/[pcre]/code/tags/pcre-8.37/pcre_jit_compile.c
ViewVC logotype

Diff of /code/tags/pcre-8.37/pcre_jit_compile.c

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

revision 678 by ph10, Sun Aug 28 15:23:03 2011 UTC revision 1549 by zherczeg, Tue Apr 21 07:13:00 2015 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2008 University of Cambridge             Copyright (c) 1997-2013 University of Cambridge
10    
11    The machine code generator part (this module) was written by Zoltan Herczeg    The machine code generator part (this module) was written by Zoltan Herczeg
12                        Copyright (c) 2010-2011                        Copyright (c) 2010-2013
13    
14  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
15  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 46  POSSIBILITY OF SUCH DAMAGE. Line 46  POSSIBILITY OF SUCH DAMAGE.
46    
47  #include "pcre_internal.h"  #include "pcre_internal.h"
48    
49  #ifdef SUPPORT_JIT  #if defined SUPPORT_JIT
50    
51  /* All-in-one: Since we use the JIT compiler only from here,  /* All-in-one: Since we use the JIT compiler only from here,
52  we just include it. This way we don't need to touch the build  we just include it. This way we don't need to touch the build
53  system files. */  system files. */
54    
55    #define SLJIT_MALLOC(size, allocator_data) (PUBL(malloc))(size)
56    #define SLJIT_FREE(ptr, allocator_data) (PUBL(free))(ptr)
57  #define SLJIT_CONFIG_AUTO 1  #define SLJIT_CONFIG_AUTO 1
58    #define SLJIT_CONFIG_STATIC 1
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    int calllimit;    int real_offset_count;
173    uschar notbol;    int offset_count;
174    uschar noteol;    pcre_uint8 notbol;
175    uschar notempty;    pcre_uint8 noteol;
176    uschar notempty_atstart;    pcre_uint8 notempty;
177      pcre_uint8 notempty_atstart;
178  } jit_arguments;  } jit_arguments;
179    
180  typedef struct executable_function {  typedef struct executable_functions {
181    void *executable_func;    void *executable_funcs[JIT_NUMBER_OF_COMPILE_MODES];
182    pcre_jit_callback callback;    void *read_only_data_heads[JIT_NUMBER_OF_COMPILE_MODES];
183      sljit_uw executable_sizes[JIT_NUMBER_OF_COMPILE_MODES];
184      PUBL(jit_callback) callback;
185    void *userdata;    void *userdata;
186  } executable_function;    pcre_uint32 top_bracket;
187      pcre_uint32 limit_match;
188    } executable_functions;
189    
190  typedef struct jump_list {  typedef struct jump_list {
191    struct sljit_jump *jump;    struct sljit_jump *jump;
192    struct jump_list *next;    struct jump_list *next;
193  } jump_list;  } jump_list;
194    
 enum stub_types { stack_alloc, 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 256  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;    jump_list *calllimit;
408    jump_list *stackalloc;    jump_list *stackalloc;
# Line 290  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 310  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))
 /* Head of the saved local variables */  
 #define LOCALS_HEAD      (4 * sizeof(sljit_w))  
 /* Head of the last recursion. */  
 #define RECURSIVE_HEAD   (5 * sizeof(sljit_w))  
 /* Number of recursions. */  
 #define CALL_COUNT       (6 * sizeof(sljit_w))  
492  /* Max limit of recursions. */  /* Max limit of recursions. */
493  #define CALL_LIMIT       (7 * sizeof(sljit_w))  #define LIMIT_MATCH      (4 * sizeof(sljit_sw))
 /* Last known position of the requested byte. */  
 #define REQ_BYTE_PTR     (8 * sizeof(sljit_w))  
 /* End pointer of the first line. */  
 #define FIRSTLINE_END    (9 * 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    (10 * 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 389  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 405  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 435  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 447  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    
1068  while (cc < ccend)  while (cc < ccend)
1069    {    {
1070      space = 0;
1071      size = 0;
1072      bracketlen = 0;
1073      if (private_data_ptr > SLJIT_MAX_LOCAL_SIZE)
1074        return;
1075    
1076      if (*cc == OP_ONCE || *cc == OP_ONCE_NC || *cc == OP_BRA || *cc == OP_CBRA || *cc == OP_COND)
1077        if (detect_repeat(common, cc))
1078          {
1079          /* These brackets are converted to repeats, so no global
1080          based single character repeat is allowed. */
1081          if (cc >= end)
1082            end = bracketend(cc);
1083          }
1084    
1085    switch(*cc)    switch(*cc)
1086      {      {
1087        case OP_KET:
1088        if (common->private_data_ptrs[cc + 1 - common->start] != 0)
1089          {
1090          common->private_data_ptrs[cc - common->start] = private_data_ptr;
1091          private_data_ptr += sizeof(sljit_sw);
1092          cc += common->private_data_ptrs[cc + 1 - common->start];
1093          }
1094        cc += 1 + LINK_SIZE;
1095        break;
1096    
1097      case OP_ASSERT:      case OP_ASSERT:
1098      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1099      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1100      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1101      case OP_ONCE:      case OP_ONCE:
1102        case OP_ONCE_NC:
1103      case OP_BRAPOS:      case OP_BRAPOS:
1104      case OP_SBRA:      case OP_SBRA:
1105      case OP_SBRAPOS:      case OP_SBRAPOS:
1106      case OP_SCOND:      case OP_SCOND:
1107      localsize++;      common->private_data_ptrs[cc - common->start] = private_data_ptr;
1108      cc += 1 + LINK_SIZE;      private_data_ptr += sizeof(sljit_sw);
1109      break;      bracketlen = 1 + LINK_SIZE;
   
