/[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 675 by ph10, Sat Aug 27 10:18:46 2011 UTC revision 1624 by zherczeg, Fri Feb 5 13:47:43 2016 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
59  #define SLJIT_VERBOSE 0  #define SLJIT_VERBOSE 0
60  #define SLJIT_DEBUG 0  #define SLJIT_DEBUG 0
61    
62  #include "sljit/sljitLir.c"  #include "sljit/sljitLir.c"
63    
64  #if defined SLJIT_CONFIG_UNSUPPORTED && SLJIT_CONFIG_UNSUPPORTED  #if defined SLJIT_CONFIG_UNSUPPORTED && SLJIT_CONFIG_UNSUPPORTED
65  #error "Unsupported architecture"  #error Unsupported architecture
66  #endif  #endif
67    
68  /* Allocate memory on the stack. Fast, but limited size. */  /* Defines for debugging purposes. */
 #define LOCAL_SPACE_SIZE 32768  
69    
70    /* 1 - Use unoptimized capturing brackets.
71       2 - Enable capture_last_ptr (includes option 1). */
72    /* #define DEBUG_FORCE_UNOPTIMIZED_CBRAS 2 */
73    
74    /* 1 - Always have a control head. */
75    /* #define DEBUG_FORCE_CONTROL_HEAD 1 */
76    
77    /* Allocate memory for the regex stack on the real machine stack.
78    Fast, but limited size. */
79    #define MACHINE_STACK_SIZE 32768
80    
81    /* Growth rate for stack allocated by the OS. Should be the multiply
82    of page size. */
83  #define STACK_GROWTH_RATE 8192  #define STACK_GROWTH_RATE 8192
84    
85  /* Enable to check that the allocation could destroy temporaries. */  /* Enable to check that the allocation could destroy temporaries. */
# Line 79  The code generator follows the recursive Line 94  The code generator follows the recursive
94  expressions. The basic blocks of regular expressions are condition checkers  expressions. The basic blocks of regular expressions are condition checkers
95  whose execute different commands depending on the result of the condition check.  whose execute different commands depending on the result of the condition check.
96  The relationship between the operators can be horizontal (concatenation) and  The relationship between the operators can be horizontal (concatenation) and
97  vertical (sub-expression) (See struct fallback_common for more details).  vertical (sub-expression) (See struct backtrack_common for more details).
98    
99    'ab' - 'a' and 'b' regexps are concatenated    'ab' - 'a' and 'b' regexps are concatenated
100    'a+' - 'a' is the sub-expression of the '+' operator    'a+' - 'a' is the sub-expression of the '+' operator
101    
102  The condition checkers are boolean (true/false) checkers. Machine code is generated  The condition checkers are boolean (true/false) checkers. Machine code is generated
103  for the checker itself and for the actions depending on the result of the checker.  for the checker itself and for the actions depending on the result of the checker.
104  The 'true' case is called as the hot path (expected path), and the other is called as  The 'true' case is called as the matching path (expected path), and the other is called as
105  the 'fallback' path. Branch instructions are expesive for all CPUs, so we avoid taken  the 'backtrack' path. Branch instructions are expesive for all CPUs, so we avoid taken
106  branches on the hot path.  branches on the matching path.
107    
108   Greedy star operator (*) :   Greedy star operator (*) :
109     Hot path: match happens.     Matching path: match happens.
110     Fallback path: match failed.     Backtrack path: match failed.
111   Non-greedy star operator (*?) :   Non-greedy star operator (*?) :
112     Hot path: no need to perform a match.     Matching path: no need to perform a match.
113     Fallback path: match is required.     Backtrack path: match is required.
114    
115  The following example shows how the code generated for a capturing bracket  The following example shows how the code generated for a capturing bracket
116  with two alternatives. Let A, B, C, D are arbirary regular expressions, and  with two alternatives. Let A, B, C, D are arbirary regular expressions, and
# Line 105  we have the following regular expression Line 120  we have the following regular expression
120    
121  The generated code will be the following:  The generated code will be the following:
122    
123   A hot path   A matching path
124   '(' hot path (pushing arguments to the stack)   '(' matching path (pushing arguments to the stack)
125   B hot path   B matching path
126   ')' hot path (pushing arguments to the stack)   ')' matching path (pushing arguments to the stack)
127   D hot path   D matching path
128   return with successful match   return with successful match
129    
130   D fallback path   D backtrack path
131   ')' fallback path (If we arrived from "C" jump to the fallback of "C")   ')' backtrack path (If we arrived from "C" jump to the backtrack of "C")
132   B fallback path   B backtrack path
133   C expected path   C expected path
134   jump to D hot path   jump to D matching path
135   C fallback path   C backtrack path
136   A fallback path   A backtrack path
137    
138   Notice, that the order of fallback code paths are the opposite of the fast   Notice, that the order of backtrack code paths are the opposite of the fast
139   code paths. In this way the topmost value on the stack is always belong   code paths. In this way the topmost value on the stack is always belong
140   to the current fallback code path. The fallback code path must check   to the current backtrack code path. The backtrack path must check
141   whether there is a next alternative. If so, it needs to jump back to   whether there is a next alternative. If so, it needs to jump back to
142   the hot path eventually. Otherwise it needs to clear out its own stack   the matching path eventually. Otherwise it needs to clear out its own stack
143   frame and continue the execution on the fallback code paths.   frame and continue the execution on the backtrack code paths.
144  */  */
145    
146  /*  /*
147  Saved stack frames:  Saved stack frames:
148    
149  Atomic blocks and asserts require reloading the values of local variables  Atomic blocks and asserts require reloading the values of private data
150  when the fallback mechanism performed. Because of OP_RECURSE, the locals  when the backtrack mechanism performed. Because of OP_RECURSE, the data
151  are not necessarly known in compile time, thus we need a dynamic restore  are not necessarly known in compile time, thus we need a dynamic restore
152  mechanism.  mechanism.
153    
154  The stack frames are stored in a chain list, and have the following format:  The stack frames are stored in a chain list, and have the following format:
155  ([ capturing bracket offset ][ start value ][ end value ])+ ... [ 0 ] [ previous head ]  ([ capturing bracket offset ][ start value ][ end value ])+ ... [ 0 ] [ previous head ]
156    
157  Thus we can restore the locals to a particular point in the stack.  Thus we can restore the private data to a particular point in the stack.
158  */  */
159    
160  typedef struct jit_arguments {  typedef struct jit_arguments {
161    /* Pointers first. */    /* Pointers first. */
162    struct sljit_stack *stack;    struct sljit_stack *stack;
163    PCRE_SPTR str;    const pcre_uchar *str;
164    PCRE_SPTR begin;    const pcre_uchar *begin;
165    PCRE_SPTR end;    const pcre_uchar *end;
166    int *offsets;    int *offsets;
167    uschar *ptr;    pcre_uchar *uchar_ptr;
168      pcre_uchar *mark_ptr;
169      void *callout_data;
170    /* Everything else after. */    /* Everything else after. */
171    int offsetcount;    pcre_uint32 limit_match;
172    uschar notbol;    int real_offset_count;
173    uschar noteol;    int offset_count;
174    uschar notempty;    pcre_uint8 notbol;
175    uschar notempty_atstart;    pcre_uint8 noteol;
176      pcre_uint8 notempty;
177      pcre_uint8 notempty_atstart;
178  } jit_arguments;  } jit_arguments;
179    
180  typedef struct executable_function {  typedef struct executable_functions {
181    void *executable_func;    void *executable_funcs[JIT_NUMBER_OF_COMPILE_MODES];
182    pcre_jit_callback callback;    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, max_index };  
   
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 255  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    struct sljit_label *acceptlabel;    sljit_sw ctypes;
387      /* Named capturing brackets. */
388      pcre_uchar *name_table;
389      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;
408    jump_list *stackalloc;    jump_list *stackalloc;
409    jump_list *revertframes;    jump_list *revertframes;
410    jump_list *wordboundary;    jump_list *wordboundary;
# Line 288  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 308  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_setmaxindex = -1,  
   frame_setstrbegin = -2  
 };  
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 MAX_INDEX     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))
492  /* Head of the saved local variables */  /* Max limit of recursions. */
493  #define LOCALS_HEAD      (4 * sizeof(sljit_w))  #define LIMIT_MATCH      (4 * sizeof(sljit_sw))
 /* Head of the last recursion. */  
 #define RECURSIVE_HEAD   (5 * sizeof(sljit_w))  
 /* Last known position of the requested byte. */  
 #define REQ_BYTE_PTR     (6 * sizeof(sljit_w))  
 /* End pointer of the first line. */  
 #define FIRSTLINE_END    (7 * sizeof(sljit_w))  
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 383  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  static uschar* bracketend(uschar* cc)  #define READ_CHAR_MAX 0x7fffffff
543    
544    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 399  cc += 1 + LINK_SIZE; Line 550  cc += 1 + LINK_SIZE;
550  return cc;  return cc;
551  }  }
552    
553  /* Functions whose might need modification for all new supported opcodes:  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:
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 429  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 441  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];
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_CLOSE:  
   cc += 3;  
   return cc;  
   
   case OP_CRRANGE:  
   case OP_CRMINRANGE:  
   return cc + 5;  
754    
755    case OP_CLASS:    case OP_ANYBYTE:
756    case OP_NCLASS:  #ifdef SUPPORT_UTF
757    return cc + 33;    if (common->utf) return NULL;
758    #endif
759      return cc + 1;
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_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_BRAPOS:  
795      case OP_SBRA:      case OP_REF:
