_data_ptrs[end - common->start - LINK_SIZE + 1] = OP_EXACT;
1001      offset = (GET2(cc, 1 + LINK_SIZE)) << 1;    common->private_data_ptrs[end - common->start - LINK_SIZE + 2] = min;
1002      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));    return TRUE;
1003      stackpos += (int)sizeof(sljit_w);    }
1004      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset));  
1005      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset + 1));  return FALSE;
1006      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);  }
1007      stackpos += (int)sizeof(sljit_w);  
1008      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);  #define CASE_ITERATOR_PRIVATE_DATA_1 \
1009      stackpos += (int)sizeof(sljit_w);      case OP_MINSTAR: \
1010        case OP_MINPLUS: \
1011        case OP_QUERY: \
1012        case OP_MINQUERY: \
1013        case OP_MINSTARI: \
1014        case OP_MINPLUSI: \
1015        case OP_QUERYI: \
1016        case OP_MINQUERYI: \
1017        case OP_NOTMINSTAR: \
1018        case OP_NOTMINPLUS: \
1019        case OP_NOTQUERY: \
1020        case OP_NOTMINQUERY: \
1021        case OP_NOTMINSTARI: \
1022        case OP_NOTMINPLUSI: \
1023        case OP_NOTQUERYI: \
1024        case OP_NOTMINQUERYI:
1025    
1026      cc += 1 + LINK_SIZE + 2;  #define CASE_ITERATOR_PRIVATE_DATA_2A \
1027      break;      case OP_STAR: \
1028        case OP_PLUS: \
1029        case OP_STARI: \
1030        case OP_PLUSI: \
1031        case OP_NOTSTAR: \
1032        case OP_NOTPLUS: \
1033        case OP_NOTSTARI: \
1034        case OP_NOTPLUSI:
1035    
1036      default:  #define CASE_ITERATOR_PRIVATE_DATA_2B \
1037      cc = next_opcode(common, cc);      case OP_UPTO: \
1038      SLJIT_ASSERT(cc != NULL);      case OP_MINUPTO: \
1039      break;      case OP_UPTOI: \
1040      }      case OP_MINUPTOI: \
1041        case OP_NOTUPTO: \
1042        case OP_NOTMINUPTO: \
1043        case OP_NOTUPTOI: \
1044        case OP_NOTMINUPTOI:
1045    
1046  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_end);  #define CASE_ITERATOR_TYPE_PRIVATE_DATA_1 \
1047  SLJIT_ASSERT(stackpos == STACK(stacktop));      case OP_TYPEMINSTAR: \
1048  }      case OP_TYPEMINPLUS: \
1049        case OP_TYPEQUERY: \
1050        case OP_TYPEMINQUERY:
1051    
1052    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2A \
1053        case OP_TYPESTAR: \
1054        case OP_TYPEPLUS:
1055    
1056    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2B \
1057        case OP_TYPEUPTO: \
1058        case OP_TYPEMINUPTO:
1059    
1060  static SLJIT_INLINE int get_localsize(compiler_common *common, uschar *cc, uschar *ccend)  static void set_private_data_ptrs(compiler_common *common, int *private_data_start, pcre_uchar *ccend)
1061  {  {
1062  int localsize = 2;  pcre_uchar *cc = common->start;
1063  uschar *alternative;  pcre_uchar *alternative;
1064  /* Calculate the sum of the local variables. */  pcre_uchar *end = NULL;
1065    int private_data_ptr = *private_data_start;
1066    int space, size, bracketlen;
1067    BOOL repeat_check = TRUE;
1068    
1069  while (cc < ccend)  while (cc < ccend)
1070    {    {
1071      space = 0;
1072      size = 0;
1073      bracketlen = 0;
1074      if (private_data_ptr > SLJIT_MAX_LOCAL_SIZE)
1075        break;
1076    
1077      if (repeat_check && (*cc == OP_ONCE || *cc == OP_ONCE_NC || *cc == OP_BRA || *cc == OP_CBRA || *cc == OP_COND))
1078        {
1079        if (detect_repeat(common, cc))
1080          {
1081          /* These brackets are converted to repeats, so no global
1082          based single character repeat is allowed. */
1083          if (cc >= end)
1084            end = bracketend(cc);
1085          }
1086        }
1087      repeat_check = TRUE;
1088    
1089    switch(*cc)    switch(*cc)
1090      {      {
1091        case OP_KET:
1092        if (common->private_data_ptrs[cc + 1 - common->start] != 0)
1093          {
1094          common->private_data_ptrs[cc - common->start] = private_data_ptr;
1095          private_data_ptr += sizeof(sljit_sw);
1096          cc += common->private_data_ptrs[cc + 1 - common->start];
1097          }
1098        cc += 1 + LINK_SIZE;
1099        break;
1100    
1101      case OP_ASSERT:      case OP_ASSERT:
1102      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1103      case OP_ASSERTBACK:      case OP_ASSERTBACK:
# Line 832  while (cc < ccend) Line 1108  while (cc < ccend)
1108      case OP_SBRA:      case OP_SBRA:
1109      case OP_SBRAPOS:      case OP_SBRAPOS:
1110      case OP_SCOND:      case OP_SCOND:
1111      localsize++;      common->private_data_ptrs[cc - common->start] = private_data_ptr;
1112      cc += 1 + LINK_SIZE;      private_data_ptr += sizeof(sljit_sw);
1113      break;      bracketlen = 1 + LINK_SIZE;
   
     case OP_CBRA:  
     case OP_SCBRA:  
     localsize++;  
     cc += 1 + LINK_SIZE + 2;  
1114      break;      break;
1115    
1116      case OP_CBRAPOS:      case OP_CBRAPOS:
1117      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1118      localsize += 2;      common->private_data_ptrs[cc - common->start] = private_data_ptr;
1119      cc += 1 + LINK_SIZE + 2;      private_data_ptr += sizeof(sljit_sw);
1120        bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
1121      break;      break;
1122    
1123      case OP_COND:      case OP_COND:
1124      /* Might be a hidden SCOND. */      /* Might be a hidden SCOND. */
1125      alternative = cc + GET(cc, 1);      alternative = cc + GET(cc, 1);
1126      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1127        localsize++;        {
1128      cc += 1 + LINK_SIZE;        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1129          private_data_ptr += sizeof(sljit_sw);
1130          }
1131        bracketlen = 1 + LINK_SIZE;
1132        break;
1133    
1134        case OP_BRA:
1135        bracketlen = 1 + LINK_SIZE;
1136        break;
1137    
1138        case OP_CBRA:
1139        case OP_SCBRA:
1140        bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
1141        break;
1142    
1143        case OP_BRAZERO:
1144        case OP_BRAMINZERO:
1145        case OP_BRAPOSZERO:
1146        repeat_check = FALSE;
1147        size = 1;
1148        break;
1149    
1150        CASE_ITERATOR_PRIVATE_DATA_1
1151        space = 1;
1152        size = -2;
1153        break;
1154    
1155        CASE_ITERATOR_PRIVATE_DATA_2A
1156        space = 2;
1157        size = -2;
1158        break;
1159    
1160        CASE_ITERATOR_PRIVATE_DATA_2B
1161        space = 2;
1162        size = -(2 + IMM2_SIZE);
1163        break;
1164    
1165        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1166        space = 1;
1167        size = 1;
1168        break;
1169    
1170        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1171        if (cc[1] != OP_ANYNL && cc[1] != OP_EXTUNI)
1172          space = 2;
1173        size = 1;
1174        break;
1175    
1176        case OP_TYPEUPTO:
1177        if (cc[1 + IMM2_SIZE] != OP_ANYNL && cc[1 + IMM2_SIZE] != OP_EXTUNI)
1178          space = 2;
1179        size = 1 + IMM2_SIZE;
1180        break;
1181    
1182        case OP_TYPEMINUPTO:
1183        space = 2;
1184        size = 1 + IMM2_SIZE;
1185        break;
1186    
1187        case OP_CLASS:
1188        case OP_NCLASS:
1189        size += 1 + 32 / sizeof(pcre_uchar);
1190        space = get_class_iterator_size(cc + size);
1191        break;
1192    
1193    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1194        case OP_XCLASS:
1195        size = GET(cc, 1);
1196        space = get_class_iterator_size(cc + size);
1197      break;      break;
1198    #endif
1199    
1200      default:      default:
1201      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
1202      SLJIT_ASSERT(cc != NULL);      SLJIT_ASSERT(cc != NULL);
1203      break;      break;
1204      }      }
1205    
1206      /* Character iterators, which are not inside a repeated bracket,
1207         gets a private slot instead of allocating it on the stack. */
1208      if (space > 0 && cc >= end)
1209        {
1210        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1211        private_data_ptr += sizeof(sljit_sw) * space;
1212        }
1213    
1214      if (size != 0)
1215        {
1216        if (size < 0)
1217          {
1218          cc += -size;
1219    #ifdef SUPPORT_UTF
1220          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1221    #endif
1222          }
1223        else
1224          cc += size;
1225        }
1226    
1227      if (bracketlen > 0)
1228        {
1229        if (cc >= end)
1230          {
1231          end = bracketend(cc);
1232          if (end[-1 - LINK_SIZE] == OP_KET)
1233            end = NULL;
1234          }
1235        cc += bracketlen;
1236        }
1237    }    }
1238  SLJIT_ASSERT(cc == ccend);  *private_data_start = private_data_ptr;
 return localsize;  
1239  }  }
1240    
1241  static void copy_locals(compiler_common *common, uschar *cc, uschar *ccend,  /* Returns with a frame_types (always < 0) if no need for frame. */
1242    BOOL save, int stackptr, int stacktop)  static int get_framesize(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, BOOL recursive, BOOL *needs_control_head)
1243  {  {
1244  DEFINE_COMPILER;  int length = 0;
1245  int srcw[2];  int possessive = 0;
1246  int count;  BOOL stack_restore = FALSE;
1247  BOOL tmp1next = TRUE;  BOOL setsom_found = recursive;
1248  BOOL tmp1empty = TRUE;  BOOL setmark_found = recursive;
1249  BOOL tmp2empty = TRUE;  /* The last capture is a local variable even for recursions. */
1250  uschar *alternative;  BOOL capture_last_found = FALSE;
1251  enum {  
1252    start,  #if defined DEBUG_FORCE_CONTROL_HEAD && DEBUG_FORCE_CONTROL_HEAD
1253    loop,  SLJIT_ASSERT(common->control_head_ptr != 0);
1254    end  *needs_control_head = TRUE;
1255  } status;  #else
1256    *needs_control_head = FALSE;
1257  status = save ? start : loop;  #endif
 stackptr = STACK(stackptr - 2);  
 stacktop = STACK(stacktop - 1);  
1258    
1259  if (!save)  if (ccend == NULL)
1260    {    {
1261    stackptr += sizeof(sljit_w);    ccend = bracketend(cc) - (1 + LINK_SIZE);
1262    if (stackptr < stacktop)    if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))
     {  
     OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);  
     stackptr += sizeof(sljit_w);  
     tmp1empty = FALSE;  
     }  
   if (stackptr < stacktop)  
1263      {      {
1264      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);      possessive = length = (common->capture_last_ptr != 0) ? 5 : 3;
1265      stackptr += sizeof(sljit_w);      /* This is correct regardless of common->capture_last_ptr. */
1266      tmp2empty = FALSE;      capture_last_found = TRUE;
1267      }      }
1268    /* The tmp1next must be TRUE in either way. */    cc = next_opcode(common, cc);
1269    }    }
1270    
1271  while (status != end)  SLJIT_ASSERT(cc != NULL);
1272    {  while (cc < ccend)
1273    count = 0;    switch(*cc)
   switch(status)  
1274      {      {
1275      case start:      case OP_SET_SOM:
1276      SLJIT_ASSERT(save);      SLJIT_ASSERT(common->has_set_som);
1277      count = 1;      stack_restore = TRUE;
1278      srcw[0] = RECURSIVE_HEAD;      if (!setsom_found)
1279      status = loop;        {
1280          length += 2;
1281          setsom_found = TRUE;
1282          }
1283        cc += 1;
1284      break;      break;
1285    
1286      case loop:      case OP_MARK:
1287      if (cc >= ccend)      case OP_PRUNE_ARG:
1288        case OP_THEN_ARG:
1289        SLJIT_ASSERT(common->mark_ptr != 0);
1290        stack_restore = TRUE;
1291        if (!setmark_found)
1292        {        {
1293        status = end;        length += 2;
1294        break;        setmark_found = TRUE;
1295        }        }
1296        if (common->control_head_ptr != 0)
1297          *needs_control_head = TRUE;
1298        cc += 1 + 2 + cc[1];
1299        break;
1300    
1301      switch(*cc)      case OP_RECURSE:
1302        stack_restore = TRUE;
1303        if (common->has_set_som && !setsom_found)
1304        {        {
1305        case OP_ASSERT:        length += 2;
1306        case OP_ASSERT_NOT:        setsom_found = TRUE;
1307        case OP_ASSERTBACK:        }
1308        case OP_ASSERTBACK_NOT:      if (common->mark_ptr != 0 && !setmark_found)
1309        case OP_ONCE:        {
1310        case OP_ONCE_NC:        length += 2;
1311        case OP_BRAPOS:        setmark_found = TRUE;
       case OP_SBRA:  
       case OP_SBRAPOS:  
       case OP_SCOND:  
       count = 1;  
       srcw[0] = PRIV(cc);  
       SLJIT_ASSERT(srcw[0] != 0);  
       cc += 1 + LINK_SIZE;  
       break;  
   
       case OP_CBRA:  
       case OP_SCBRA:  
       count = 1;  
       srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));  
       cc += 1 + LINK_SIZE + 2;  
       break;  
   
       case OP_CBRAPOS:  
       case OP_SCBRAPOS:  
       count = 2;  
       srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));  
       srcw[0] = PRIV(cc);  
       SLJIT_ASSERT(srcw[0] != 0);  
       cc += 1 + LINK_SIZE + 2;  
       break;  
   
       case OP_COND:  
       /* Might be a hidden SCOND. */  
       alternative = cc + GET(cc, 1);  
       if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)  
         {  
         count = 1;  
         srcw[0] = PRIV(cc);  
         SLJIT_ASSERT(srcw[0] != 0);  
         }  
       cc += 1 + LINK_SIZE;  
       break;  
   
       default:  
       cc = next_opcode(common, cc);  
       SLJIT_ASSERT(cc != NULL);  
       break;  
1312        }        }
1313        if (common->capture_last_ptr != 0 && !capture_last_found)
1314          {
1315          length += 2;
1316          capture_last_found = TRUE;
1317          }
1318        cc += 1 + LINK_SIZE;
1319      break;      break;
1320    
1321      case end:      case OP_CBRA:
1322      SLJIT_ASSERT_STOP();      case OP_CBRAPOS:
1323        case OP_SCBRA:
1324        case OP_SCBRAPOS:
1325        stack_restore = TRUE;
1326        if (common->capture_last_ptr != 0 && !capture_last_found)
1327          {
1328          length += 2;
1329          capture_last_found = TRUE;
1330          }
1331        length += 3;
1332        cc += 1 + LINK_SIZE + IMM2_SIZE;
1333        break;
1334    
1335        case OP_THEN:
1336        stack_restore = TRUE;
1337        if (common->control_head_ptr != 0)
1338          *needs_control_head = TRUE;
1339        cc ++;
1340        break;
1341    
1342        default:
1343        stack_restore = TRUE;
1344        /* Fall through. */
1345    
1346        case OP_NOT_WORD_BOUNDARY:
1347        case OP_WORD_BOUNDARY:
1348        case OP_NOT_DIGIT:
1349        case OP_DIGIT:
1350        case OP_NOT_WHITESPACE:
1351        case OP_WHITESPACE:
1352        case OP_NOT_WORDCHAR:
1353        case OP_WORDCHAR:
1354        case OP_ANY:
1355        case OP_ALLANY:
1356        case OP_ANYBYTE:
1357        case OP_NOTPROP:
1358        case OP_PROP:
1359        case OP_ANYNL:
1360        case OP_NOT_HSPACE:
1361        case OP_HSPACE:
1362        case OP_NOT_VSPACE:
1363        case OP_VSPACE:
1364        case OP_EXTUNI:
1365        case OP_EODN:
1366        case OP_EOD:
1367        case OP_CIRC:
1368        case OP_CIRCM:
1369        case OP_DOLL:
1370        case OP_DOLLM:
1371        case OP_CHAR:
1372        case OP_CHARI:
1373        case OP_NOT:
1374        case OP_NOTI:
1375    
1376        case OP_EXACT:
1377        case OP_POSSTAR:
1378        case OP_POSPLUS:
1379        case OP_POSQUERY:
1380        case OP_POSUPTO:
1381    
1382        case OP_EXACTI:
1383        case OP_POSSTARI:
1384        case OP_POSPLUSI:
1385        case OP_POSQUERYI:
1386        case OP_POSUPTOI:
1387    
1388        case OP_NOTEXACT:
1389        case OP_NOTPOSSTAR:
1390        case OP_NOTPOSPLUS:
1391        case OP_NOTPOSQUERY:
1392        case OP_NOTPOSUPTO:
1393    
1394        case OP_NOTEXACTI:
1395        case OP_NOTPOSSTARI:
1396        case OP_NOTPOSPLUSI:
1397        case OP_NOTPOSQUERYI:
1398        case OP_NOTPOSUPTOI:
1399    
1400        case OP_TYPEEXACT:
1401        case OP_TYPEPOSSTAR:
1402        case OP_TYPEPOSPLUS:
1403        case OP_TYPEPOSQUERY:
1404        case OP_TYPEPOSUPTO:
1405    
1406        case OP_CLASS:
1407        case OP_NCLASS:
1408        case OP_XCLASS:
1409    
1410        cc = next_opcode(common, cc);
1411        SLJIT_ASSERT(cc != NULL);
1412      break;      break;
1413      }      }
1414    
1415    while (count > 0)  /* Possessive quantifiers can use a special case. */
1416    if (SLJIT_UNLIKELY(possessive == length))
1417      return stack_restore ? no_frame : no_stack;
1418    
1419    if (length > 0)
1420      return length + 1;
1421    return stack_restore ? no_frame : no_stack;
1422    }
1423    
1424    static void init_frame(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, int stackpos, int stacktop, BOOL recursive)
1425    {
1426    DEFINE_COMPILER;
1427    BOOL setsom_found = recursive;
1428    BOOL setmark_found = recursive;
1429    /* The last capture is a local variable even for recursions. */
1430    BOOL capture_last_found = FALSE;
1431    int offset;
1432    
1433    /* >= 1 + shortest item size (2) */
1434    SLJIT_UNUSED_ARG(stacktop);
1435    SLJIT_ASSERT(stackpos >= stacktop + 2);
1436    
1437    stackpos = STACK(stackpos);
1438    if (ccend == NULL)
1439      {
1440      ccend = bracketend(cc) - (1 + LINK_SIZE);
1441      if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))
1442        cc = next_opcode(common, cc);
1443      }
1444    
1445    SLJIT_ASSERT(cc != NULL);
1446    while (cc < ccend)
1447      switch(*cc)
1448      {      {
1449      count--;      case OP_SET_SOM:
1450      if (save)      SLJIT_ASSERT(common->has_set_som);
1451        if (!setsom_found)
1452        {        {
1453        if (tmp1next)        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(0));
1454          {        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1455          if (!tmp1empty)        stackpos += (int)sizeof(sljit_sw);
1456            {        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1457            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        stackpos += (int)sizeof(sljit_sw);
1458            stackptr += sizeof(sljit_w);        setsom_found = TRUE;
           }  
         OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);  
         tmp1empty = FALSE;  
         tmp1next = FALSE;  
         }  
       else  
         {  
         if (!tmp2empty)  
           {  
           OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);  
           stackptr += sizeof(sljit_w);  
           }  
         OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);  
         tmp2empty = FALSE;  
         tmp1next = TRUE;  
         }  
1459        }        }
1460      else      cc += 1;
1461        break;
1462    
1463        case OP_MARK:
1464        case OP_PRUNE_ARG:
1465        case OP_THEN_ARG:
1466        SLJIT_ASSERT(common->mark_ptr != 0);
1467        if (!setmark_found)
1468          {
1469          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->mark_ptr);
1470          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
1471          stackpos += (int)sizeof(sljit_sw);
1472          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1473          stackpos += (int)sizeof(sljit_sw);
1474          setmark_found = TRUE;
1475          }
1476        cc += 1 + 2 + cc[1];
1477        break;
1478    
1479        case OP_RECURSE:
1480        if (common->has_set_som && !setsom_found)
1481        {        {
1482        if (tmp1next)        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(0));
1483          {        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1484          SLJIT_ASSERT(!tmp1empty);        stackpos += (int)sizeof(sljit_sw);
1485          OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count], TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1486          tmp1empty = stackptr >= stacktop;        stackpos += (int)sizeof(sljit_sw);
1487          if (!tmp1empty)        setsom_found = TRUE;
1488            {        }
1489            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);      if (common->mark_ptr != 0 && !setmark_found)
1490            stackptr += sizeof(sljit_w);        {
1491            }        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->mark_ptr);
1492          tmp1next = FALSE;        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
1493          }        stackpos += (int)sizeof(sljit_sw);
1494        else        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1495          {        stackpos += (int)sizeof(sljit_sw);
1496          SLJIT_ASSERT(!tmp2empty);        setmark_found = TRUE;
1497          OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count], TMP2, 0);        }
1498          tmp2empty = stackptr >= stacktop;      if (common->capture_last_ptr != 0 && !capture_last_found)
1499          if (!tmp2empty)        {
1500            {        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->capture_last_ptr);
1501            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1502            stackptr += sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1503            }        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1504          tmp1next = TRUE;        stackpos += (int)sizeof(sljit_sw);
1505          }        capture_last_found = TRUE;
1506          }
1507        cc += 1 + LINK_SIZE;
1508        break;
1509    
1510        case OP_CBRA:
1511        case OP_CBRAPOS:
1512        case OP_SCBRA:
1513        case OP_SCBRAPOS:
1514        if (common->capture_last_ptr != 0 && !capture_last_found)
1515          {
1516          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->capture_last_ptr);
1517          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1518          stackpos += (int)sizeof(sljit_sw);
1519          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1520          stackpos += (int)sizeof(sljit_sw);
1521          capture_last_found = TRUE;
1522        }        }
1523        offset = (GET2(cc, 1 + LINK_SIZE)) << 1;
1524        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));
1525        stackpos += (int)sizeof(sljit_sw);
1526        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(offset));
1527        OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(offset + 1));
1528        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1529        stackpos += (int)sizeof(sljit_sw);
1530        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);
1531        stackpos += (int)sizeof(sljit_sw);
1532    
1533        cc += 1 + LINK_SIZE + IMM2_SIZE;
1534        break;
1535    
1536        default:
1537        cc = next_opcode(common, cc);
1538        SLJIT_ASSERT(cc != NULL);
1539        break;
1540      }      }
   }  
1541    
1542  if (save)  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, 0);
1543    SLJIT_ASSERT(stackpos == STACK(stacktop));
1544    }
1545    
1546    static SLJIT_INLINE int get_private_data_copy_length(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, BOOL needs_control_head)
1547    {
1548    int private_data_length = needs_control_head ? 3 : 2;
1549    int size;
1550    pcre_uchar *alternative;
1551    /* Calculate the sum of the private machine words. */
1552    while (cc < ccend)
1553      {
1554      size = 0;
1555      switch(*cc)
1556        {
1557        case OP_KET:
1558        if (PRIVATE_DATA(cc) != 0)
1559          {
1560          private_data_length++;
1561          SLJIT_ASSERT(PRIVATE_DATA(cc + 1) != 0);
1562          cc += PRIVATE_DATA(cc + 1);
1563          }
1564        cc += 1 + LINK_SIZE;
1565        break;
1566    
1567        case OP_ASSERT:
1568        case OP_ASSERT_NOT:
1569        case OP_ASSERTBACK:
1570        case OP_ASSERTBACK_NOT:
1571        case OP_ONCE:
1572        case OP_ONCE_NC:
1573        case OP_BRAPOS:
1574        case OP_SBRA:
1575        case OP_SBRAPOS:
1576        case OP_SCOND:
1577        private_data_length++;
1578        SLJIT_ASSERT(PRIVATE_DATA(cc) != 0);