796      case OP_SBRAPOS:      case OP_REFI:
797      case OP_SCOND:      common->optimized_cbracket[GET2(cc, 1)] = 0;
798      localspace += sizeof(sljit_w);      cc += 1 + IMM2_SIZE;
     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_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. */
 uschar *end;  
 int length = 0;  
 BOOL possessive = FALSE;  
 BOOL needs_frame = FALSE;  
 BOOL needs_maxindex = FALSE;  
 BOOL setsom_found = FALSE;  
   
 if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))  
   {  
   length = 3 + 2;  
   needs_maxindex = TRUE;  
   possessive = TRUE;  
   }  
857    
858  cc = next_opcode(common, cc);      case OP_PRUNE_ARG:
859  SLJIT_ASSERT(cc != NULL);      common->needs_start_ptr = TRUE;
860  while (cc < ccend)      /* Fall through. */
861    switch(*cc)  
862      {      case OP_MARK:
863      case OP_SET_SOM:      if (common->mark_ptr == 0)
     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_ASSERT:      case OP_THEN:
872      case OP_ASSERT_NOT:      common->has_then = TRUE;
873      case OP_ASSERTBACK:      common->control_head_ptr = 1;
874      case OP_ASSERTBACK_NOT:      /* Fall through. */
875      case OP_ONCE:  
876      if (needs_frame || length > 0)      case OP_PRUNE:
877        {      case OP_SKIP:
878        cc = bracketend(cc);      common->needs_start_ptr = TRUE;
879        break;      cc += 1;
       }  
     /* Check whether a frame must be created. */  
     end = bracketend(cc);  
     while (cc < end)  
       {  
       if (*cc == OP_SET_SOM || *cc == OP_CBRA || *cc == OP_CBRAPOS  
           || *cc == OP_SCBRA || *cc == OP_SCBRAPOS || *cc == OP_RECURSE)  
         needs_frame = TRUE;  
       cc = next_opcode(common, cc);  
       SLJIT_ASSERT(cc != NULL);  
       }  
880      break;      break;
881    
882      case OP_CBRA:      case OP_SKIP_ARG:
883      case OP_CBRAPOS:      common->control_head_ptr = 1;
884      case OP_SCBRA:      common->has_skip_arg = TRUE;
885      case OP_SCBRAPOS:      cc += 1 + 2 + cc[1];
     if (!needs_maxindex)  
       {  
       needs_maxindex = TRUE;  
       length += 2;  
       }  
     length += 3;  
     cc += 1 + LINK_SIZE + 2;  
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) && !needs_frame && length == 3 + 2)  
   return -1;  
   
 if (length > 0)  
   return length + 2;  
 return needs_frame ? 0 : -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  /* TMP2 must contain STACK_TOP - (-STACK(stackpos)) */  switch(*cc)
901  DEFINE_COMPILER;    {
902  uschar *ccend = bracketend(cc);    case OP_CRSTAR:
903  BOOL needs_maxindex = FALSE;    case OP_CRPLUS:
904  BOOL setsom_found = FALSE;    return 2;
905  int offset;  
906      case OP_CRMINSTAR:
907      case OP_CRMINPLUS:
908      case OP_CRQUERY:
909      case OP_CRMINQUERY:
910      return 1;
911    
912      case OP_CRRANGE:
913      case OP_CRMINRANGE:
914      if (GET2(cc, 1) == GET2(cc, 1 + IMM2_SIZE))
915        return 0;
916      return 2;
917    
918      default:
919      return 0;
920      }
921    }
922    
923  if (stackpos < stacktop)  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    SLJIT_ASSERT(stackpos + 1 == stacktop);    if (*next != *begin)
946    return;      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  stackpos = STACK(stackpos);  if (min == 2)
955  OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS_HEAD);    return FALSE;
 OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS_HEAD, TMP2, 0);  
 OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), STACK(stacktop), TMP1, 0);  
956    
957  if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))  max = 0;
958    cc = next_opcode(common, cc);  max_end = next;
959  SLJIT_ASSERT(cc != NULL);  if (*next == OP_BRAZERO || *next == OP_BRAMINZERO)
960  while (cc < ccend)    {
961    switch(*cc)    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_ASSERT:  if (min >= 3)
998      case OP_ASSERT_NOT:    {
999      case OP_ASSERTBACK:    common->private_data_ptrs[end - common->start - LINK_SIZE] = max_end - end;
1000      case OP_ASSERTBACK_NOT:    common->private_data_ptrs[end - common->start - LINK_SIZE + 1] = OP_EXACT;
1001      case OP_ONCE:    common->private_data_ptrs[end - common->start - LINK_SIZE + 2] = min;
1002      cc = bracketend(cc);    return TRUE;
1003      break;    }
1004    
1005    return FALSE;
1006    }
1007    
1008    #define CASE_ITERATOR_PRIVATE_DATA_1 \
1009        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      case OP_CBRA:  #define CASE_ITERATOR_PRIVATE_DATA_2A \
1027      case OP_CBRAPOS:      case OP_STAR: \
1028      case OP_SCBRA:      case OP_PLUS: \
1029      case OP_SCBRAPOS:      case OP_STARI: \
1030      if (!needs_maxindex)      case OP_PLUSI: \
1031        {      case OP_NOTSTAR: \
1032        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setmaxindex);      case OP_NOTPLUS: \
1033        stackpos += (int)sizeof(sljit_w);      case OP_NOTSTARI: \
1034        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, MAX_INDEX, 0);      case OP_NOTPLUSI:
       stackpos += (int)sizeof(sljit_w);  
       needs_maxindex = TRUE;  
       }  
     offset = (GET2(cc, 1 + LINK_SIZE)) << 1;  
     OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));  
     stackpos += (int)sizeof(sljit_w);  
     OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset));  
     OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset + 1));  
     OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);  
     stackpos += (int)sizeof(sljit_w);  
     OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);  
     stackpos += (int)sizeof(sljit_w);  
1035    
1036      cc += 1 + LINK_SIZE + 2;  #define CASE_ITERATOR_PRIVATE_DATA_2B \
1037      break;      case OP_UPTO: \
1038        case OP_MINUPTO: \
1039        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      default:  #define CASE_ITERATOR_TYPE_PRIVATE_DATA_1 \
1047      cc = next_opcode(common, cc);      case OP_TYPEMINSTAR: \
1048      SLJIT_ASSERT(cc != NULL);      case OP_TYPEMINPLUS: \
1049      break;      case OP_TYPEQUERY: \
1050      }      case OP_TYPEMINQUERY:
1051    
1052  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_end);  #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2A \
1053  SLJIT_ASSERT(stackpos == STACK(stacktop + 1));      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:
1104      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1105      case OP_ONCE:      case OP_ONCE:
1106        case OP_ONCE_NC:
1107      case OP_BRAPOS:      case OP_BRAPOS:
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;      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;
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))
1263      {      {
1264      OP1(SLJIT_MOV, TMP1, 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      tmp1empty = FALSE;      capture_last_found = TRUE;
     }  
   if (stackptr < stacktop)  
     {  
     OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);  
     stackptr += sizeof(sljit_w);  
     tmp2empty = FALSE;  
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_BRAPOS:        length += 2;
1311        case OP_SBRA:        setmark_found = TRUE;
1312        case OP_SBRAPOS:        }
1313        case OP_SCOND:      if (common->capture_last_ptr != 0 && !capture_last_found)
1314        count = 1;        {
1315        srcw[0] = PRIV(cc);        length += 2;
1316        SLJIT_ASSERT(srcw[0] != 0);        capture_last_found = TRUE;
1317        cc += 1 + LINK_SIZE;        }
1318        break;      cc += 1 + LINK_SIZE;
1319        break;
1320    
1321        case OP_CBRA:      case OP_CBRA:
1322        case OP_SCBRA:      case OP_CBRAPOS:
1323        count = 1;      case OP_SCBRA:
1324        srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));      case OP_SCBRAPOS:
1325        cc += 1 + LINK_SIZE + 2;      stack_restore = TRUE;
1326        break;      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_CBRAPOS:      case OP_THEN:
1336        case OP_SCBRAPOS:      stack_restore = TRUE;
1337        count = 2;      if (common->control_head_ptr != 0)
1338        srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));        *needs_control_head = TRUE;
1339        srcw[0] = PRIV(cc);      cc ++;
1340        SLJIT_ASSERT(srcw[0] != 0);      break;
       cc += 1 + LINK_SIZE + 2;  
       break;  
1341    
1342        case OP_COND:      default:
1343        /* Might be a hidden SCOND. */      stack_restore = TRUE;
1344        alternative = cc + GET(cc, 1);      /* Fall through. */
       if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)  
         {  
         count = 1;  
         srcw[0] = PRIV(cc);  
         SLJIT_ASSERT(srcw[0] != 0);  
         }  
       cc += 1 + LINK_SIZE;  
       break;  
1345    
1346        default:      case OP_NOT_WORD_BOUNDARY:
1347        cc = next_opcode(common, cc);      case OP_WORD_BOUNDARY:
1348        SLJIT_ASSERT(cc != NULL);      case OP_NOT_DIGIT:
1349        break;      case OP_DIGIT:
1350        }      case OP_NOT_WHITESPACE:
1351      break;      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 end:      case OP_EXACT:
1377      SLJIT_ASSERT_STOP();      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        case OP_CALLOUT:
1410    
1411        cc = next_opcode(common, cc);
1412        SLJIT_ASSERT(cc != NULL);
1413      break;      break;
1414      }      }
1415    
1416    while (count > 0)  /* Possessive quantifiers can use a special case. */
1417    if (SLJIT_UNLIKELY(possessive == length))
1418      return stack_restore ? no_frame : no_stack;
1419    
1420    if (length > 0)
1421      return length + 1;
1422    return stack_restore ? no_frame : no_stack;
1423    }
1424    
1425    static void init_frame(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, int stackpos, int stacktop, BOOL recursive)
1426    {
1427    DEFINE_COMPILER;
1428    BOOL setsom_found = recursive;
1429    BOOL setmark_found = recursive;
1430    /* The last capture is a local variable even for recursions. */
1431    BOOL capture_last_found = FALSE;
1432    int offset;
1433    
1434    /* >= 1 + shortest item size (2) */
1435    SLJIT_UNUSED_ARG(stacktop);
1436    SLJIT_ASSERT(stackpos >= stacktop + 2);
1437    
1438    stackpos = STACK(stackpos);
1439    if (ccend == NULL)
1440      {
1441      ccend = bracketend(cc) - (1 + LINK_SIZE);
1442      if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))
1443        cc = next_opcode(common, cc);
1444      }
1445    