1579        cc += 1 + LINK_SIZE;
1580        break;
1581    
1582        case OP_CBRA:
1583        case OP_SCBRA:
1584        if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1585          private_data_length++;
1586        cc += 1 + LINK_SIZE + IMM2_SIZE;
1587        break;
1588    
1589        case OP_CBRAPOS:
1590        case OP_SCBRAPOS:
1591        private_data_length += 2;
1592        cc += 1 + LINK_SIZE + IMM2_SIZE;
1593        break;
1594    
1595        case OP_COND:
1596        /* Might be a hidden SCOND. */
1597        alternative = cc + GET(cc, 1);
1598        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1599          private_data_length++;
1600        cc += 1 + LINK_SIZE;
1601        break;
1602    
1603        CASE_ITERATOR_PRIVATE_DATA_1
1604        if (PRIVATE_DATA(cc))
1605          private_data_length++;
1606        cc += 2;
1607    #ifdef SUPPORT_UTF
1608        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1609    #endif
1610        break;
1611    
1612        CASE_ITERATOR_PRIVATE_DATA_2A
1613        if (PRIVATE_DATA(cc))
1614          private_data_length += 2;
1615        cc += 2;
1616    #ifdef SUPPORT_UTF
1617        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1618    #endif
1619        break;
1620    
1621        CASE_ITERATOR_PRIVATE_DATA_2B
1622        if (PRIVATE_DATA(cc))
1623          private_data_length += 2;
1624        cc += 2 + IMM2_SIZE;
1625    #ifdef SUPPORT_UTF
1626        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1627    #endif
1628        break;
1629    
1630        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1631        if (PRIVATE_DATA(cc))
1632          private_data_length++;
1633        cc += 1;
1634        break;
1635    
1636        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1637        if (PRIVATE_DATA(cc))
1638          private_data_length += 2;
1639        cc += 1;
1640        break;
1641    
1642        CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1643        if (PRIVATE_DATA(cc))
1644          private_data_length += 2;
1645        cc += 1 + IMM2_SIZE;
1646        break;
1647    
1648        case OP_CLASS:
1649        case OP_NCLASS:
1650    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1651        case OP_XCLASS:
1652        size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
1653    #else
1654        size = 1 + 32 / (int)sizeof(pcre_uchar);
1655    #endif
1656        if (PRIVATE_DATA(cc))
1657          private_data_length += get_class_iterator_size(cc + size);
1658        cc += size;
1659        break;
1660    
1661        default:
1662        cc = next_opcode(common, cc);
1663        SLJIT_ASSERT(cc != NULL);
1664        break;
1665        }
1666      }
1667    SLJIT_ASSERT(cc == ccend);
1668    return private_data_length;
1669    }
1670    
1671    static void copy_private_data(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend,
1672      BOOL save, int stackptr, int stacktop, BOOL needs_control_head)
1673    {
1674    DEFINE_COMPILER;
1675    int srcw[2];
1676    int count, size;
1677    BOOL tmp1next = TRUE;
1678    BOOL tmp1empty = TRUE;
1679    BOOL tmp2empty = TRUE;
1680    pcre_uchar *alternative;
1681    enum {
1682      start,
1683      loop,
1684      end
1685    } status;
1686    
1687    status = save ? start : loop;
1688    stackptr = STACK(stackptr - 2);
1689    stacktop = STACK(stacktop - 1);
1690    
1691    if (!save)
1692      {
1693      stackptr += (needs_control_head ? 2 : 1) * sizeof(sljit_sw);
1694      if (stackptr < stacktop)
1695        {
1696        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1697        stackptr += sizeof(sljit_sw);
1698        tmp1empty = FALSE;
1699        }
1700      if (stackptr < stacktop)
1701        {
1702        OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1703        stackptr += sizeof(sljit_sw);
1704        tmp2empty = FALSE;
1705        }
1706      /* The tmp1next must be TRUE in either way. */
1707      }
1708    
1709    do
1710      {
1711      count = 0;
1712      switch(status)
1713        {
1714        case start:
1715        SLJIT_ASSERT(save && common->recursive_head_ptr != 0);
1716        count = 1;
1717        srcw[0] = common->recursive_head_ptr;
1718        if (needs_control_head)
1719          {
1720          SLJIT_ASSERT(common->control_head_ptr != 0);
1721          count = 2;
1722          srcw[1] = common->control_head_ptr;
1723          }
1724        status = loop;
1725        break;
1726    
1727        case loop:
1728        if (cc >= ccend)
1729          {
1730          status = end;
1731          break;
1732          }
1733    
1734        switch(*cc)
1735          {
1736          case OP_KET:
1737          if (PRIVATE_DATA(cc) != 0)
1738            {
1739            count = 1;
1740            srcw[0] = PRIVATE_DATA(cc);
1741            SLJIT_ASSERT(PRIVATE_DATA(cc + 1) != 0);
1742            cc += PRIVATE_DATA(cc + 1);
1743            }
1744          cc += 1 + LINK_SIZE;
1745          break;
1746    
1747          case OP_ASSERT:
1748          case OP_ASSERT_NOT:
1749          case OP_ASSERTBACK:
1750          case OP_ASSERTBACK_NOT:
1751          case OP_ONCE:
1752          case OP_ONCE_NC:
1753          case OP_BRAPOS:
1754          case OP_SBRA:
1755          case OP_SBRAPOS:
1756          case OP_SCOND:
1757          count = 1;
1758          srcw[0] = PRIVATE_DATA(cc);
1759          SLJIT_ASSERT(srcw[0] != 0);
1760          cc += 1 + LINK_SIZE;
1761          break;
1762    
1763          case OP_CBRA:
1764          case OP_SCBRA:
1765          if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1766            {
1767            count = 1;
1768            srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1769            }
1770          cc += 1 + LINK_SIZE + IMM2_SIZE;
1771          break;
1772    
1773          case OP_CBRAPOS:
1774          case OP_SCBRAPOS:
1775          count = 2;
1776          srcw[0] = PRIVATE_DATA(cc);
1777          srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1778          SLJIT_ASSERT(srcw[0] != 0 && srcw[1] != 0);
1779          cc += 1 + LINK_SIZE + IMM2_SIZE;
1780          break;
1781    
1782          case OP_COND:
1783          /* Might be a hidden SCOND. */
1784          alternative = cc + GET(cc, 1);
1785          if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1786            {
1787            count = 1;
1788            srcw[0] = PRIVATE_DATA(cc);
1789            SLJIT_ASSERT(srcw[0] != 0);
1790            }
1791          cc += 1 + LINK_SIZE;
1792          break;
1793    
1794          CASE_ITERATOR_PRIVATE_DATA_1
1795          if (PRIVATE_DATA(cc))
1796            {
1797            count = 1;
1798            srcw[0] = PRIVATE_DATA(cc);
1799            }
1800          cc += 2;
1801    #ifdef SUPPORT_UTF
1802          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1803    #endif
1804          break;
1805    
1806          CASE_ITERATOR_PRIVATE_DATA_2A
1807          if (PRIVATE_DATA(cc))
1808            {
1809            count = 2;
1810            srcw[0] = PRIVATE_DATA(cc);
1811            srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1812            }
1813          cc += 2;
1814    #ifdef SUPPORT_UTF
1815          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1816    #endif
1817          break;
1818    
1819          CASE_ITERATOR_PRIVATE_DATA_2B
1820          if (PRIVATE_DATA(cc))
1821            {
1822            count = 2;
1823            srcw[0] = PRIVATE_DATA(cc);
1824            srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1825            }
1826          cc += 2 + IMM2_SIZE;
1827    #ifdef SUPPORT_UTF
1828          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1829    #endif
1830          break;
1831    
1832          CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1833          if (PRIVATE_DATA(cc))
1834            {
1835            count = 1;
1836            srcw[0] = PRIVATE_DATA(cc);
1837            }
1838          cc += 1;
1839          break;
1840    
1841          CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1842          if (PRIVATE_DATA(cc))
1843            {
1844            count = 2;
1845            srcw[0] = PRIVATE_DATA(cc);
1846            srcw[1] = srcw[0] + sizeof(sljit_sw);
1847            }
1848          cc += 1;
1849          break;
1850    
1851          CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1852          if (PRIVATE_DATA(cc))
1853            {
1854            count = 2;
1855            srcw[0] = PRIVATE_DATA(cc);
1856            srcw[1] = srcw[0] + sizeof(sljit_sw);
1857            }
1858          cc += 1 + IMM2_SIZE;
1859          break;
1860    
1861          case OP_CLASS:
1862          case OP_NCLASS:
1863    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1864          case OP_XCLASS:
1865          size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
1866    #else
1867          size = 1 + 32 / (int)sizeof(pcre_uchar);
1868    #endif
1869          if (PRIVATE_DATA(cc))
1870            switch(get_class_iterator_size(cc + size))
1871              {
1872              case 1:
1873              count = 1;
1874              srcw[0] = PRIVATE_DATA(cc);
1875              break;
1876    
1877              case 2:
1878              count = 2;
1879              srcw[0] = PRIVATE_DATA(cc);
1880              srcw[1] = srcw[0] + sizeof(sljit_sw);
1881              break;
1882    
1883              default:
1884              SLJIT_ASSERT_STOP();
1885              break;
1886              }
1887          cc += size;
1888          break;
1889    
1890          default:
1891          cc = next_opcode(common, cc);
1892          SLJIT_ASSERT(cc != NULL);
1893          break;
1894          }
1895        break;
1896    
1897        case end:
1898        SLJIT_ASSERT_STOP();
1899        break;
1900        }
1901    
1902      while (count > 0)
1903        {
1904        count--;
1905        if (save)
1906          {
1907          if (tmp1next)
1908            {
1909            if (!tmp1empty)
1910              {
1911              OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1912              stackptr += sizeof(sljit_sw);
1913              }
1914            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), srcw[count]);
1915            tmp1empty = FALSE;
1916            tmp1next = FALSE;
1917            }
1918          else
1919            {
1920            if (!tmp2empty)
1921              {
1922              OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1923              stackptr += sizeof(sljit_sw);
1924              }
1925            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_SP), srcw[count]);
1926            tmp2empty = FALSE;
1927            tmp1next = TRUE;
1928            }
1929          }
1930        else
1931          {
1932          if (tmp1next)
1933            {
1934            SLJIT_ASSERT(!tmp1empty);
1935            OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), srcw[count], TMP1, 0);
1936            tmp1empty = stackptr >= stacktop;
1937            if (!tmp1empty)
1938              {
1939              OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1940              stackptr += sizeof(sljit_sw);
1941              }
1942            tmp1next = FALSE;
1943            }
1944          else
1945            {
1946            SLJIT_ASSERT(!tmp2empty);
1947            OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), srcw[count], TMP2, 0);
1948            tmp2empty = stackptr >= stacktop;
1949            if (!tmp2empty)
1950              {
1951              OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1952              stackptr += sizeof(sljit_sw);
1953              }
1954            tmp1next = TRUE;
1955            }
1956          }
1957        }
1958      }
1959    while (status != end);
1960    
1961    if (save)
1962      {
1963      if (tmp1next)
1964        {
1965        if (!tmp1empty)
1966          {
1967          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1968          stackptr += sizeof(sljit_sw);