1446    SLJIT_ASSERT(cc != NULL);
1447    while (cc < ccend)
1448      switch(*cc)
1449      {      {
1450      count--;      case OP_SET_SOM:
1451      if (save)      SLJIT_ASSERT(common->has_set_som);
1452        if (!setsom_found)
1453        {        {
1454        if (tmp1next)        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(0));
1455          {        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1456          if (!tmp1empty)        stackpos += (int)sizeof(sljit_sw);
1457            {        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1458            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        stackpos += (int)sizeof(sljit_sw);
1459            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;  
         }  
1460        }        }
1461      else      cc += 1;
1462        {      break;
1463        if (tmp1next)  
1464          {      case OP_MARK:
1465          SLJIT_ASSERT(!tmp1empty);      case OP_PRUNE_ARG:
1466          OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count], TMP1, 0);      case OP_THEN_ARG:
1467          tmp1empty = stackptr >= stacktop;      SLJIT_ASSERT(common->mark_ptr != 0);
1468          if (!tmp1empty)      if (!setmark_found)
1469            {        {
1470            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->mark_ptr);
1471            stackptr += sizeof(sljit_w);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
1472            }        stackpos += (int)sizeof(sljit_sw);
1473          tmp1next = FALSE;        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1474          }        stackpos += (int)sizeof(sljit_sw);
1475        else        setmark_found = TRUE;
         {  
         SLJIT_ASSERT(!tmp2empty);  
         OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count], TMP2, 0);  
         tmp2empty = stackptr >= stacktop;  
         if (!tmp2empty)  
           {  
           OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);  
           stackptr += sizeof(sljit_w);  
           }  
         tmp1next = TRUE;  
         }  
1476        }        }
1477      }      cc += 1 + 2 + cc[1];
1478    }      break;
1479    
1480  if (save)      case OP_RECURSE:
1481    {      if (common->has_set_som && !setsom_found)
   if (tmp1next)  
     {  
     if (!tmp1empty)  
1482        {        {
1483        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(0));
1484        stackptr += sizeof(sljit_w);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1485          stackpos += (int)sizeof(sljit_sw);
1486          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1487          stackpos += (int)sizeof(sljit_sw);
1488          setsom_found = TRUE;
1489        }        }
1490      if (!tmp2empty)      if (common->mark_ptr != 0 && !setmark_found)
1491        {        {
1492        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->mark_ptr);
1493        stackptr += sizeof(sljit_w);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
1494          stackpos += (int)sizeof(sljit_sw);
1495          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1496          stackpos += (int)sizeof(sljit_sw);
1497          setmark_found = TRUE;
1498        }        }
1499      }      if (common->capture_last_ptr != 0 && !capture_last_found)
   else  
     {  
     if (!tmp2empty)  
1500        {        {
1501        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->capture_last_ptr);
1502        stackptr += sizeof(sljit_w);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1503          stackpos += (int)sizeof(sljit_sw);
1504          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1505          stackpos += (int)sizeof(sljit_sw);
1506          capture_last_found = TRUE;
1507        }        }
1508      if (!tmp1empty)      cc += 1 + LINK_SIZE;
1509        break;
1510    
1511        case OP_CBRA:
1512        case OP_CBRAPOS:
1513        case OP_SCBRA:
1514        case OP_SCBRAPOS:
1515        if (common->capture_last_ptr != 0 && !capture_last_found)
1516        {        {
1517        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->capture_last_ptr);
1518        stackptr += sizeof(sljit_w);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1519          stackpos += (int)sizeof(sljit_sw);
1520          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1521          stackpos += (int)sizeof(sljit_sw);
1522          capture_last_found = TRUE;
1523        }        }
1524        offset = (GET2(cc, 1 + LINK_SIZE)) << 1;
1525        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));
1526        stackpos += (int)sizeof(sljit_sw);
1527        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(offset));
1528        OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(offset + 1));
1529        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1530        stackpos += (int)sizeof(sljit_sw);
1531        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);
1532        stackpos += (int)sizeof(sljit_sw);
1533    
1534        cc += 1 + LINK_SIZE + IMM2_SIZE;
1535        break;
1536    
1537        default:
1538        cc = next_opcode(common, cc);
1539        SLJIT_ASSERT(cc != NULL);
1540        break;
1541      }      }
   }  
 SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));  
 }  
1542    
1543  static SLJIT_INLINE BOOL ispowerof2(unsigned int value)  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, 0);
1544  {  SLJIT_ASSERT(stackpos == STACK(stacktop));
 return (value & (value - 1)) == 0;  
1545  }  }
1546    
1547  static SLJIT_INLINE void set_jumps(jump_list *list, struct sljit_label *label)  static SLJIT_INLINE int get_private_data_copy_length(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, BOOL needs_control_head)
1548  {  {
1549  while (list)  int private_data_length = needs_control_head ? 3 : 2;
1550    int size;
1551    pcre_uchar *alternative;
1552    /* Calculate the sum of the private machine words. */
1553    while (cc < ccend)
1554    {    {
1555    /* sljit_set_label is clever enough to do nothing    size = 0;
1556    if either the jump or the label is NULL */    switch(*cc)
1557    sljit_set_label(list->jump, label);      {
1558    list = list->next;      case OP_KET:
1559    }      if (PRIVATE_DATA(cc) != 0)
1560  }        {
1561          private_data_length++;
1562          SLJIT_ASSERT(PRIVATE_DATA(cc + 1) != 0);
1563          cc += PRIVATE_DATA(cc + 1);
1564          }
1565        cc += 1 + LINK_SIZE;
1566        break;
1567    
1568  static SLJIT_INLINE void add_jump(struct sljit_compiler *compiler, jump_list **list, struct sljit_jump* jump)      case OP_ASSERT:
1569  {      case OP_ASSERT_NOT:
1570  jump_list *list_item = sljit_alloc_memory(compiler, sizeof(jump_list));      case OP_ASSERTBACK:
1571  if (list_item)      case OP_ASSERTBACK_NOT:
1572    {      case OP_ONCE:
1573    list_item->next = *list;      case OP_ONCE_NC:
1574    list_item->jump = jump;      case OP_BRAPOS:
1575    *list = list_item;      case OP_SBRA:
1576    }      case OP_SBRAPOS:
1577  }      case OP_SCOND:
1578        private_data_length++;
1579        SLJIT_ASSERT(PRIVATE_DATA(cc) != 0);
1580        cc += 1 + LINK_SIZE;
1581        break;
1582    
1583  static void add_stub(compiler_common *common, enum stub_types type, int data, struct sljit_jump *start)      case OP_CBRA:
1584  {      case OP_SCBRA:
1585  DEFINE_COMPILER;      if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1586  stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));        private_data_length++;
1587        cc += 1 + LINK_SIZE + IMM2_SIZE;
1588        break;
1589    
1590  if (list_item)      case OP_CBRAPOS:
1591        case OP_SCBRAPOS:
1592        private_data_length += 2;
1593        cc += 1 + LINK_SIZE + IMM2_SIZE;
1594        break;
1595    
1596        case OP_COND:
1597        /* Might be a hidden SCOND. */
1598        alternative = cc + GET(cc, 1);
1599        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1600          private_data_length++;
1601        cc += 1 + LINK_SIZE;
1602        break;
1603    
1604        CASE_ITERATOR_PRIVATE_DATA_1
1605        if (PRIVATE_DATA(cc))
1606          private_data_length++;
1607        cc += 2;
1608    #ifdef SUPPORT_UTF
1609        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1610    #endif
1611        break;
1612    
1613        CASE_ITERATOR_PRIVATE_DATA_2A
1614        if (PRIVATE_DATA(cc))
1615          private_data_length += 2;
1616        cc += 2;
1617    #ifdef SUPPORT_UTF
1618        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1619    #endif
1620        break;
1621    
1622        CASE_ITERATOR_PRIVATE_DATA_2B
1623        if (PRIVATE_DATA(cc))
1624          private_data_length += 2;
1625        cc += 2 + IMM2_SIZE;
1626    #ifdef SUPPORT_UTF
1627        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1628    #endif
1629        break;
1630    
1631        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1632        if (PRIVATE_DATA(cc))
1633          private_data_length++;
1634        cc += 1;
1635        break;
1636    
1637        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1638        if (PRIVATE_DATA(cc))
1639          private_data_length += 2;
1640        cc += 1;
1641        break;
1642    
1643        CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1644        if (PRIVATE_DATA(cc))
1645          private_data_length += 2;
1646        cc += 1 + IMM2_SIZE;
1647        break;
1648    
1649        case OP_CLASS:
1650        case OP_NCLASS:
1651    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1652        case OP_XCLASS:
1653        size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
1654    #else
1655        size = 1 + 32 / (int)sizeof(pcre_uchar);
1656    #endif
1657        if (PRIVATE_DATA(cc))
1658          private_data_length += get_class_iterator_size(cc + size);
1659        cc += size;
1660        break;
1661    
1662        default:
1663        cc = next_opcode(common, cc);
1664        SLJIT_ASSERT(cc != NULL);
1665        break;
1666        }
1667      }
1668    SLJIT_ASSERT(cc == ccend);
1669    return private_data_length;
1670    }
1671    
1672    static void copy_private_data(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend,
1673      BOOL save, int stackptr, int stacktop, BOOL needs_control_head)
1674    {
1675    DEFINE_COMPILER;
1676    int srcw[2];
1677    int count, size;
1678    BOOL tmp1next = TRUE;
1679    BOOL tmp1empty = TRUE;
1680    BOOL tmp2empty = TRUE;
1681    pcre_uchar *alternative;
1682    enum {