1969          }
1970        if (!tmp2empty)
1971          {
1972          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1973          stackptr += sizeof(sljit_sw);
1974          }
1975        }
1976      else
1977        {
1978        if (!tmp2empty)
1979          {
1980          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1981          stackptr += sizeof(sljit_sw);
1982          }
1983        if (!tmp1empty)
1984          {
1985          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1986          stackptr += sizeof(sljit_sw);
1987          }
1988        }
1989      }
1990    SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));
1991    }
1992    
1993    static SLJIT_INLINE pcre_uchar *set_then_offsets(compiler_common *common, pcre_uchar *cc, pcre_uint8 *current_offset)
1994    {
1995    pcre_uchar *end = bracketend(cc);
1996    BOOL has_alternatives = cc[GET(cc, 1)] == OP_ALT;
1997    
1998    /* Assert captures then. */
1999    if (*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT)
2000      current_offset = NULL;
2001    /* Conditional block does not. */
2002    if (*cc == OP_COND || *cc == OP_SCOND)
2003      has_alternatives = FALSE;
2004    
2005    cc = next_opcode(common, cc);
2006    if (has_alternatives)
2007      current_offset = common->then_offsets + (cc - common->start);
2008    
2009    while (cc < end)
2010      {
2011      if ((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND))
2012        cc = set_then_offsets(common, cc, current_offset);
2013      else
2014        {
2015        if (*cc == OP_ALT && has_alternatives)
2016          current_offset = common->then_offsets + (cc + 1 + LINK_SIZE - common->start);
2017        if (*cc >= OP_THEN && *cc <= OP_THEN_ARG && current_offset != NULL)
2018          *current_offset = 1;
2019        cc = next_opcode(common, cc);
2020        }
2021      }
2022    
2023    return end;
2024    }
2025    
2026    #undef CASE_ITERATOR_PRIVATE_DATA_1
2027    #undef CASE_ITERATOR_PRIVATE_DATA_2A
2028    #undef CASE_ITERATOR_PRIVATE_DATA_2B
2029    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_1
2030    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
2031    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
2032    
2033    static SLJIT_INLINE BOOL is_powerof2(unsigned int value)
2034    {
2035    return (value & (value - 1)) == 0;
2036    }
2037    
2038    static SLJIT_INLINE void set_jumps(jump_list *list, struct sljit_label *label)
2039    {
2040    while (list)
2041      {
2042      /* sljit_set_label is clever enough to do nothing
2043      if either the jump or the label is NULL. */
2044      SET_LABEL(list->jump, label);
2045      list = list->next;
2046      }
2047    }
2048    
2049    static SLJIT_INLINE void add_jump(struct sljit_compiler *compiler, jump_list **list, struct sljit_jump *jump)
2050    {
2051    jump_list *list_item = sljit_alloc_memory(compiler, sizeof(jump_list));
2052    if (list_item)
2053      {
2054      list_item->next = *list;
2055      list_item->jump = jump;
2056      *list = list_item;
2057      }
2058    }
2059    
2060    static void add_stub(compiler_common *common, struct sljit_jump *start)
2061    {
2062    DEFINE_COMPILER;
2063    stub_list *list_item = sljit_alloc_memory(compiler, sizeof(stub_list));
2064    
2065    if (list_item)
2066      {
2067      list_item->start = start;
2068      list_item->quit = LABEL();
2069      list_item->next = common->stubs;
2070      common->stubs = list_item;
2071      }
2072    }
2073    
2074    static void flush_stubs(compiler_common *common)
2075    {
2076    DEFINE_COMPILER;
2077    stub_list *list_item = common->stubs;
2078    
2079    while (list_item)
2080      {
2081      JUMPHERE(list_item->start);
2082      add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));
2083      JUMPTO(SLJIT_JUMP, list_item->quit);
2084      list_item = list_item->next;
2085      }
2086    common->stubs = NULL;
2087    }
2088    
2089    static void add_label_addr(compiler_common *common, sljit_uw *update_addr)
2090    {
2091    DEFINE_COMPILER;
2092    label_addr_list *label_addr;
2093    
2094    label_addr = sljit_alloc_memory(compiler, sizeof(label_addr_list));
2095    if (label_addr == NULL)
2096      return;
2097    label_addr->label = LABEL();
2098    label_addr->update_addr = update_addr;
2099    label_addr->next = common->label_addrs;
2100    common->label_addrs = label_addr;
2101    }
2102    
2103    static SLJIT_INLINE void count_match(compiler_common *common)
2104    {
2105    DEFINE_COMPILER;
2106    
2107    OP2(SLJIT_SUB | SLJIT_SET_E, COUNT_MATCH, 0, COUNT_MATCH, 0, SLJIT_IMM, 1);
2108    add_jump(compiler, &common->calllimit, JUMP(SLJIT_ZERO));
2109    }
2110    
2111    static SLJIT_INLINE void allocate_stack(compiler_common *common, int size)
2112    {
2113    /* May destroy all locals and registers except TMP2. */
2114    DEFINE_COMPILER;
2115    
2116    OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_sw));
2117    #ifdef DESTROY_REGISTERS
2118    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);
2119    OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);
2120    OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);
2121    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), LOCALS0, TMP1, 0);
2122    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), LOCALS1, TMP1, 0);
2123    #endif
2124    add_stub(common, CMP(SLJIT_GREATER, STACK_TOP, 0, STACK_LIMIT, 0));
2125    }
2126    
2127    static SLJIT_INLINE void free_stack(compiler_common *common, int size)
2128    {
2129    DEFINE_COMPILER;
2130    OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_sw));
2131    }
2132    
2133    static sljit_uw * allocate_read_only_data(compiler_common *common, sljit_uw size)
2134    {
2135    DEFINE_COMPILER;
2136    sljit_uw *result;
2137    
2138    if (SLJIT_UNLIKELY(sljit_get_compiler_error(compiler)))
2139      return NULL;
2140    
2141    result = (sljit_uw *)SLJIT_MALLOC(size + sizeof(sljit_uw), compiler->allocator_data);
2142    if (SLJIT_UNLIKELY(result == NULL))
2143      {
2144      sljit_set_compiler_memory_error(compiler);
2145      return NULL;
2146      }
2147    
2148    *(void**)result = common->read_only_data_head;
2149    common->read_only_data_head = (void *)result;
2150    return result + 1;
2151    }
2152    
2153    static void free_read_only_data(void *current, void *allocator_data)
2154    {
2155    void *next;
2156    
2157    SLJIT_UNUSED_ARG(allocator_data);
2158    
2159    while (current != NULL)
2160      {
2161      next = *(void**)current;
2162      SLJIT_FREE(current, allocator_data);
2163      current = next;
2164      }
2165    }
2166    
2167    static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)
2168    {
2169    DEFINE_COMPILER;
2170    struct sljit_label *loop;
2171    int i;
2172    
2173    /* At this point we can freely use all temporary registers. */
2174    SLJIT_ASSERT(length > 1);
2175    /* TMP1 returns with begin - 1. */
2176    OP2(SLJIT_SUB, SLJIT_R0, 0, SLJIT_MEM1(SLJIT_S0), SLJIT_OFFSETOF(jit_arguments, begin), SLJIT_IMM, IN_UCHARS(1));
2177    if (length < 8)
2178      {
2179      for (i = 1; i < length; i++)
2180        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), OVECTOR(i), SLJIT_R0, 0);
2181      }
2182    else
2183      {
2184      GET_LOCAL_BASE(SLJIT_R1, 0, OVECTOR_START);
2185      OP1(SLJIT_MOV, SLJIT_R2, 0, SLJIT_IMM, length - 1);
2186      loop = LABEL();
2187      OP1(SLJIT_MOVU, SLJIT_MEM1(SLJIT_R1), sizeof(sljit_sw), SLJIT_R0, 0);
2188      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_IMM, 1);
2189      JUMPTO(SLJIT_NOT_ZERO, loop);
2190      }
2191    }
2192    
2193    static SLJIT_INLINE void do_reset_match(compiler_common *common, int length)
2194    {
2195    DEFINE_COMPILER;
2196    struct sljit_label *loop;
2197    int i;
2198    
2199    SLJIT_ASSERT(length > 1);
2200    /* OVECTOR(1) contains the "string begin - 1" constant. */
2201    if (length > 2)
2202      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(1));
2203    if (length < 8)
2204      {
2205      for (i = 2; i < length; i++)
2206        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), OVECTOR(i), TMP1, 0);
2207      }
2208    else
2209      {
2210      GET_LOCAL_BASE(TMP2, 0, OVECTOR_START + sizeof(sljit_sw));
2211      OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_IMM, length - 2);
2212      loop = LABEL();
2213      OP1(SLJIT_MOVU, SLJIT_MEM1(TMP2), sizeof(sljit_sw), TMP1, 0);
2214      OP2(SLJIT_SUB | SLJIT_SET_E, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, 1);
2215      JUMPTO(SLJIT_NOT_ZERO, loop);
2216      }
2217    
2218    OP1(SLJIT_MOV, STACK_TOP, 0, ARGUMENTS, 0);
2219    if (common->mark_ptr != 0)
2220      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->mark_ptr, SLJIT_IMM, 0);
2221    if (common->control_head_ptr != 0)
2222      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->control_head_ptr, SLJIT_IMM, 0);
2223    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(jit_arguments, stack));
2224    OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->start_ptr);
2225    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(struct sljit_stack, base));
2226    }
2227    
2228    static sljit_sw SLJIT_CALL do_search_mark(sljit_sw *current, const pcre_uchar *skip_arg)
2229    {
2230    while (current != NULL)
2231      {
2232      switch (current[-2])
2233        {
2234        case type_then_trap:
2235        break;
2236    
2237        case type_mark:
2238        if (STRCMP_UC_UC(skip_arg, (pcre_uchar *)current[-3]) == 0)
2239          return current[-4];
2240        break;
2241    
2242        default:
2243        SLJIT_ASSERT_STOP();
2244        break;
2245        }
2246      SLJIT_ASSERT(current > (sljit_sw*)current[-1]);
2247      current = (sljit_sw*)current[-1];
2248      }
2249    return -1;
2250    }
2251    
2252    static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)
2253    {
2254    DEFINE_COMPILER;
2255    struct sljit_label *loop;
2256    struct sljit_jump *early_quit;
2257    
2258    /* At this point we can freely use all registers. */
2259    OP1(SLJIT_MOV, SLJIT_S2, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(1));
2260    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), OVECTOR(1), STR_PTR, 0);
2261    
2262    OP1(SLJIT_MOV, SLJIT_R0, 0, ARGUMENTS, 0);
2263    if (common->mark_ptr != 0)
2264      OP1(SLJIT_MOV, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_SP), common->mark_ptr);
2265    OP1(SLJIT_MOV_SI, SLJIT_R1, 0, SLJIT_MEM1(SLJIT_R0), SLJIT_OFFSETOF(jit_arguments, offset_count));
2266    if (common->mark_ptr != 0)
2267      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_R0), SLJIT_OFFSETOF(jit_arguments, mark_ptr), SLJIT_R2, 0);
2268    OP2(SLJIT_SUB, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_R0), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));
2269    OP1(SLJIT_MOV, SLJIT_R0, 0, SLJIT_MEM1(SLJIT_R0), SLJIT_OFFSETOF(jit_arguments, begin));
2270    GET_LOCAL_BASE(SLJIT_S0, 0, OVECTOR_START);
2271    /* Unlikely, but possible */
2272    early_quit = CMP(SLJIT_EQUAL, SLJIT_R1, 0, SLJIT_IMM, 0);
2273    loop = LABEL();
2274    OP2(SLJIT_SUB, SLJIT_S1, 0, SLJIT_MEM1(SLJIT_S0), 0, SLJIT_R0, 0);
2275    OP2(SLJIT_ADD, SLJIT_S0, 0, SLJIT_S0, 0, SLJIT_IMM, sizeof(sljit_sw));
2276    /* Copy the integer value to the output buffer */
2277    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2278    OP2(SLJIT_ASHR, SLJIT_S1, 0, SLJIT_S1, 0, SLJIT_IMM, UCHAR_SHIFT);
2279    #endif
2280    OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_R2), sizeof(int), SLJIT_S1, 0);
2281    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_R1, 0, SLJIT_R1, 0, SLJIT_IMM, 1);
2282    JUMPTO(SLJIT_NOT_ZERO, loop);
2283    JUMPHERE(early_quit);
2284    
2285    /* Calculate the return value, which is the maximum ovector value. */
2286    if (topbracket > 1)
2287      {
2288      GET_LOCAL_BASE(SLJIT_R0, 0, OVECTOR_START + topbracket * 2 * sizeof(sljit_sw));
2289      OP1(SLJIT_MOV, SLJIT_R1, 0, SLJIT_IMM, topbracket + 1);
2290    
2291      /* OVECTOR(0) is never equal to SLJIT_S2. */
2292      loop = LABEL();
2293      OP1(SLJIT_MOVU, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_R0), -(2 * (sljit_sw)sizeof(sljit_sw)));
2294      OP2(SLJIT_SUB, SLJIT_R1, 0, SLJIT_R1, 0, SLJIT_IMM, 1);
2295      CMPTO(SLJIT_EQUAL, SLJIT_R2, 0, SLJIT_S2, 0, loop);
2296      OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_R1, 0);
2297      }
2298    else
2299      OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, 1);
2300    }
2301    
2302    static SLJIT_INLINE void return_with_partial_match(compiler_common *common, struct sljit_label *quit)
2303    {
2304    DEFINE_COMPILER;
2305    struct sljit_jump *jump;
2306    
2307    SLJIT_COMPILE_ASSERT(STR_END == SLJIT_S1, str_end_must_be_saved_reg2);
2308    SLJIT_ASSERT(common->start_used_ptr != 0 && common->start_ptr != 0
2309      && (common->mode == JIT_PARTIAL_SOFT_COMPILE ? common->hit_start != 0 : common->hit_start == 0));
2310    
2311    OP1(SLJIT_MOV, SLJIT_R1, 0, ARGUMENTS, 0);
2312    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, PCRE_ERROR_PARTIAL);
2313    OP1(SLJIT_MOV_SI, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_R1), SLJIT_OFFSETOF(jit_arguments, real_offset_count));
2314    CMPTO(SLJIT_SIG_LESS, SLJIT_R2, 0, SLJIT_IMM, 2, quit);
2315    
2316    /* Store match begin and end. */
2317    OP1(SLJIT_MOV, SLJIT_S0, 0, SLJIT_MEM1(SLJIT_R1), SLJIT_OFFSETOF(jit_arguments, begin));
2318    OP1(SLJIT_MOV, SLJIT_R1, 0, SLJIT_MEM1(SLJIT_R1), SLJIT_OFFSETOF(jit_arguments, offsets));
2319    
2320    jump = CMP(SLJIT_SIG_LESS, SLJIT_R2, 0, SLJIT_IMM, 3);
2321    OP2(SLJIT_SUB, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_SP), common->mode == JIT_PARTIAL_HARD_COMPILE ? common->start_ptr : (common->hit_start + (int)sizeof(sljit_sw)), SLJIT_S0, 0);
2322    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2323    OP2(SLJIT_ASHR, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_IMM, UCHAR_SHIFT);
2324    #endif
2325    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_R1), 2 * sizeof(int), SLJIT_R2, 0);
2326    JUMPHERE(jump);
2327    
2328    OP1(SLJIT_MOV, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_SP), common->mode == JIT_PARTIAL_HARD_COMPILE ? common->start_used_ptr : common->hit_start);
2329    OP2(SLJIT_SUB, SLJIT_S1, 0, STR_END, 0, SLJIT_S0, 0);
2330    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2331    OP2(SLJIT_ASHR, SLJIT_S1, 0, SLJIT_S1, 0, SLJIT_IMM, UCHAR_SHIFT);
2332    #endif
2333    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_R1), sizeof(int), SLJIT_S1, 0);
2334    
2335    OP2(SLJIT_SUB, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_S0, 0);
2336    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2337    OP2(SLJIT_ASHR, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_IMM, UCHAR_SHIFT);
2338    #endif
2339    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_R1), 0, SLJIT_R2, 0);
2340    
2341    JUMPTO(SLJIT_JUMP, quit);
2342    }
2343    
2344    static SLJIT_INLINE void check_start_used_ptr(compiler_common *common)
2345    {
2346    /* May destroy TMP1. */
2347    DEFINE_COMPILER;
2348    struct sljit_jump *jump;
2349    
2350    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2351      {
2352      /* The value of -1 must be kept for start_used_ptr! */
2353      OP2(SLJIT_ADD, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, SLJIT_IMM, 1);
2354      /* Jumps if start_used_ptr < STR_PTR, or start_used_ptr == -1. Although overwriting
2355      is not necessary if start_used_ptr == STR_PTR, it does not hurt as well. */
2356      jump = CMP(SLJIT_LESS_EQUAL, TMP1, 0, STR_PTR, 0);
2357      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0);
2358      JUMPHERE(jump);
2359      }
2360    else if (common->mode == JIT_PARTIAL_HARD_COMPILE)
2361      {
2362      jump = CMP(SLJIT_LESS_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0);
2363      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0);
2364      JUMPHERE(jump);
2365      }
2366    }
2367    
2368    static SLJIT_INLINE BOOL char_has_othercase(compiler_common *common, pcre_uchar *cc)
2369    {
2370    /* Detects if the character has an othercase. */
2371    unsigned int c;
2372    
2373    #ifdef SUPPORT_UTF
2374    if (common->utf)
2375      {
2376      GETCHAR(c, cc);
2377      if (c > 127)
2378        {
2379    #ifdef SUPPORT_UCP
2380        return c != UCD_OTHERCASE(c);
2381    #else
2382        return FALSE;
2383    #endif
2384        }
2385    #ifndef COMPILE_PCRE8
2386      return common->fcc[c] != c;
2387    #endif
2388      }
2389    else
2390    #endif
2391      c = *cc;
2392    return MAX_255(c) ? common->fcc[c] != c : FALSE;
2393    }
2394    
2395    static SLJIT_INLINE unsigned int char_othercase(compiler_common *common, unsigned int c)
2396    {
2397    /* Returns with the othercase. */
2398    #ifdef SUPPORT_UTF
2399    if (common->utf && c > 127)
2400      {
2401    #ifdef SUPPORT_UCP
2402      return UCD_OTHERCASE(c);
2403    #else
2404      return c;
2405    #endif
2406      }
2407    #endif
2408    return TABLE_GET(c, common->fcc, c);
2409    }
2410    
2411    static unsigned int char_get_othercase_bit(compiler_common *common, pcre_uchar *cc)
2412    {
2413    /* Detects if the character and its othercase has only 1 bit difference. */
2414    unsigned int c, oc, bit;
2415    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2416    int n;
2417    #endif
2418    
2419    #ifdef SUPPORT_UTF
2420    if (common->utf)
2421      {
2422      GETCHAR(c, cc);
2423      if (c <= 127)
2424        oc = common->fcc[c];
2425      else
2426        {
2427    #ifdef SUPPORT_UCP
2428        oc = UCD_OTHERCASE(c);
2429    #else
2430        oc = c;
2431    #endif
2432        }
2433      }
2434    else
2435      {
2436      c = *cc;
2437      oc = TABLE_GET(c, common->fcc, c);
2438      }
2439    #else
2440    c = *cc;
2441    oc = TABLE_GET(c, common->fcc, c);
2442    #endif
2443    
2444    SLJIT_ASSERT(c != oc);
2445    
2446    bit = c ^ oc;
2447    /* Optimized for English alphabet. */
2448    if (c <= 127 && bit == 0x20)
2449      return (0 << 8) | 0x20;
2450    
2451    /* Since c != oc, they must have at least 1 bit difference. */
2452    if (!is_powerof2(bit))
2453      return 0;
2454    
2455    #if defined COMPILE_PCRE8
2456    
2457    #ifdef SUPPORT_UTF
2458    if (common->utf && c > 127)
2459      {
2460      n = GET_EXTRALEN(*cc);
2461      while ((bit & 0x3f) == 0)
2462        {
2463        n--;
2464        bit >>= 6;
2465        }
2466      return (n << 8) | bit;
2467      }
2468    #endif /* SUPPORT_UTF */
2469    return (0 << 8) | bit;
2470    
2471    #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2472    
2473    #ifdef SUPPORT_UTF
2474    if (common->utf && c > 65535)
2475      {
2476      if (bit >= (1 << 10))
2477        bit >>= 10;
2478      else
2479        return (bit < 256) ? ((2 << 8) | bit) : ((3 << 8) | (bit >> 8));
2480      }
2481    #endif /* SUPPORT_UTF */
2482    return (bit < 256) ? ((0 << 8) | bit) : ((1 << 8) | (bit >> 8));
2483    
2484    #endif /* COMPILE_PCRE[8|16|32] */
2485    }
2486    
2487    static void check_partial(compiler_common *common, BOOL force)
2488    {
2489    /* Checks whether a partial matching is occurred. Does not modify registers. */
2490    DEFINE_COMPILER;
2491    struct sljit_jump *jump = NULL;
2492    
2493    SLJIT_ASSERT(!force || common->mode != JIT_COMPILE);
2494    
2495    if (common->mode == JIT_COMPILE)
2496      return;
2497    
2498    if (!force)
2499      jump = CMP(SLJIT_GREATER_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0);
2500    else if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2501      jump = CMP(SLJIT_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, SLJIT_IMM, -1);
2502    
2503    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2504      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->hit_start, SLJIT_IMM, 0);
2505    else
2506      {
2507      if (common->partialmatchlabel != NULL)
2508        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2509      else
2510        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2511      }
2512    
2513    if (jump != NULL)
2514      JUMPHERE(jump);
2515    }
2516    
2517    static void check_str_end(compiler_common *common, jump_list **end_reached)
2518    {
2519    /* Does not affect registers. Usually used in a tight spot. */
2520    DEFINE_COMPILER;
2521    struct sljit_jump *jump;
2522    
2523    if (common->mode == JIT_COMPILE)
2524      {
2525      add_jump(compiler, end_reached, CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2526      return;
2527      }
2528    
2529    jump = CMP(SLJIT_LESS, STR_PTR, 0, STR_END, 0);
2530    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2531      {
2532      add_jump(compiler, end_reached, CMP(SLJIT_GREATER_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0));
2533      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->hit_start, SLJIT_IMM, 0);
2534      add_jump(compiler, end_reached, JUMP(SLJIT_JUMP));
2535      }
2536    else
2537      {
2538      add_jump(compiler, end_reached, CMP(SLJIT_GREATER_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0));
2539      if (common->partialmatchlabel != NULL)
2540        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2541      else
2542        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2543      }
2544    JUMPHERE(jump);
2545    }
2546    
2547    static void detect_partial_match(compiler_common *common, jump_list **backtracks)
2548    {
2549    DEFINE_COMPILER;
2550    struct sljit_jump *jump;
2551    
2552    if (common->mode == JIT_COMPILE)
2553      {
2554      add_jump(compiler, backtracks, CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2555      return;
2556      }
2557    
2558    /* Partial matching mode. */
2559    jump = CMP(SLJIT_LESS, STR_PTR, 0, STR_END, 0);
2560    add_jump(compiler, backtracks, CMP(SLJIT_GREATER_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0));
2561    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2562      {
2563      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->hit_start, SLJIT_IMM, 0);
2564      add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));
2565      }
2566    else
2567      {
2568      if (common->partialmatchlabel != NULL)
2569        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2570      else
2571        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2572      }
2573    JUMPHERE(jump);
2574    }
2575    
2576    static void peek_char(compiler_common *common, pcre_uint32 max)
2577    {
2578    /* Reads the character into TMP1, keeps STR_PTR.
2579    Does not check STR_END. TMP2 Destroyed. */
2580    DEFINE_COMPILER;
2581    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2582    struct sljit_jump *jump;
2583    #endif
2584    
2585    SLJIT_UNUSED_ARG(max);
2586    
2587    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2588    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2589    if (common->utf)
2590      {
2591      if (max < 128) return;
2592    
2593      jump = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2594      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2595      add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2596      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2597      JUMPHERE(jump);
2598      }
2599    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2600    
2601    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2602    if (common->utf)
2603      {
2604      if (max < 0xd800) return;
2605    
2606      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2607      jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2608      /* TMP2 contains the high surrogate. */
2609      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2610      OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2611      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2612      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2613      OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2614      JUMPHERE(jump);