1683      start,
1684      loop,
1685      end
1686    } status;
1687    
1688    status = save ? start : loop;
1689    stackptr = STACK(stackptr - 2);
1690    stacktop = STACK(stacktop - 1);
1691    
1692    if (!save)
1693      {
1694      stackptr += (needs_control_head ? 2 : 1) * sizeof(sljit_sw);
1695      if (stackptr < stacktop)
1696        {
1697        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1698        stackptr += sizeof(sljit_sw);
1699        tmp1empty = FALSE;
1700        }
1701      if (stackptr < stacktop)
1702        {
1703        OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1704        stackptr += sizeof(sljit_sw);
1705        tmp2empty = FALSE;
1706        }
1707      /* The tmp1next must be TRUE in either way. */
1708      }
1709    
1710    do
1711      {
1712      count = 0;
1713      switch(status)
1714        {
1715        case start:
1716        SLJIT_ASSERT(save && common->recursive_head_ptr != 0);
1717        count = 1;
1718        srcw[0] = common->recursive_head_ptr;
1719        if (needs_control_head)
1720          {
1721          SLJIT_ASSERT(common->control_head_ptr != 0);
1722          count = 2;
1723          srcw[1] = common->control_head_ptr;
1724          }
1725        status = loop;
1726        break;
1727    
1728        case loop:
1729        if (cc >= ccend)
1730          {
1731          status = end;
1732          break;
1733          }
1734    
1735        switch(*cc)
1736          {
1737          case OP_KET:
1738          if (PRIVATE_DATA(cc) != 0)
1739            {
1740            count = 1;
1741            srcw[0] = PRIVATE_DATA(cc);
1742            SLJIT_ASSERT(PRIVATE_DATA(cc + 1) != 0);
1743            cc += PRIVATE_DATA(cc + 1);
1744            }
1745          cc += 1 + LINK_SIZE;
1746          break;
1747    
1748          case OP_ASSERT:
1749          case OP_ASSERT_NOT:
1750          case OP_ASSERTBACK:
1751          case OP_ASSERTBACK_NOT:
1752          case OP_ONCE:
1753          case OP_ONCE_NC:
1754          case OP_BRAPOS:
1755          case OP_SBRA:
1756          case OP_SBRAPOS:
1757          case OP_SCOND:
1758          count = 1;
1759          srcw[0] = PRIVATE_DATA(cc);
1760          SLJIT_ASSERT(srcw[0] != 0);
1761          cc += 1 + LINK_SIZE;
1762          break;
1763    
1764          case OP_CBRA:
1765          case OP_SCBRA:
1766          if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1767            {
1768            count = 1;
1769            srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1770            }
1771          cc += 1 + LINK_SIZE + IMM2_SIZE;
1772          break;
1773    
1774          case OP_CBRAPOS:
1775          case OP_SCBRAPOS:
1776          count = 2;
1777          srcw[0] = PRIVATE_DATA(cc);
1778          srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1779          SLJIT_ASSERT(srcw[0] != 0 && srcw[1] != 0);
1780          cc += 1 + LINK_SIZE + IMM2_SIZE;
1781          break;
1782    
1783          case OP_COND:
1784          /* Might be a hidden SCOND. */
1785          alternative = cc + GET(cc, 1);
1786          if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1787            {
1788            count = 1;
1789            srcw[0] = PRIVATE_DATA(cc);
1790            SLJIT_ASSERT(srcw[0] != 0);
1791            }
1792          cc += 1 + LINK_SIZE;
1793          break;
1794    
1795          CASE_ITERATOR_PRIVATE_DATA_1
1796          if (PRIVATE_DATA(cc))
1797            {
1798            count = 1;
1799            srcw[0] = PRIVATE_DATA(cc);
1800            }
1801          cc += 2;
1802    #ifdef SUPPORT_UTF
1803          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1804    #endif
1805          break;
1806    
1807          CASE_ITERATOR_PRIVATE_DATA_2A
1808          if (PRIVATE_DATA(cc))
1809            {
1810            count = 2;
1811            srcw[0] = PRIVATE_DATA(cc);
1812            srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1813            }
1814          cc += 2;
1815    #ifdef SUPPORT_UTF
1816          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1817    #endif
1818          break;
1819    
1820          CASE_ITERATOR_PRIVATE_DATA_2B
1821          if (PRIVATE_DATA(cc))
1822            {
1823            count = 2;
1824            srcw[0] = PRIVATE_DATA(cc);
1825            srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1826            }
1827          cc += 2 + IMM2_SIZE;
1828    #ifdef SUPPORT_UTF
1829          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1830    #endif
1831          break;
1832    
1833          CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1834          if (PRIVATE_DATA(cc))
1835            {
1836            count = 1;
1837            srcw[0] = PRIVATE_DATA(cc);
1838            }
1839          cc += 1;
1840          break;
1841    
1842          CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1843          if (PRIVATE_DATA(cc))
1844            {
1845            count = 2;
1846            srcw[0] = PRIVATE_DATA(cc);
1847            srcw[1] = srcw[0] + sizeof(sljit_sw);
1848            }
1849          cc += 1;
1850          break;
1851    
1852          CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1853          if (PRIVATE_DATA(cc))
1854            {
1855            count = 2;
1856            srcw[0] = PRIVATE_DATA(cc);
1857            srcw[1] = srcw[0] + sizeof(sljit_sw);
1858            }
1859          cc += 1 + IMM2_SIZE;
1860          break;
1861    
1862          case OP_CLASS:
1863          case OP_NCLASS:
1864    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1865          case OP_XCLASS:
1866          size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
1867    #else
1868          size = 1 + 32 / (int)sizeof(pcre_uchar);
1869    #endif
1870          if (PRIVATE_DATA(cc))
1871            switch(get_class_iterator_size(cc + size))
1872              {
1873              case 1:
1874              count = 1;
1875              srcw[0] = PRIVATE_DATA(cc);
1876              break;
1877    
1878              case 2:
1879              count = 2;
1880              srcw[0] = PRIVATE_DATA(cc);
1881              srcw[1] = srcw[0] + sizeof(sljit_sw);
1882              break;
1883    
1884              default:
1885              SLJIT_ASSERT_STOP();
1886              break;
1887              }
1888          cc += size;
1889          break;
1890    
1891          default:
1892          cc = next_opcode(common, cc);
1893          SLJIT_ASSERT(cc != NULL);
1894          break;
1895          }
1896        break;
1897    
1898        case end:
1899        SLJIT_ASSERT_STOP();
1900        break;
1901        }
1902    
1903      while (count > 0)
1904        {
1905        count--;
1906        if (save)
1907          {
1908          if (tmp1next)
1909            {
1910            if (!tmp1empty)
1911              {
1912              OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1913              stackptr += sizeof(sljit_sw);
1914              }
1915            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), srcw[count]);
1916            tmp1empty = FALSE;
1917            tmp1next = FALSE;
1918            }
1919          else
1920            {
1921            if (!tmp2empty)
1922              {
1923              OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1924              stackptr += sizeof(sljit_sw);
1925              }
1926            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_SP), srcw[count]);
1927            tmp2empty = FALSE;
1928            tmp1next = TRUE;
1929            }
1930          }
1931        else
1932          {
1933          if (tmp1next)
1934            {
1935            SLJIT_ASSERT(!tmp1empty);
1936            OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), srcw[count], TMP1, 0);
1937            tmp1empty = stackptr >= stacktop;
1938            if (!tmp1empty)
1939              {
1940              OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1941              stackptr += sizeof(sljit_sw);
1942              }
1943            tmp1next = FALSE;
1944            }
1945          else
1946            {
1947            SLJIT_ASSERT(!tmp2empty);
1948            OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), srcw[count], TMP2, 0);
1949            tmp2empty = stackptr >= stacktop;
1950            if (!tmp2empty)
1951              {
1952              OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1953              stackptr += sizeof(sljit_sw);
1954              }
1955            tmp1next = TRUE;
1956            }
1957          }
1958        }
1959      }
1960    while (status != end);
1961    
1962    if (save)
1963      {
1964      if (tmp1next)
1965        {
1966        if (!tmp1empty)
1967          {
1968          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1969          stackptr += sizeof(sljit_sw);
1970          }
1971        if (!tmp2empty)
1972          {
1973          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1974          stackptr += sizeof(sljit_sw);
1975          }
1976        }
1977      else
1978        {
1979        if (!tmp2empty)
1980          {
1981          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1982          stackptr += sizeof(sljit_sw);
1983          }
1984        if (!tmp1empty)
1985          {
1986          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1987          stackptr += sizeof(sljit_sw);
1988          }
1989        }
1990      }
1991    SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));
1992    }
1993    
1994    static SLJIT_INLINE pcre_uchar *set_then_offsets(compiler_common *common, pcre_uchar *cc, pcre_uint8 *current_offset)
1995    {
1996    pcre_uchar *end = bracketend(cc);
1997    BOOL has_alternatives = cc[GET(cc, 1)] == OP_ALT;
1998    
1999    /* Assert captures then. */
2000    if (*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT)
2001      current_offset = NULL;
2002    /* Conditional block does not. */
2003    if (*cc == OP_COND || *cc == OP_SCOND)
2004      has_alternatives = FALSE;
2005    
2006    cc = next_opcode(common, cc);
2007    if (has_alternatives)
2008      current_offset = common->then_offsets + (cc - common->start);
2009    
2010    while (cc < end)
2011      {
2012      if ((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND))
2013        cc = set_then_offsets(common, cc, current_offset);
2014      else
2015        {
2016        if (*cc == OP_ALT && has_alternatives)
2017          current_offset = common->then_offsets + (cc + 1 + LINK_SIZE - common->start);
2018        if (*cc >= OP_THEN && *cc <= OP_THEN_ARG && current_offset != NULL)
2019          *current_offset = 1;
2020        cc = next_opcode(common, cc);
2021        }
2022      }
2023    
2024    return end;
2025    }
2026    
2027    #undef CASE_ITERATOR_PRIVATE_DATA_1
2028    #undef CASE_ITERATOR_PRIVATE_DATA_2A
2029    #undef CASE_ITERATOR_PRIVATE_DATA_2B
2030    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_1
2031    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
2032    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
2033    
2034    static SLJIT_INLINE BOOL is_powerof2(unsigned int value)
2035    {
2036    return (value & (value - 1)) == 0;
2037    }
2038    
2039    static SLJIT_INLINE void set_jumps(jump_list *list, struct sljit_label *label)
2040    {
2041    while (list)
2042      {
2043      /* sljit_set_label is clever enough to do nothing
2044      if either the jump or the label is NULL. */
2045      SET_LABEL(list->jump, label);
2046      list = list->next;
2047      }
2048    }
2049    
2050    static SLJIT_INLINE void add_jump(struct sljit_compiler *compiler, jump_list **list, struct sljit_jump *jump)
2051    {
2052    jump_list *list_item = sljit_alloc_memory(compiler, sizeof(jump_list));
2053    if (list_item)
2054      {
2055      list_item->next = *list;
2056      list_item->jump = jump;
2057      *list = list_item;
2058      }
2059    }
2060    
2061    static void add_stub(compiler_common *common, struct sljit_jump *start)
2062    {
2063    DEFINE_COMPILER;
2064    stub_list *list_item = sljit_alloc_memory(compiler, sizeof(stub_list));
2065    
2066    if (list_item)
2067      {
2068      list_item->start = start;
2069      list_item->quit = LABEL();
2070      list_item->next = common->stubs;
2071      common->stubs = list_item;
2072      }
2073    }
2074    
2075    static void flush_stubs(compiler_common *common)
2076    {
2077    DEFINE_COMPILER;
2078    stub_list *list_item = common->stubs;
2079    
2080    while (list_item)
2081      {
2082      JUMPHERE(list_item->start);
2083      add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));
2084      JUMPTO(SLJIT_JUMP, list_item->quit);
2085      list_item = list_item->next;
2086      }
2087    common->stubs = NULL;
2088    }
2089    
2090    static void add_label_addr(compiler_common *common, sljit_uw *update_addr)
2091    {
2092    DEFINE_COMPILER;
2093    label_addr_list *label_addr;
2094    
2095    label_addr = sljit_alloc_memory(compiler, sizeof(label_addr_list));
2096    if (label_addr == NULL)
2097      return;
2098    label_addr->label = LABEL();
2099    label_addr->update_addr = update_addr;
2100    label_addr->next = common->label_addrs;
2101    common->label_addrs = label_addr;
2102    }
2103    
2104    static SLJIT_INLINE void count_match(compiler_common *common)
2105    {
2106    DEFINE_COMPILER;
2107    
2108    OP2(SLJIT_SUB | SLJIT_SET_E, COUNT_MATCH, 0, COUNT_MATCH, 0, SLJIT_IMM, 1);
2109    add_jump(compiler, &common->calllimit, JUMP(SLJIT_ZERO));
2110    }
2111    
2112    static SLJIT_INLINE void allocate_stack(compiler_common *common, int size)
2113    {
2114    /* May destroy all locals and registers except TMP2. */
2115    DEFINE_COMPILER;
2116    
2117    SLJIT_ASSERT(size > 0);
2118    OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_sw));
2119    #ifdef DESTROY_REGISTERS
2120    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);
2121    OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);
2122    OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);
2123    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), LOCALS0, TMP1, 0);
2124    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), LOCALS1, TMP1, 0);
2125    #endif
2126    add_stub(common, CMP(SLJIT_GREATER, STACK_TOP, 0, STACK_LIMIT, 0));
2127    }
2128    
2129    static SLJIT_INLINE void free_stack(compiler_common *common, int size)
2130    {
2131    DEFINE_COMPILER;
2132    
2133    SLJIT_ASSERT(size > 0);
2134    OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_sw));
2135    }
2136    
2137    static sljit_uw * allocate_read_only_data(compiler_common *common, sljit_uw size)
2138    {
2139    DEFINE_COMPILER;
2140    sljit_uw *result;
2141    
2142    if (SLJIT_UNLIKELY(sljit_get_compiler_error(compiler)))
2143      return NULL;
2144    
2145    result = (sljit_uw *)SLJIT_MALLOC(size + sizeof(sljit_uw), compiler->allocator_data);
2146    if (SLJIT_UNLIKELY(result == NULL))
2147      {
2148      sljit_set_compiler_memory_error(compiler);
2149      return NULL;
2150      }
2151    
2152    *(void**)result = common->read_only_data_head;
2153    common->read_only_data_head = (void *)result;
2154    return result + 1;
2155    }
2156    
2157    static void free_read_only_data(void *current, void *allocator_data)
2158    {
2159    void *next;
2160    
2161    SLJIT_UNUSED_ARG(allocator_data);
2162    
2163    while (current != NULL)
2164      {
2165      next = *(void**)current;
2166      SLJIT_FREE(current, allocator_data);
2167      current = next;
2168      }
2169    }
2170    
2171    static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)
2172    {
2173    DEFINE_COMPILER;
2174    struct sljit_label *loop;
2175    int i;
2176    
2177    /* At this point we can freely use all temporary registers. */
2178    SLJIT_ASSERT(length > 1);
2179    /* TMP1 returns with begin - 1. */
2180    OP2(SLJIT_SUB, SLJIT_R0, 0, SLJIT_MEM1(SLJIT_S0), SLJIT_OFFSETOF(jit_arguments, begin), SLJIT_IMM, IN_UCHARS(1));
2181    if (length < 8)
2182      {
2183      for (i = 1; i < length; i++)
2184        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), OVECTOR(i), SLJIT_R0, 0);
2185      }
2186    else
2187      {
2188      GET_LOCAL_BASE(SLJIT_R1, 0, OVECTOR_START);
2189      OP1(SLJIT_MOV, SLJIT_R2, 0, SLJIT_IMM, length - 1);
2190      loop = LABEL();
2191      OP1(SLJIT_MOVU, SLJIT_MEM1(SLJIT_R1), sizeof(sljit_sw), SLJIT_R0, 0);
2192      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_IMM, 1);
2193      JUMPTO(SLJIT_NOT_ZERO, loop);
2194      }
2195    }
2196    
2197    static SLJIT_INLINE void do_reset_match(compiler_common *common, int length)
2198    {
2199    DEFINE_COMPILER;
2200    struct sljit_label *loop;
2201    int i;
2202    
2203    SLJIT_ASSERT(length > 1);
2204    /* OVECTOR(1) contains the "string begin - 1" constant. */
2205    if (length > 2)
2206      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(1));
2207    if (length < 8)
2208      {
2209      for (i = 2; i < length; i++)
2210        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), OVECTOR(i), TMP1, 0);
2211      }
2212    else
2213      {
2214      GET_LOCAL_BASE(TMP2, 0, OVECTOR_START + sizeof(sljit_sw));
2215      OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_IMM, length - 2);
2216      loop = LABEL();
2217      OP1(SLJIT_MOVU, SLJIT_MEM1(TMP2), sizeof(sljit_sw), TMP1, 0);
2218      OP2(SLJIT_SUB | SLJIT_SET_E, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, 1);
2219      JUMPTO(SLJIT_NOT_ZERO, loop);
2220      }
2221    
2222    OP1(SLJIT_MOV, STACK_TOP, 0, ARGUMENTS, 0);
2223    if (common->mark_ptr != 0)
2224      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->mark_ptr, SLJIT_IMM, 0);
2225    if (common->control_head_ptr != 0)
2226      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->control_head_ptr, SLJIT_IMM, 0);
2227    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(jit_arguments, stack));
2228    OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->start_ptr);
2229    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(struct sljit_stack, base));
2230    }
2231    
2232    static sljit_sw SLJIT_CALL do_search_mark(sljit_sw *current, const pcre_uchar *skip_arg)
2233    {
2234    while (current != NULL)
2235      {
2236      switch (current[-2])
2237        {
2238        case type_then_trap:
2239        break;
2240    
2241        case type_mark:
2242        if (STRCMP_UC_UC(skip_arg, (pcre_uchar *)current[-3]) == 0)
2243          return current[-4];
2244        break;
2245    
2246        default:
2247        SLJIT_ASSERT_STOP();
2248        break;
2249        }
2250      SLJIT_ASSERT(current > (sljit_sw*)current[-1]);
2251      current = (sljit_sw*)current[-1];
2252      }
2253    return -1;
2254    }
2255    
2256    static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)
2257    {
2258    DEFINE_COMPILER;
2259    struct sljit_label *loop;
2260    struct sljit_jump *early_quit;
2261    
2262    /* At this point we can freely use all registers. */
2263    OP1(SLJIT_MOV, SLJIT_S2, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(1));
2264    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), OVECTOR(1), STR_PTR, 0);
2265    
2266    OP1(SLJIT_MOV, SLJIT_R0, 0, ARGUMENTS, 0);
2267    if (common->mark_ptr != 0)
2268      OP1(SLJIT_MOV, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_SP), common->mark_ptr);
2269    OP1(SLJIT_MOV_SI, SLJIT_R1, 0, SLJIT_MEM1(SLJIT_R0), SLJIT_OFFSETOF(jit_arguments, offset_count));
2270    if (common->mark_ptr != 0)
2271      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_R0), SLJIT_OFFSETOF(jit_arguments, mark_ptr), SLJIT_R2, 0);
2272    OP2(SLJIT_SUB, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_R0), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));
2273    OP1(SLJIT_MOV, SLJIT_R0, 0, SLJIT_MEM1(SLJIT_R0), SLJIT_OFFSETOF(jit_arguments, begin));
2274    GET_LOCAL_BASE(SLJIT_S0, 0, OVECTOR_START);
2275    /* Unlikely, but possible */
2276    early_quit = CMP(SLJIT_EQUAL, SLJIT_R1, 0, SLJIT_IMM, 0);
2277    loop = LABEL();
2278    OP2(SLJIT_SUB, SLJIT_S1, 0, SLJIT_MEM1(SLJIT_S0), 0, SLJIT_R0, 0);
2279    OP2(SLJIT_ADD, SLJIT_S0, 0, SLJIT_S0, 0, SLJIT_IMM, sizeof(sljit_sw));
2280    /* Copy the integer value to the output buffer */
2281    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2282    OP2(SLJIT_ASHR, SLJIT_S1, 0, SLJIT_S1, 0, SLJIT_IMM, UCHAR_SHIFT);
2283    #endif
2284    OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_R2), sizeof(int), SLJIT_S1, 0);
2285    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_R1, 0, SLJIT_R1, 0, SLJIT_IMM, 1);
2286    JUMPTO(SLJIT_NOT_ZERO, loop);
2287    JUMPHERE(early_quit);
2288    
2289    /* Calculate the return value, which is the maximum ovector value. */
2290    if (topbracket > 1)
2291      {
2292      GET_LOCAL_BASE(SLJIT_R0, 0, OVECTOR_START + topbracket * 2 * sizeof(sljit_sw));
2293      OP1(SLJIT_MOV, SLJIT_R1, 0, SLJIT_IMM, topbracket + 1);
2294    
2295      /* OVECTOR(0) is never equal to SLJIT_S2. */