2615      }
2616    #endif
2617    }
2618    
2619    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2620    
2621    static BOOL is_char7_bitset(const pcre_uint8 *bitset, BOOL nclass)
2622    {
2623    /* Tells whether the character codes below 128 are enough
2624    to determine a match. */
2625    const pcre_uint8 value = nclass ? 0xff : 0;
2626    const pcre_uint8 *end = bitset + 32;
2627    
2628    bitset += 16;
2629    do
2630      {
2631      if (*bitset++ != value)
2632        return FALSE;
2633      }
2634    while (bitset < end);
2635    return TRUE;
2636    }
2637    
2638    static void read_char7_type(compiler_common *common, BOOL full_read)
2639    {
2640    /* Reads the precise character type of a character into TMP1, if the character
2641    is less than 128. Otherwise it returns with zero. Does not check STR_END. The
2642    full_read argument tells whether characters above max are accepted or not. */
2643    DEFINE_COMPILER;
2644    struct sljit_jump *jump;
2645    
2646    SLJIT_ASSERT(common->utf);
2647    
2648    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2649    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2650    
2651    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2652    
2653    if (full_read)
2654      {
2655      jump = CMP(SLJIT_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2656      OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2657      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2658      JUMPHERE(jump);
2659      }
2660    }
2661    
2662    #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2663    
2664    static void read_char_range(compiler_common *common, pcre_uint32 min, pcre_uint32 max, BOOL update_str_ptr)
2665    {
2666    /* Reads the precise value of a character into TMP1, if the character is
2667    between min and max (c >= min && c <= max). Otherwise it returns with a value
2668    outside the range. Does not check STR_END. */
2669    DEFINE_COMPILER;
2670    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2671    struct sljit_jump *jump;
2672    #endif
2673    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2674    struct sljit_jump *jump2;
2675    #endif
2676    
2677    SLJIT_UNUSED_ARG(update_str_ptr);
2678    SLJIT_UNUSED_ARG(min);
2679    SLJIT_UNUSED_ARG(max);
2680    SLJIT_ASSERT(min <= max);
2681    
2682    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2683    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2684    
2685    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2686    if (common->utf)
2687      {
2688      if (max < 128 && !update_str_ptr) return;
2689    
2690      jump = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2691      if (min >= 0x10000)
2692        {
2693        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xf0);
2694        if (update_str_ptr)
2695          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2696        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2697        jump2 = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0x7);
2698        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2699        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2700        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2701        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2702        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2703        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2704        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2705        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2706        if (!update_str_ptr)
2707          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2708        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2709        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2710        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2711        JUMPHERE(jump2);
2712        if (update_str_ptr)
2713          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2714        }
2715      else if (min >= 0x800 && max <= 0xffff)
2716        {
2717        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xe0);
2718        if (update_str_ptr)
2719          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2720        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2721        jump2 = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xf);
2722        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2723        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2724        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2725        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2726        if (!update_str_ptr)
2727          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2728        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2729        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2730        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2731        JUMPHERE(jump2);
2732        if (update_str_ptr)
2733          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2734        }
2735      else if (max >= 0x800)
2736        add_jump(compiler, (max < 0x10000) ? &common->utfreadchar16 : &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2737      else if (max < 128)
2738        {
2739        OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2740        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2741        }
2742      else
2743        {
2744        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2745        if (!update_str_ptr)
2746          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2747        else
2748          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2749        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2750        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2751        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2752        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2753        if (update_str_ptr)
2754          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2755        }
2756      JUMPHERE(jump);
2757      }
2758    #endif
2759    
2760    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2761    if (common->utf)
2762      {
2763      if (max >= 0x10000)
2764        {
2765        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2766        jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2767        /* TMP2 contains the high surrogate. */
2768        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2769        OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2770        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2771        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2772        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2773        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2774        JUMPHERE(jump);
2775        return;
2776        }
2777    
2778      if (max < 0xd800 && !update_str_ptr) return;
2779    
2780      /* Skip low surrogate if necessary. */
2781      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2782      jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2783      if (update_str_ptr)
2784        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2785      if (max >= 0xd800)
2786        OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0x10000);
2787      JUMPHERE(jump);
2788      }
2789    #endif
2790    }
2791    
2792    static SLJIT_INLINE void read_char(compiler_common *common)
2793    {
2794    read_char_range(common, 0, READ_CHAR_MAX, TRUE);
2795    }
2796    
2797    static void read_char8_type(compiler_common *common, BOOL update_str_ptr)
2798    {
2799    /* Reads the character type into TMP1, updates STR_PTR. Does not check STR_END. */
2800    DEFINE_COMPILER;
2801    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2802    struct sljit_jump *jump;
2803    #endif
2804    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2805    struct sljit_jump *jump2;
2806    #endif
2807    
2808    SLJIT_UNUSED_ARG(update_str_ptr);
2809    
2810    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2811    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2812    
2813    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2814    if (common->utf)
2815      {
2816      /* This can be an extra read in some situations, but hopefully
2817      it is needed in most cases. */
2818      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2819      jump = CMP(SLJIT_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2820      if (!update_str_ptr)
2821        {
2822        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2823        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2824        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2825        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2826        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2827        OP2(SLJIT_OR, TMP2, 0, TMP2, 0, TMP1, 0);
2828        OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2829        jump2 = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 255);
2830        OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2831        JUMPHERE(jump2);
2832        }
2833      else
2834        add_jump(compiler, &common->utfreadtype8, JUMP(SLJIT_FAST_CALL));
2835      JUMPHERE(jump);
2836      return;
2837      }
2838    #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2839    
2840    #if !defined COMPILE_PCRE8
2841    /* The ctypes array contains only 256 values. */
2842    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2843    jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 255);
2844    #endif
2845    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2846    #if !defined COMPILE_PCRE8
2847    JUMPHERE(jump);
2848    #endif
2849    
2850    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2851    if (common->utf && update_str_ptr)
2852      {
2853      /* Skip low surrogate if necessary. */
2854      OP2(SLJIT_SUB, TMP2, 0, TMP2, 0, SLJIT_IMM, 0xd800);
2855      jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2856      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2857      JUMPHERE(jump);
2858      }
2859    #endif /* SUPPORT_UTF && COMPILE_PCRE16 */
2860    }
2861    
2862    static void skip_char_back(compiler_common *common)
2863    {
2864    /* Goes one character back. Affects STR_PTR and TMP1. Does not check begin. */
2865    DEFINE_COMPILER;
2866    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2867    #if defined COMPILE_PCRE8
2868    struct sljit_label *label;
2869    
2870    if (common->utf)
2871      {
2872      label = LABEL();
2873      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
2874      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2875      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xc0);
2876      CMPTO(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, 0x80, label);
2877      return;
2878      }
2879    #elif defined COMPILE_PCRE16
2880    if (common->utf)
2881      {
2882      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
2883      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2884      /* Skip low surrogate if necessary. */
2885      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
2886      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xdc00);
2887      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
2888      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
2889      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2890      return;
2891      }
2892    #endif /* COMPILE_PCRE[8|16] */
2893    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2894    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2895    }
2896    
2897    static void check_newlinechar(compiler_common *common, int nltype, jump_list **backtracks, BOOL jumpifmatch)