2296      loop = LABEL();
2297      OP1(SLJIT_MOVU, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_R0), -(2 * (sljit_sw)sizeof(sljit_sw)));
2298      OP2(SLJIT_SUB, SLJIT_R1, 0, SLJIT_R1, 0, SLJIT_IMM, 1);
2299      CMPTO(SLJIT_EQUAL, SLJIT_R2, 0, SLJIT_S2, 0, loop);
2300      OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_R1, 0);
2301      }
2302    else
2303      OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, 1);
2304    }
2305    
2306    static SLJIT_INLINE void return_with_partial_match(compiler_common *common, struct sljit_label *quit)
2307    {
2308    DEFINE_COMPILER;
2309    struct sljit_jump *jump;
2310    
2311    SLJIT_COMPILE_ASSERT(STR_END == SLJIT_S1, str_end_must_be_saved_reg2);
2312    SLJIT_ASSERT(common->start_used_ptr != 0 && common->start_ptr != 0
2313      && (common->mode == JIT_PARTIAL_SOFT_COMPILE ? common->hit_start != 0 : common->hit_start == 0));
2314    
2315    OP1(SLJIT_MOV, SLJIT_R1, 0, ARGUMENTS, 0);
2316    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, PCRE_ERROR_PARTIAL);
2317    OP1(SLJIT_MOV_SI, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_R1), SLJIT_OFFSETOF(jit_arguments, real_offset_count));
2318    CMPTO(SLJIT_SIG_LESS, SLJIT_R2, 0, SLJIT_IMM, 2, quit);
2319    
2320    /* Store match begin and end. */
2321    OP1(SLJIT_MOV, SLJIT_S0, 0, SLJIT_MEM1(SLJIT_R1), SLJIT_OFFSETOF(jit_arguments, begin));
2322    OP1(SLJIT_MOV, SLJIT_R1, 0, SLJIT_MEM1(SLJIT_R1), SLJIT_OFFSETOF(jit_arguments, offsets));
2323    
2324    jump = CMP(SLJIT_SIG_LESS, SLJIT_R2, 0, SLJIT_IMM, 3);
2325    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);
2326    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2327    OP2(SLJIT_ASHR, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_IMM, UCHAR_SHIFT);
2328    #endif
2329    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_R1), 2 * sizeof(int), SLJIT_R2, 0);
2330    JUMPHERE(jump);
2331    
2332    OP1(SLJIT_MOV, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_SP), common->mode == JIT_PARTIAL_HARD_COMPILE ? common->start_used_ptr : common->hit_start);
2333    OP2(SLJIT_SUB, SLJIT_S1, 0, STR_END, 0, SLJIT_S0, 0);
2334    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2335    OP2(SLJIT_ASHR, SLJIT_S1, 0, SLJIT_S1, 0, SLJIT_IMM, UCHAR_SHIFT);
2336    #endif
2337    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_R1), sizeof(int), SLJIT_S1, 0);
2338    
2339    OP2(SLJIT_SUB, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_S0, 0);
2340    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2341    OP2(SLJIT_ASHR, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_IMM, UCHAR_SHIFT);
2342    #endif
2343    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_R1), 0, SLJIT_R2, 0);
2344    
2345    JUMPTO(SLJIT_JUMP, quit);
2346    }
2347    
2348    static SLJIT_INLINE void check_start_used_ptr(compiler_common *common)
2349    {
2350    /* May destroy TMP1. */
2351    DEFINE_COMPILER;
2352    struct sljit_jump *jump;
2353    
2354    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2355      {
2356      /* The value of -1 must be kept for start_used_ptr! */
2357      OP2(SLJIT_ADD, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, SLJIT_IMM, 1);
2358      /* Jumps if start_used_ptr < STR_PTR, or start_used_ptr == -1. Although overwriting
2359      is not necessary if start_used_ptr == STR_PTR, it does not hurt as well. */
2360      jump = CMP(SLJIT_LESS_EQUAL, TMP1, 0, STR_PTR, 0);
2361      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0);
2362      JUMPHERE(jump);
2363      }
2364    else if (common->mode == JIT_PARTIAL_HARD_COMPILE)
2365      {
2366      jump = CMP(SLJIT_LESS_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0);
2367      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0);
2368      JUMPHERE(jump);
2369      }
2370    }
2371    
2372    static SLJIT_INLINE BOOL char_has_othercase(compiler_common *common, pcre_uchar *cc)
2373    {
2374    /* Detects if the character has an othercase. */
2375    unsigned int c;
2376    
2377    #ifdef SUPPORT_UTF
2378    if (common->utf)
2379      {
2380      GETCHAR(c, cc);
2381      if (c > 127)
2382        {
2383    #ifdef SUPPORT_UCP
2384        return c != UCD_OTHERCASE(c);
2385    #else
2386        return FALSE;
2387    #endif
2388        }
2389    #ifndef COMPILE_PCRE8
2390      return common->fcc[c] != c;
2391    #endif
2392      }
2393    else
2394    #endif
2395      c = *cc;
2396    return MAX_255(c) ? common->fcc[c] != c : FALSE;
2397    }
2398    
2399    static SLJIT_INLINE unsigned int char_othercase(compiler_common *common, unsigned int c)
2400    {
2401    /* Returns with the othercase. */
2402    #ifdef SUPPORT_UTF
2403    if (common->utf && c > 127)
2404      {
2405    #ifdef SUPPORT_UCP
2406      return UCD_OTHERCASE(c);
2407    #else
2408      return c;
2409    #endif
2410      }
2411    #endif
2412    return TABLE_GET(c, common->fcc, c);
2413    }
2414    
2415    static unsigned int char_get_othercase_bit(compiler_common *common, pcre_uchar *cc)
2416    {
2417    /* Detects if the character and its othercase has only 1 bit difference. */
2418    unsigned int c, oc, bit;
2419    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2420    int n;
2421    #endif
2422    
2423    #ifdef SUPPORT_UTF
2424    if (common->utf)
2425      {
2426      GETCHAR(c, cc);
2427      if (c <= 127)
2428        oc = common->fcc[c];
2429      else
2430        {
2431    #ifdef SUPPORT_UCP
2432        oc = UCD_OTHERCASE(c);
2433    #else
2434        oc = c;
2435    #endif
2436        }
2437      }
2438    else
2439      {
2440      c = *cc;
2441      oc = TABLE_GET(c, common->fcc, c);
2442      }
2443    #else
2444    c = *cc;
2445    oc = TABLE_GET(c, common->fcc, c);
2446    #endif
2447    
2448    SLJIT_ASSERT(c != oc);
2449    
2450    bit = c ^ oc;
2451    /* Optimized for English alphabet. */
2452    if (c <= 127 && bit == 0x20)
2453      return (0 << 8) | 0x20;
2454    
2455    /* Since c != oc, they must have at least 1 bit difference. */
2456    if (!is_powerof2(bit))
2457      return 0;
2458    
2459    #if defined COMPILE_PCRE8
2460    
2461    #ifdef SUPPORT_UTF
2462    if (common->utf && c > 127)
2463      {
2464      n = GET_EXTRALEN(*cc);
2465      while ((bit & 0x3f) == 0)
2466        {
2467        n--;
2468        bit >>= 6;
2469        }
2470      return (n << 8) | bit;
2471      }
2472    #endif /* SUPPORT_UTF */
2473    return (0 << 8) | bit;
2474    
2475    #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2476    
2477    #ifdef SUPPORT_UTF
2478    if (common->utf && c > 65535)
2479      {
2480      if (bit >= (1 << 10))
2481        bit >>= 10;
2482      else
2483        return (bit < 256) ? ((2 << 8) | bit) : ((3 << 8) | (bit >> 8));
2484      }
2485    #endif /* SUPPORT_UTF */
2486    return (bit < 256) ? ((0 << 8) | bit) : ((1 << 8) | (bit >> 8));
2487    
2488    #endif /* COMPILE_PCRE[8|16|32] */
2489    }
2490    
2491    static void check_partial(compiler_common *common, BOOL force)
2492    {
2493    /* Checks whether a partial matching is occurred. Does not modify registers. */
2494    DEFINE_COMPILER;
2495    struct sljit_jump *jump = NULL;
2496    
2497    SLJIT_ASSERT(!force || common->mode != JIT_COMPILE);
2498    
2499    if (common->mode == JIT_COMPILE)
2500      return;
2501    
2502    if (!force)
2503      jump = CMP(SLJIT_GREATER_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0);
2504    else if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2505      jump = CMP(SLJIT_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, SLJIT_IMM, -1);
2506    
2507    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2508      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->hit_start, SLJIT_IMM, 0);
2509    else
2510      {
2511      if (common->partialmatchlabel != NULL)
2512        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2513      else
2514        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2515      }
2516    
2517    if (jump != NULL)
2518      JUMPHERE(jump);
2519    }
2520    
2521    static void check_str_end(compiler_common *common, jump_list **end_reached)
2522    {
2523    /* Does not affect registers. Usually used in a tight spot. */
2524    DEFINE_COMPILER;
2525    struct sljit_jump *jump;
2526    
2527    if (common->mode == JIT_COMPILE)
2528      {
2529      add_jump(compiler, end_reached, CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2530      return;
2531      }
2532    
2533    jump = CMP(SLJIT_LESS, STR_PTR, 0, STR_END, 0);
2534    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2535      {
2536      add_jump(compiler, end_reached, CMP(SLJIT_GREATER_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0));
2537      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->hit_start, SLJIT_IMM, 0);
2538      add_jump(compiler, end_reached, JUMP(SLJIT_JUMP));
2539      }
2540    else
2541      {
2542      add_jump(compiler, end_reached, CMP(SLJIT_GREATER_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0));
2543      if (common->partialmatchlabel != NULL)
2544        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2545      else
2546        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2547      }
2548    JUMPHERE(jump);
2549    }
2550    
2551    static void detect_partial_match(compiler_common *common, jump_list **backtracks)
2552    {
2553    DEFINE_COMPILER;
2554    struct sljit_jump *jump;
2555    
2556    if (common->mode == JIT_COMPILE)
2557      {
2558      add_jump(compiler, backtracks, CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2559      return;
2560      }
2561    
2562    /* Partial matching mode. */
2563    jump = CMP(SLJIT_LESS, STR_PTR, 0, STR_END, 0);
2564    add_jump(compiler, backtracks, CMP(SLJIT_GREATER_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0));
2565    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2566      {
2567      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->hit_start, SLJIT_IMM, 0);
2568      add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));
2569      }
2570    else
2571      {
2572      if (common->partialmatchlabel != NULL)
2573        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2574      else
2575        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2576      }
2577    JUMPHERE(jump);
2578    }
2579    
2580    static void peek_char(compiler_common *common, pcre_uint32 max)
2581    {
2582    /* Reads the character into TMP1, keeps STR_PTR.