2898    {
2899    /* Character comes in TMP1. Checks if it is a newline. TMP2 may be destroyed. */
2900    DEFINE_COMPILER;
2901    struct sljit_jump *jump;
2902    
2903    if (nltype == NLTYPE_ANY)
2904      {
2905      add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));
2906      add_jump(compiler, backtracks, JUMP(jumpifmatch ? SLJIT_NOT_ZERO : SLJIT_ZERO));
2907      }
2908    else if (nltype == NLTYPE_ANYCRLF)
2909      {
2910      if (jumpifmatch)
2911        {
2912        add_jump(compiler, backtracks, CMP(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR));
2913        add_jump(compiler, backtracks, CMP(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL));
2914        }
2915      else
2916        {
2917        jump = CMP(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR);
2918        add_jump(compiler, backtracks, CMP(SLJIT_NOT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL));
2919        JUMPHERE(jump);
2920        }
2921      }
2922    else
2923      {
2924      SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline < 256);
2925      add_jump(compiler, backtracks, CMP(jumpifmatch ? SLJIT_EQUAL : SLJIT_NOT_EQUAL, TMP1, 0, SLJIT_IMM, common->newline));
2926      }
2927    }
2928    
2929    #ifdef SUPPORT_UTF
2930    
2931    #if defined COMPILE_PCRE8
2932    static void do_utfreadchar(compiler_common *common)
2933    {
2934    /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2935    of the character (>= 0xc0). Return char value in TMP1, length in TMP2. */
2936    DEFINE_COMPILER;
2937    struct sljit_jump *jump;
2938    
2939    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2940    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2941    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2942    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2943    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2944    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2945    
2946    /* Searching for the first zero. */
2947    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
2948    jump = JUMP(SLJIT_NOT_ZERO);
2949    /* Two byte sequence. */
2950    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2951    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(2));
2952    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2953    
2954    JUMPHERE(jump);
2955    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2956    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
2957    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2958    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2959    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2960    
2961    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2962    jump = JUMP(SLJIT_NOT_ZERO);
2963    /* Three byte sequence. */
2964    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2965    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(3));
2966    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2967    
2968    /* Four byte sequence. */
2969    JUMPHERE(jump);
2970    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2971    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2972    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2973    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2974    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2975    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2976    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(4));
2977    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2978    }
2979    
2980    static void do_utfreadchar16(compiler_common *common)
2981    {
2982    /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2983    of the character (>= 0xc0). Return value in TMP1. */
2984    DEFINE_COMPILER;
2985    struct sljit_jump *jump;
2986    
2987    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2988    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2989    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2990    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2991    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2992    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2993    
2994    /* Searching for the first zero. */
2995    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
2996    jump = JUMP(SLJIT_NOT_ZERO);
2997    /* Two byte sequence. */
2998    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2999    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3000    
3001    JUMPHERE(jump);
3002    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x400);
3003    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_NOT_ZERO);
3004    /* This code runs only in 8 bit mode. No need to shift the value. */
3005    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
3006    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
3007    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
3008    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
3009    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
3010    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
3011    /* Three byte sequence. */
3012    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
3013    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3014    }
3015    
3016    static void do_utfreadtype8(compiler_common *common)
3017    {
3018    /* Fast decoding a UTF-8 character type. TMP2 contains the first byte
3019    of the character (>= 0xc0). Return value in TMP1. */
3020    DEFINE_COMPILER;
3021    struct sljit_jump *jump;
3022    struct sljit_jump *compare;
3023    
3024    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
3025    
3026    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0x20);
3027    jump = JUMP(SLJIT_NOT_ZERO);
3028    /* Two byte sequence. */
3029    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
3030    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3031    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);
3032    /* The upper 5 bits are known at this point. */
3033    compare = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0x3);
3034    OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
3035    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
3036    OP2(SLJIT_OR, TMP2, 0, TMP2, 0, TMP1, 0);
3037    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
3038    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3039    
3040    JUMPHERE(compare);
3041    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
3042    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3043    
3044    /* We only have types for characters less than 256. */
3045    JUMPHERE(jump);
3046    OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
3047    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
3048    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
3049    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3050    }
3051    
3052    #endif /* COMPILE_PCRE8 */
3053    
3054    #endif /* SUPPORT_UTF */
3055    
3056    #ifdef SUPPORT_UCP
3057    
3058    /* UCD_BLOCK_SIZE must be 128 (see the assert below). */
3059    #define UCD_BLOCK_MASK 127
3060    #define UCD_BLOCK_SHIFT 7
3061    
3062    static void do_getucd(compiler_common *common)
3063    {
3064    /* Search the UCD record for the character comes in TMP1.
3065    Returns chartype in TMP1 and UCD offset in TMP2. */
3066    DEFINE_COMPILER;
3067    
3068    SLJIT_ASSERT(UCD_BLOCK_SIZE == 128 && sizeof(ucd_record) == 8);
3069    
3070    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
3071    OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
3072    OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(ucd_stage1));
3073    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);
3074    OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
3075    OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, TMP2, 0);
3076    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_stage2));
3077    OP1(SLJIT_MOV_UH, TMP2, 0, SLJIT_MEM2(TMP2, TMP1), 1);
3078    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_records) + SLJIT_OFFSETOF(ucd_record, chartype));
3079    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM2(TMP1, TMP2), 3);
3080    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3081    }
3082    #endif
3083    
3084    static SLJIT_INLINE struct sljit_label *mainloop_entry(compiler_common *common, BOOL hascrorlf, BOOL firstline)
3085    {
3086    DEFINE_COMPILER;
3087    struct sljit_label *mainloop;
3088    struct sljit_label *newlinelabel = NULL;
3089    struct sljit_jump *start;
3090    struct sljit_jump *end = NULL;
3091    struct sljit_jump *nl = NULL;
3092    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3093    struct sljit_jump *singlechar;
3094    #endif
3095    jump_list *newline = NULL;
3096    BOOL newlinecheck = FALSE;
3097    BOOL readuchar = FALSE;
3098    
3099    if (!(hascrorlf || firstline) && (common->nltype == NLTYPE_ANY ||
3100        common->nltype == NLTYPE_ANYCRLF || common->newline > 255))
3101      newlinecheck = TRUE;
3102    
3103    if (firstline)
3104    {    {
3105    if (tmp1next)    /* Search for the end of the first line. */
3106      SLJIT_ASSERT(common->first_line_end != 0);
3107      OP1(SLJIT_MOV, TMP3, 0, STR_PTR, 0);
3108    
3109      if (common->nltype == NLTYPE_FIXED && common->newline > 255)
3110      {      {
3111      if (!tmp1empty)      mainloop = LABEL();
3112        {      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3113        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);      end = CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3114        stackptr += sizeof(sljit_w);      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(-1));
3115        }      OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
3116      if (!tmp2empty)      CMPTO(SLJIT_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, mainloop);
3117        {      CMPTO(SLJIT_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, mainloop);
3118        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);      JUMPHERE(end);
3119        stackptr += sizeof(sljit_w);      OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_SP), common->first_line_end, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
       }  
3120      }      }
3121    else    else
3122      {      {
3123      if (!tmp2empty)      end = CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3124        {      mainloop = LABEL();
3125        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);      /* Continual stores does not cause data dependency. */
3126        stackptr += sizeof(sljit_w);      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->first_line_end, STR_PTR, 0);
3127        }      read_char_range(common, common->nlmin, common->nlmax, TRUE);
3128      if (!tmp1empty)      check_newlinechar(common, common->nltype, &newline, TRUE);
3129        {      CMPTO(SLJIT_LESS, STR_PTR, 0, STR_END, 0, mainloop);
3130        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);      JUMPHERE(end);
3131        stackptr += sizeof(sljit_w);      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->first_line_end, STR_PTR, 0);
3132        }      set_jumps(newline, LABEL());
3133      }      }
3134    
3135      OP1(SLJIT_MOV, STR_PTR, 0, TMP3, 0);
3136    }    }
 SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));  
 }  
3137    
3138  static SLJIT_INLINE BOOL ispowerof2(unsigned int value)  start = JUMP(SLJIT_JUMP);
 {  
 return (value & (value - 1)) == 0;  
 }  
3139    
3140  static SLJIT_INLINE void set_jumps(jump_list *list, struct sljit_label *label)  if (newlinecheck)
 {  
 while (list)  
3141    {    {
3142    /* sljit_set_label is clever enough to do nothing    newlinelabel = LABEL();
3143    if either the jump or the label is NULL */    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3144    sljit_set_label(list->jump, label);    end = CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3145    list = list->next;    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3146      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, common->newline & 0xff);
3147      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
3148    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
3149      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, UCHAR_SHIFT);
3150    #endif
3151      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3152      nl = JUMP(SLJIT_JUMP);
3153    }    }
 }  
3154    
3155  static SLJIT_INLINE void add_jump(struct sljit_compiler *compiler, jump_list **list, struct sljit_jump* jump)  mainloop = LABEL();
3156  {  
3157  jump_list *list_item = sljit_alloc_memory(compiler, sizeof(jump_list));  /* Increasing the STR_PTR here requires one less jump in the most common case. */
3158  if (list_item)  #ifdef SUPPORT_UTF
3159    if (common->utf) readuchar = TRUE;
3160    #endif
3161    if (newlinecheck) readuchar = TRUE;
3162    
3163    if (readuchar)
3164      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3165    
3166    if (newlinecheck)
3167      CMPTO(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, newlinelabel);
3168    
3169    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3170    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3171    #if defined COMPILE_PCRE8
3172    if (common->utf)
3173    {    {
3174    list_item->next = *list;    singlechar = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
3175    list_item->jump = jump;    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
3176    *list = list_item;    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3177      JUMPHERE(singlechar);
3178      }
3179    #elif defined COMPILE_PCRE16
3180    if (common->utf)
3181      {
3182      singlechar = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xd800);
3183      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
3184      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);
3185      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
3186      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
3187      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3188      JUMPHERE(singlechar);
3189      }
3190    #endif /* COMPILE_PCRE[8|16] */
3191    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
3192    JUMPHERE(start);
3193    
3194    if (newlinecheck)
3195      {
3196      JUMPHERE(end);
3197      JUMPHERE(nl);
3198    }    }
3199    
3200    return mainloop;
3201  }  }
3202    
3203  static void add_stub(compiler_common *common, enum stub_types type, int data, struct sljit_jump *start)  #define MAX_N_CHARS 16
3204    #define MAX_N_BYTES 8
3205    
3206    static SLJIT_INLINE void add_prefix_byte(pcre_uint8 byte, pcre_uint8 *bytes)
3207  {  {
3208  DEFINE_COMPILER;  pcre_uint8 len = bytes[0];
3209  stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));  int i;
3210    
3211  if (list_item)  if (len == 255)
3212      return;
3213    
3214    if (len == 0)
3215    {    {
3216    list_item->type = type;    bytes[0] = 1;
3217    list_item->data = data;    bytes[1] = byte;
3218    list_item->start = start;    return;
   list_item->leave = LABEL();  
   list_item->next = common->stubs;  
   common->stubs = list_item;  
3219    }    }
 }  
3220    
3221  static void flush_stubs(compiler_common *common)  for (i = len; i > 0; i--)
3222  {    if (bytes[i] == byte)
3223  DEFINE_COMPILER;      return;
 stub_list* list_item = common->stubs;  
3224    
3225  while (list_item)  if (len >= MAX_N_BYTES - 1)
3226    {    {
3227    JUMPHERE(list_item->start);    bytes[0] = 255;
3228    switch(list_item->type)    return;
3229      }
3230    
3231    len++;
3232    bytes[len] = byte;
3233    bytes[0] = len;
3234    }
3235    
3236    static int scan_prefix(compiler_common *common, pcre_uchar *cc, pcre_uint32 *chars, pcre_uint8 *bytes, int max_chars, pcre_uint32 *rec_count)
3237    {
3238    /* Recursive function, which scans prefix literals. */
3239    BOOL last, any, caseless;
3240    int len, repeat, len_save, consumed = 0;
3241    pcre_uint32 chr, mask;
3242    pcre_uchar *alternative, *cc_save, *oc;
3243    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3244    pcre_uchar othercase[8];
3245    #elif defined SUPPORT_UTF && defined COMPILE_PCRE16
3246    pcre_uchar othercase[2];
3247    #else
3248    pcre_uchar othercase[1];
3249    #endif
3250    
3251    repeat = 1;
3252    while (TRUE)
3253      {
3254      if (*rec_count == 0)
3255        return 0;
3256      (*rec_count)--;
3257    
3258      last = TRUE;
3259      any = FALSE;
3260      caseless = FALSE;
3261    
3262      switch (*cc)
3263      {      {
3264      case stack_alloc:      case OP_CHARI:
3265      add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));      caseless = TRUE;
3266        case OP_CHAR:
3267        last = FALSE;
3268        cc++;
3269      break;      break;
     }  
   JUMPTO(SLJIT_JUMP, list_item->leave);  
   list_item = list_item->next;  
   }  
 common->stubs = NULL;  
 }  
3270    
3271  static SLJIT_INLINE void decrease_call_count(compiler_common *common)      case OP_SOD:
3272  {      case OP_SOM:
3273  DEFINE_COMPILER;      case OP_SET_SOM:
3274        case OP_NOT_WORD_BOUNDARY:
3275        case OP_WORD_BOUNDARY:
3276        case OP_EODN:
3277        case OP_EOD:
3278        case OP_CIRC:
3279        case OP_CIRCM:
3280        case OP_DOLL:
3281        case OP_DOLLM:
3282        /* Zero width assertions. */
3283        cc++;
3284        continue;
3285    
3286  OP2(SLJIT_SUB | SLJIT_SET_E, CALL_COUNT, 0, CALL_COUNT, 0, SLJIT_IMM, 1);      case OP_ASSERT:
3287  add_jump(compiler, &common->calllimit, JUMP(SLJIT_C_ZERO));      case OP_ASSERT_NOT:
3288  }      case OP_ASSERTBACK:
3289        case OP_ASSERTBACK_NOT:
3290        cc = bracketend(cc);
3291        continue;
3292    
3293  static SLJIT_INLINE void allocate_stack(compiler_common *common, int size)      case OP_PLUSI:
3294  {      case OP_MINPLUSI:
3295  /* May destroy all locals and registers except TMP2. */      case OP_POSPLUSI:
3296  DEFINE_COMPILER;      caseless = TRUE;
3297        case OP_PLUS:
3298        case OP_MINPLUS:
3299        case OP_POSPLUS:
3300        cc++;
3301        break;
3302    
3303  OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_w));      case OP_EXACTI:
3304  #ifdef DESTROY_REGISTERS      caseless = TRUE;
3305  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);      case OP_EXACT:
3306  OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);      repeat = GET2(cc, 1);
3307  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);      last = FALSE;
3308  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);      cc += 1 + IMM2_SIZE;
3309  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);      break;
3310    
3311        case OP_QUERYI:
3312        case OP_MINQUERYI:
3313        case OP_POSQUERYI:
3314        caseless = TRUE;
3315        case OP_QUERY:
3316        case OP_MINQUERY:
3317        case OP_POSQUERY:
3318        len = 1;
3319        cc++;
3320    #ifdef SUPPORT_UTF
3321        if (common->utf && HAS_EXTRALEN(*cc)) len += GET_EXTRALEN(*cc);
3322  #endif  #endif
3323  add_stub(common, stack_alloc, 0, CMP(SLJIT_C_GREATER, STACK_TOP, 0, STACK_LIMIT, 0));      max_chars = scan_prefix(common, cc + len, chars, bytes, max_chars, rec_count);
3324  }      if (max_chars == 0)
3325          return consumed;
3326        last = FALSE;
3327        break;
3328    
3329  static SLJIT_INLINE void free_stack(compiler_common *common, int size)      case OP_KET:
3330  {      cc += 1 + LINK_SIZE;
3331  DEFINE_COMPILER;      continue;
 OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_w));  
 }  
3332    
3333  static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)      case OP_ALT:
3334  {      cc += GET(cc, 1);
3335  DEFINE_COMPILER;      continue;
 struct sljit_label *loop;  
 int i;  
 /* At this point we can freely use all temporary registers. */  
 /* TMP1 returns with begin - 1. */  
 OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG1, 0, SLJIT_MEM1(SLJIT_GENERAL_REG1), SLJIT_OFFSETOF(jit_arguments, begin), SLJIT_IMM, 1);  
 if (length < 8)  
   {  
   for (i = 0; i < length; i++)  
     OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), SLJIT_TEMPORARY_REG1, 0);  
   }  
 else  
   {  
   OP2(SLJIT_ADD, SLJIT_TEMPORARY_REG2, 0, SLJIT_LOCALS_REG, 0, SLJIT_IMM, OVECTOR_START - sizeof(sljit_w));  
   OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, length);  
   loop = LABEL();  
   OP1(SLJIT_MOVU, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), sizeof(sljit_w), SLJIT_TEMPORARY_REG1, 0);  
   OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_TEMPORARY_REG3, 0, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, 1);  
   JUMPTO(SLJIT_C_NOT_ZERO, loop);  
   }  
 }  
3336    
3337  static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)      case OP_ONCE:
3338  {      case OP_ONCE_NC:
3339  DEFINE_COMPILER;      case OP_BRA:
3340  struct sljit_label *loop;      case OP_BRAPOS:
3341  struct sljit_jump *earlyexit;      case OP_CBRA:
3342        case OP_CBRAPOS:
3343        alternative = cc + GET(cc, 1);
3344        while (*alternative == OP_ALT)
3345          {
3346          max_chars = scan_prefix(common, alternative + 1 + LINK_SIZE, chars, bytes, max_chars, rec_count);
3347          if (max_chars == 0)
3348            return consumed;
3349          alternative += GET(alternative, 1);
3350          }
3351    
3352  /* At this point we can freely use all registers. */      if (*cc == OP_CBRA || *cc == OP_CBRAPOS)
3353  OP1(SLJIT_MOV, SLJIT_GENERAL_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));        cc += IMM2_SIZE;
3354  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1), STR_PTR, 0);      cc += 1 + LINK_SIZE;
3355        continue;
3356    
3357  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, ARGUMENTS, 0);      case OP_CLASS:
3358  OP1(SLJIT_MOV_SI, SLJIT_TEMPORARY_REG2, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, offsetcount));  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3359  OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));      if (common->utf && !is_char7_bitset((const pcre_uint8 *)(cc + 1), FALSE)) return consumed;
3360  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, begin));  #endif
3361  OP2(SLJIT_ADD, SLJIT_GENERAL_REG1, 0, SLJIT_LOCALS_REG, 0, SLJIT_IMM, OVECTOR_START);      any = TRUE;
3362  /* Unlikely, but possible */      cc += 1 + 32 / sizeof(pcre_uchar);
3363  earlyexit = CMP(SLJIT_C_EQUAL, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, 0);      break;
 loop = LABEL();  
 OP2(SLJIT_SUB, SLJIT_GENERAL_REG2, 0, SLJIT_MEM1(SLJIT_GENERAL_REG1), 0, SLJIT_TEMPORARY_REG1, 0);  
 OP2(SLJIT_ADD, SLJIT_GENERAL_REG1, 0, SLJIT_GENERAL_REG1, 0, SLJIT_IMM, sizeof(sljit_w));  
 /* Copy the integer value to the output buffer */  
 OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_TEMPORARY_REG3), sizeof(int), SLJIT_GENERAL_REG2, 0);  
 OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_TEMPORARY_REG2, 0, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, 1);  
 JUMPTO(SLJIT_C_NOT_ZERO, loop);  
 JUMPHERE(earlyexit);  
3364    
3365  /* Calculate the return value, which is the maximum ovector value. */      case OP_NCLASS:
3366  if (topbracket > 1)  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3367    {      if (common->utf) return consumed;
3368    OP2(SLJIT_ADD, SLJIT_TEMPORARY_REG1, 0, SLJIT_LOCALS_REG, 0, SLJIT_IMM, OVECTOR_START + topbracket * 2 * sizeof(sljit_w));  #endif
3369    OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, topbracket + 1);      any = TRUE;
3370        cc += 1 + 32 / sizeof(pcre_uchar);
3371        break;
3372    
3373    /* OVECTOR(0) is never equal to SLJIT_GENERAL_REG3. */  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3374    loop = LABEL();      case OP_XCLASS:
3375    OP1(SLJIT_MOVU, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), -(2 * (sljit_w)sizeof(sljit_w)));  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3376    OP2(SLJIT_SUB, SLJIT_TEMPORARY_R