2583    Does not check STR_END. TMP2 Destroyed. */
2584    DEFINE_COMPILER;
2585    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2586    struct sljit_jump *jump;
2587    #endif
2588    
2589    SLJIT_UNUSED_ARG(max);
2590    
2591    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2592    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2593    if (common->utf)
2594      {
2595      if (max < 128) return;
2596    
2597      jump = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2598      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2599      add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2600      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2601      JUMPHERE(jump);
2602      }
2603    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2604    
2605    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2606    if (common->utf)
2607      {
2608      if (max < 0xd800) return;
2609    
2610      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2611      jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2612      /* TMP2 contains the high surrogate. */
2613      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2614      OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2615      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2616      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2617      OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2618      JUMPHERE(jump);
2619      }
2620    #endif
2621    }
2622    
2623    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2624    
2625    static BOOL is_char7_bitset(const pcre_uint8 *bitset, BOOL nclass)
2626    {
2627    /* Tells whether the character codes below 128 are enough
2628    to determine a match. */
2629    const pcre_uint8 value = nclass ? 0xff : 0;
2630    const pcre_uint8 *end = bitset + 32;
2631    
2632    bitset += 16;
2633    do
2634      {
2635      if (*bitset++ != value)
2636        return FALSE;
2637      }
2638    while (bitset < end);
2639    return TRUE;
2640    }
2641    
2642    static void read_char7_type(compiler_common *common, BOOL full_read)
2643    {
2644    /* Reads the precise character type of a character into TMP1, if the character
2645    is less than 128. Otherwise it returns with zero. Does not check STR_END. The
2646    full_read argument tells whether characters above max are accepted or not. */
2647    DEFINE_COMPILER;
2648    struct sljit_jump *jump;
2649    
2650    SLJIT_ASSERT(common->utf);
2651    
2652    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2653    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2654    
2655    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2656    
2657    if (full_read)
2658      {
2659      jump = CMP(SLJIT_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2660      OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2661      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2662      JUMPHERE(jump);
2663      }
2664    }
2665    
2666    #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2667    
2668    static void read_char_range(compiler_common *common, pcre_uint32 min, pcre_uint32 max, BOOL update_str_ptr)
2669    {
2670    /* Reads the precise value of a character into TMP1, if the character is
2671    between min and max (c >= min && c <= max). Otherwise it returns with a value
2672    outside the range. Does not check STR_END. */
2673    DEFINE_COMPILER;
2674    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2675    struct sljit_jump *jump;
2676    #endif
2677    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2678    struct sljit_jump *jump2;
2679    #endif
2680    
2681    SLJIT_UNUSED_ARG(update_str_ptr);
2682    SLJIT_UNUSED_ARG(min);
2683    SLJIT_UNUSED_ARG(max);
2684    SLJIT_ASSERT(min <= max);
2685    
2686    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2687    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2688    
2689    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2690    if (common->utf)
2691      {
2692      if (max < 128 && !update_str_ptr) return;
2693    
2694      jump = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2695      if (min >= 0x10000)
2696        {
2697        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xf0);
2698        if (update_str_ptr)
2699          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2700        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2701        jump2 = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0x7);
2702        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2703        OP2(SLJIT_AND, TMP1, 0, TMP1, 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(1));
2706        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2707        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2708        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2709        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2710        if (!update_str_ptr)
2711          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2712        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2713        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2714        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2715        JUMPHERE(jump2);
2716        if (update_str_ptr)
2717          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2718        }
2719      else if (min >= 0x800 && max <= 0xffff)
2720        {
2721        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xe0);
2722        if (update_str_ptr)
2723          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2724        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2725        jump2 = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xf);
2726        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2727        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2728        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2729        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2730        if (!update_str_ptr)
2731          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2732        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2733        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2734        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2735        JUMPHERE(jump2);
2736        if (update_str_ptr)
2737          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2738        }
2739      else if (max >= 0x800)
2740        add_jump(compiler, (max < 0x10000) ? &common->utfreadchar16 : &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2741      else if (max < 128)
2742        {
2743        OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2744        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2745        }
2746      else
2747        {
2748        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2749        if (!update_str_ptr)
2750          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2751        else
2752          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2753        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2754        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2755        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2756        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2757        if (update_str_ptr)
2758          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2759        }
2760      JUMPHERE(jump);
2761      }
2762    #endif
2763    
2764    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2765    if (common->utf)
2766      {
2767      if (max >= 0x10000)
2768        {
2769        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2770        jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2771        /* TMP2 contains the high surrogate. */
2772        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2773        OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2774        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2775        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2776        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2777        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2778        JUMPHERE(jump);
2779        return;
2780        }
2781    
2782      if (max < 0xd800 && !update_str_ptr) return;
2783    
2784      /* Skip low surrogate if necessary. */
2785      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2786      jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2787      if (update_str_ptr)
2788        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2789      if (max >= 0xd800)
2790        OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0x10000);
2791      JUMPHERE(jump);
2792      }
2793    #endif
2794    }
2795    
2796    static SLJIT_INLINE void read_char(compiler_common *common)
2797    {
2798    read_char_range(common, 0, READ_CHAR_MAX, TRUE);
2799    }
2800    
2801    static void read_char8_type(compiler_common *common, BOOL update_str_ptr)
2802    {
2803    /* Reads the character type into TMP1, updates STR_PTR. Does not check STR_END. */
2804    DEFINE_COMPILER;
2805    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2806    struct sljit_jump *jump;
2807    #endif
2808    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2809    struct sljit_jump *jump2;
2810    #endif
2811    
2812    SLJIT_UNUSED_ARG(update_str_ptr);
2813    
2814    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2815    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2816    
2817    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2818    if (common->utf)
2819      {
2820      /* This can be an extra read in some situations, but hopefully
2821      it is needed in most cases. */
2822      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2823      jump = CMP(SLJIT_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2824      if (!update_str_ptr)
2825        {
2826        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2827        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2828        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2829        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2830        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2831        OP2(SLJIT_OR, TMP2, 0, TMP2, 0, TMP1, 0);
2832        OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2833        jump2 = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 255);
2834        OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2835        JUMPHERE(jump2);
2836        }
2837      else
2838        add_jump(compiler, &common->utfreadtype8, JUMP(SLJIT_FAST_CALL));
2839      JUMPHERE(jump);
2840      return;
2841      }
2842    #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2843    
2844    #if !defined COMPILE_PCRE8
2845    /* The ctypes array contains only 256 values. */
2846    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2847    jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 255);
2848    #endif
2849    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2850    #if !defined COMPILE_PCRE8
2851    JUMPHERE(jump);
2852    #endif
2853    
2854    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2855    if (common->utf && update_str_ptr)
2856      {
2857      /* Skip low surrogate if necessary. */
2858      OP2(SLJIT_SUB, TMP2, 0, TMP2, 0, SLJIT_IMM, 0xd800);
2859      jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2860      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2861      JUMPHERE(jump);
2862      }
2863    #endif /* SUPPORT_UTF && COMPILE_PCRE16 */
2864    }
2865    
2866    static void skip_char_back(compiler_common *common)
2867    {
2868    /* Goes one character back. Affects STR_PTR and TMP1. Does not check begin. */
2869    DEFINE_COMPILER;
2870    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2871    #if defined COMPILE_PCRE8
2872    struct sljit_label *label;
2873    
2874    if (common->utf)
2875      {
2876      label = LABEL();
2877      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
2878      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2879      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xc0);
2880      CMPTO(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, 0x80, label);
2881      return;
2882      }
2883    #elif defined COMPILE_PCRE16
2884    if (common->utf)
2885      {
2886      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
2887      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2888      /* Skip low surrogate if necessary. */
2889      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
2890      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xdc00);
2891      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
2892      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
2893      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2894      return;
2895      }
2896    #endif /* COMPILE_PCRE[8|16] */
2897    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2898    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2899    }
2900    
2901    static void check_newlinechar(compiler_common *common, int nltype, jump_list **backtracks, BOOL jumpifmatch)
2902    {
2903    /* Character comes in TMP1. Checks if it is a newline. TMP2 may be destroyed. */
2904    DEFINE_COMPILER;
2905    struct sljit_jump *jump;
2906    
2907    if (nltype == NLTYPE_ANY)
2908      {
2909      add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));
2910      add_jump(compiler, backtracks, JUMP(jumpifmatch ? SLJIT_NOT_ZERO : SLJIT_ZERO));
2911      }
2912    else if (nltype == NLTYPE_ANYCRLF)
2913      {
2914      if (jumpifmatch)
2915        {
2916        add_jump(compiler, backtracks, CMP(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR));
2917        add_jump(compiler, backtracks, CMP(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL));
2918        }
2919      else
2920        {
2921        jump = CMP(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR);
2922        add_jump(compiler, backtracks, CMP(SLJIT_NOT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL));
2923        JUMPHERE(jump);
2924        }
2925      }
2926    else
2927      {
2928      SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline < 256);
2929      add_jump(compiler, backtracks, CMP(jumpifmatch ? SLJIT_EQUAL : SLJIT_NOT_EQUAL, TMP1, 0, SLJIT_IMM, common->newline));
2930      }
2931    }
2932    
2933    #ifdef SUPPORT_UTF
2934    
2935    #if defined COMPILE_PCRE8
2936    static void do_utfreadchar(compiler_common *common)
2937    {
2938    /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2939    of the character (>= 0xc0). Return char value in TMP1, length in TMP2. */
2940    DEFINE_COMPILER;
2941    struct sljit_jump *jump;
2942    
2943    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2944    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2945    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2946    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2947    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2948    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2949    
2950    /* Searching for the first zero. */
2951    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
2952    jump = JUMP(SLJIT_NOT_ZERO);
2953    /* Two byte sequence. */
2954    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2955    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(2));
2956    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2957    
2958    JUMPHERE(jump);
2959    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2960    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
2961    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2962    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2963    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2964    
2965    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2966    jump = JUMP(SLJIT_NOT_ZERO);
2967    /* Three byte sequence. */
2968    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2969    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(3));
2970    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2971    
2972    /* Four byte sequence. */
2973    JUMPHERE(jump);
2974    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2975    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2976    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2977    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2978    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2979    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2980    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(4));
2981    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2982    }
2983    
2984    static void do_utfreadchar16(compiler_common *common)
2985    {
2986    /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2987    of the character (>= 0xc0). Return value in TMP1. */
2988    DEFINE_COMPILER;
2989    struct sljit_jump *jump;
2990    
2991    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2992    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2993    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2994    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2995    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2996    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2997    
2998    /* Searching for the first zero. */
2999    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
3000    jump = JUMP(SLJIT_NOT_ZERO);
3001    /* Two byte sequence. */
3002    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3003    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3004    
3005    JUMPHERE(jump);
3006    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x400);
3007    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_NOT_ZERO);
3008    /* This code runs only in 8 bit mode. No need to shift the value. */
3009    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
3010    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
3011    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
3012    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
3013    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
3014    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
3015    /* Three byte sequence. */
3016    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
3017    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3018    }
3019    
3020    static void do_utfreadtype8(compiler_common *common)
3021    {
3022    /* Fast decoding a UTF-8 character type. TMP2 contains the first byte
3023    of the character (>= 0xc0). Return value in TMP1. */
3024    DEFINE_COMPILER;
3025    struct sljit_jump *jump;
3026    struct sljit_jump *compare;
3027    
3028    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
3029    
3030    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0x20);
3031    jump = JUMP(SLJIT_NOT_ZERO);
3032    /* Two byte sequence. */
3033    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
3034    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3035    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);
3036    /* The upper 5 bits are known at this point. */
3037    compare = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0x3);
3038    OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
3039    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
3040    OP2(SLJIT_OR, TMP2, 0, TMP2, 0, TMP1, 0);
3041    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
3042    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3043    
3044    JUMPHERE(compare);
3045    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
3046    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3047    
3048    /* We only have types for characters less than 256. */
3049    JUMPHERE(jump);
3050    OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
3051    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
3052    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
3053    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3054    }
3055    
3056    #endif /* COMPILE_PCRE8 */
3057    
3058    #endif /* SUPPORT_UTF */
3059    
3060    #ifdef SUPPORT_UCP
3061    
3062    /* UCD_BLOCK_SIZE must be 128 (see the assert below). */
3063    #define UCD_BLOCK_MASK 127
3064    #define UCD_BLOCK_SHIFT 7
3065    
3066    static void do_getucd(compiler_common *common)
3067    {
3068    /* Search the UCD record for the character comes in TMP1.
3069    Returns chartype in TMP1 and UCD offset in TMP2. */
3070    DEFINE_COMPILER;
3071    
3072    SLJIT_ASSERT(UCD_BLOCK_SIZE == 128 && sizeof(ucd_record) == 8);
3073    
3074    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
3075    OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
3076    OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(ucd_stage1));
3077    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);
3078    OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
3079    OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, TMP2, 0);
3080    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_stage2));
3081    OP1(SLJIT_MOV_UH, TMP2, 0, SLJIT_MEM2(TMP2, TMP1), 1);
3082    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_records) + SLJIT_OFFSETOF(ucd_record, chartype));
3083    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM2(TMP1, TMP2), 3);
3084    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3085    }
3086    #endif
3087    
3088    static SLJIT_INLINE struct sljit_label *mainloop_entry(compiler_common *common, BOOL hascrorlf, BOOL firstline)
3089    {
3090    DEFINE_COMPILER;
3091    struct sljit_label *mainloop;
3092    struct sljit_label *newlinelabel = NULL;
3093    struct sljit_jump *start;
3094    struct sljit_jump *end = NULL;
3095    struct sljit_jump *nl = NULL;
3096    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3097    struct sljit_jump *singlechar;
3098    #endif
3099    jump_list *newline = NULL;
3100    BOOL newlinecheck = FALSE;
3101    BOOL readuchar = FALSE;
3102    
3103    if (!(hascrorlf || firstline) && (common->nltype == NLTYPE_ANY ||
3104        common->nltype == NLTYPE_ANYCRLF || common->newline > 255))
3105      newlinecheck = TRUE;
3106    
3107    if (firstline)
3108      {
3109      /* Search for the end of the first line. */
3110      SLJIT_ASSERT(common->first_line_end != 0);
3111      OP1(SLJIT_MOV, TMP3, 0, STR_PTR, 0);
3112    
3113      if (common->nltype == NLTYPE_FIXED && common->newline > 255)
3114        {
3115        mainloop = LABEL();
3116        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3117        end = CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3118        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(-1));
3119        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
3120        CMPTO(SLJIT_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, mainloop);
3121        CMPTO(SLJIT_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, mainloop);
3122        JUMPHERE(end);
3123        OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_SP), common->first_line_end, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3124        }
3125      else
3126        {
3127        end = CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3128        mainloop = LABEL();
3129        /* Continual stores does not cause data dependency. */
3130        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->first_line_end, STR_PTR, 0);
3131        read_char_range(common, common->nlmin, common->nlmax, TRUE);
3132        check_newlinechar(common, common->nltype, &newline, TRUE);
3133        CMPTO(SLJIT_LESS, STR_PTR, 0, STR_END, 0, mainloop);
3134        JUMPHERE(end);
3135        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->first_line_end, STR_PTR, 0);
3136        set_jumps(newline, LABEL());
3137        }
3138    
3139      OP1(SLJIT_MOV, STR_PTR, 0, TMP3, 0);
3140      }
3141    
3142    start = JUMP(SLJIT_JUMP);
3143    
3144    if (newlinecheck)
3145      {
3146      newlinelabel = LABEL();
3147      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3148      end = CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3149      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3150      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, common->newline & 0xff);
3151      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
3152    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
3153      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, UCHAR_SHIFT);
3154    #endif
3155      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3156      nl = JUMP(SLJIT_JUMP);
3157      }
3158    
3159    mainloop = LABEL();
3160    
3161    /* Increasing the STR_PTR here requires one less jump in the most common case. */
3162    #ifdef SUPPORT_UTF
3163    if (common->utf) readuchar = TRUE;
3164    #endif
3165    if (newlinecheck) readuchar = TRUE;
3166    
3167    if (readuchar)
3168      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3169    
3170    if (newlinecheck)
3171      CMPTO(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, newlinelabel);
3172    
3173    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3174    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3175    #if defined COMPILE_PCRE8
3176    if (common->utf)
3177      {
3178      singlechar = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
3179      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
3180      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3181      JUMPHERE(singlechar);
3182      }
3183    #elif defined COMPILE_PCRE16
3184    if (common->utf)
3185      {
3186      singlechar = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xd800);
3187      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
3188      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);
3189      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
3190      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
3191      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3192      JUMPHERE(singlechar);
3193      }
3194    #endif /* COMPILE_PCRE[8|16] */
3195    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
3196    JUMPHERE(start);
3197    
3198    if (newlinecheck)
3199      {
3200      JUMPHERE(end);
3201      JUMPHERE(nl);
3202      }
3203    
3204    return mainloop;
3205    }
3206    
3207    #define MAX_N_CHARS 16
3208    #define MAX_N_BYTES 8
3209    
3210    static SLJIT_INLINE void add_prefix_byte(pcre_uint8 byte, pcre_uint8 *bytes)
3211    {
3212    pcre_uint8 len = bytes[0];
3213    int i;
3214    
3215    if (len == 255)
3216      return;
3217    
3218    if (len == 0)
3219      {
3220      bytes[0] = 1;
3221      bytes[1] = byte;
3222      return;
3223      }
3224    
3225    for (i = len; i > 0; i--)
3226      if (bytes[i] == byte)
3227        return;
3228    
3229    if (len >= MAX_N_BYTES - 1)
3230    {    {
3231    list_item->type = type;    bytes[0] = 255;
3232    list_item->data = data;    return;
   list_item->start = start;  
   list_item->leave = LABEL();  
   list_item->next = common->stubs;  
   common->stubs = list_item;  
3233    }    }
 }  
3234    
3235  static void flush_stubs(compiler_common *common)  len++;
3236  {  bytes[len] = byte;
3237  DEFINE_COMPILER;  bytes[0] = len;
3238  stub_list* list_item = common->stubs;  }
3239    
3240    static int scan_prefix(compiler_common *common, pcre_uchar *cc, pcre_uint32 *chars, pcre_uint8 *bytes, int max_chars, pcre_uint32 *rec_count)
3241    {
3242    /* Recursive function, which scans prefix literals. */
3243    BOOL last, any, caseless;
3244    int len, repeat, len_save, consumed = 0;
3245    pcre_uint32 chr, mask;
3246    pcre_uchar *alternative, *cc_save, *oc;
3247    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3248    pcre_uchar othercase[8];
3249    #elif defined SUPPORT_UTF && defined COMPILE_PCRE16
3250    pcre_uchar othercase[2];
3251    #else
3252    pcre_uchar othercase[1];
3253    #endif
3254    
3255  while (list_item)  repeat = 1;
3256    while (TRUE)
3257    {    {
3258    JUMPHERE(list_item->start);    if (*rec_count == 0)
3259    switch(list_item->type)      return 0;
3260      (*rec_count)--;
3261    
3262      last = TRUE;
3263      any = FALSE;
3264      caseless = FALSE;
3265    
3266      switch (*cc)
3267      {      {
3268      case stack_alloc:      case OP_CHARI:
3269      add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));      caseless = TRUE;
3270        case OP_CHAR:
3271        last = FALSE;
3272        cc++;
3273      break;      break;
3274    
3275      case max_index:      case OP_SOD:
3276      OP1(SLJIT_MOV, MAX_INDEX, 0, SLJIT_IMM, list_item->data);      case OP_SOM:
3277        case OP_SET_SOM:
3278        case OP_NOT_WORD_BOUNDARY:
3279        case OP_WORD_BOUNDARY:
3280        case OP_EODN:
3281        case OP_EOD:
3282        case OP_CIRC:
3283        case OP_CIRCM:
3284        case OP_DOLL:
3285        case OP_DOLLM:
3286        /* Zero width assertions. */
3287        cc++;
3288        continue;
3289    
3290        case OP_ASSERT:
3291        case OP_ASSERT_NOT:
3292        case OP_ASSERTBACK:
3293        case OP_ASSERTBACK_NOT:
3294        cc = bracketend(cc);
3295        continue;
3296    
3297        case OP_PLUSI:
3298        case OP_MINPLUSI:
3299        case OP_POSPLUSI:
3300        caseless = TRUE;
3301        case OP_PLUS:
3302        case OP_MINPLUS:
3303        case OP_POSPLUS:
3304        cc++;
3305      break;      break;
     }  
   JUMPTO(SLJIT_JUMP, list_item->leave);  
   list_item = list_item->next;  
   }  
 common->stubs = NULL;  
 }  
3306    
3307  static SLJIT_INLINE void allocate_stack(compiler_common *common, int size)      case OP_EXACTI:
3308  {      caseless = TRUE;
3309  /* May destroy all locals and registers except TMP2. */      case OP_EXACT:
3310  DEFINE_COMPILER;      repeat = GET2(cc, 1);
3311        last = FALSE;
3312        cc += 1 + IMM2_SIZE;
3313        break;
3314    
3315  OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_w));      case OP_QUERYI:
3316  #ifdef DESTROY_REGISTERS      case OP_MINQUERYI:
3317  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);      case OP_POSQUERYI:
3318  OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);      caseless = TRUE;
3319  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);      case OP_QUERY:
3320  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);      case OP_MINQUERY:
3321  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);      case OP_POSQUERY:
3322        len = 1;
3323        cc++;
3324    #ifdef SUPPORT_UTF
3325        if (common->utf && HAS_EXTRALEN(*cc)) len += GET_EXTRALEN(*cc);
3326  #endif  #endif
3327  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);
3328  }      if (max_chars == 0)
3329          return consumed;
3330        last = FALSE;
3331        break;
3332    
3333  static SLJIT_INLINE void free_stack(compiler_common *common, int size)      case OP_KET:
3334  {      cc += 1 + LINK_SIZE;
3335  DEFINE_COMPILER;      continue;
 OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_w));  
 }  
3336    
3337  static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)      case OP_ALT:
3338  {      cc += GET(cc, 1);
3339  DEFINE_COMPILER;      continue;
 struct sljit_label *loop;  
 int i;  
 /* At this point we can freely use all temporary registers. */  
 /* TMP1 returns with begin - 1. */  
 OP1(SLJIT_MOV, MAX_INDEX, 0, SLJIT_IMM, 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);  
   }  
 }  
3340    
3341  static SLJIT_INLINE void copy_ovector(compiler_common *common)      case OP_ONCE:
3342  {      case OP_ONCE_NC:
3343  DEFINE_COMPILER;      case OP_BRA:
3344  struct sljit_label *loop;      case OP_BRAPOS:
3345  struct sljit_jump *earlyexit;      case OP_CBRA:
3346        case OP_CBRAPOS:
3347        alternative = cc + GET(cc, 1);
3348        while (*alternative == OP_ALT)
3349          {
3350          max_chars = scan_prefix(common, alternative + 1 + LINK_SIZE, chars, bytes, max_chars, rec_count);
3351          if (max_chars == 0)
3352            return consumed;
3353          alternative += GET(alternative, 1);
3354          }
3355    
3356  /* At this point we can freely use all registers. */      if (*cc == OP_CBRA || *cc == OP_CBRAPOS)
3357  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, ARGUMENTS, 0);        cc += IMM2_SIZE;
3358  OP1(SLJIT_MOV_SI, SLJIT_TEMPORARY_REG2, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, offsetcount));      cc += 1 + LINK_SIZE;
3359  OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));      continue;
 OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, begin));  
 OP2(SLJIT_ADD, SLJIT_GENERAL_REG1, 0, SLJIT_LOCALS_REG, 0, SLJIT_IMM, OVECTOR_START);  
 /* Unlikely, but possible */  
 earlyexit = CMP(SLJIT_C_EQUAL, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, 0);  
 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);  
 }  
3360    
3361  stati