     case OP_CBRA:  
     case OP_SCBRA:  
     localsize++;  
     cc += 1 + LINK_SIZE + 2;  
1110      break;      break;
1111    
1112      case OP_CBRAPOS:      case OP_CBRAPOS:
1113      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1114      localsize += 2;      common->private_data_ptrs[cc - common->start] = private_data_ptr;
1115      cc += 1 + LINK_SIZE + 2;      private_data_ptr += sizeof(sljit_sw);
1116        bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
1117      break;      break;
1118    
1119      case OP_COND:      case OP_COND:
1120      /* Might be a hidden SCOND. */      /* Might be a hidden SCOND. */
1121      alternative = cc + GET(cc, 1);      alternative = cc + GET(cc, 1);
1122      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1123        localsize++;        {
1124      cc += 1 + LINK_SIZE;        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1125          private_data_ptr += sizeof(sljit_sw);
1126          }
1127        bracketlen = 1 + LINK_SIZE;
1128        break;
1129    
1130        case OP_BRA:
1131        bracketlen = 1 + LINK_SIZE;
1132        break;
1133    
1134        case OP_CBRA:
1135        case OP_SCBRA:
1136        bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
1137        break;
1138    
1139        CASE_ITERATOR_PRIVATE_DATA_1
1140        space = 1;
1141        size = -2;
1142        break;
1143    
1144        CASE_ITERATOR_PRIVATE_DATA_2A
1145        space = 2;
1146        size = -2;
1147        break;
1148    
1149        CASE_ITERATOR_PRIVATE_DATA_2B
1150        space = 2;
1151        size = -(2 + IMM2_SIZE);
1152      break;      break;
1153    
1154        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1155        space = 1;
1156        size = 1;
1157        break;
1158    
1159        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1160        if (cc[1] != OP_ANYNL && cc[1] != OP_EXTUNI)
1161          space = 2;
1162        size = 1;
1163        break;
1164    
1165        CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1166        if (cc[1 + IMM2_SIZE] != OP_ANYNL && cc[1 + IMM2_SIZE] != OP_EXTUNI)
1167          space = 2;
1168        size = 1 + IMM2_SIZE;
1169        break;
1170    
1171        case OP_CLASS:
1172        case OP_NCLASS:
1173        size += 1 + 32 / sizeof(pcre_uchar);
1174        space = get_class_iterator_size(cc + size);
1175        break;
1176    
1177    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1178        case OP_XCLASS:
1179        size = GET(cc, 1);
1180        space = get_class_iterator_size(cc + size);
1181        break;
1182    #endif
1183    
1184      default:      default:
1185      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
1186      SLJIT_ASSERT(cc != NULL);      SLJIT_ASSERT(cc != NULL);
1187      break;      break;
1188      }      }
1189    
1190      /* Character iterators, which are not inside a repeated bracket,
1191         gets a private slot instead of allocating it on the stack. */
1192      if (space > 0 && cc >= end)
1193        {
1194        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1195        private_data_ptr += sizeof(sljit_sw) * space;
1196        }
1197    
1198      if (size != 0)
1199        {
1200        if (size < 0)
1201          {
1202          cc += -size;
1203    #ifdef SUPPORT_UTF
1204          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1205    #endif
1206          }
1207        else
1208          cc += size;
1209        }
1210    
1211      if (bracketlen > 0)
1212        {
1213        if (cc >= end)
1214          {
1215          end = bracketend(cc);
1216          if (end[-1 - LINK_SIZE] == OP_KET)
1217            end = NULL;
1218          }
1219        cc += bracketlen;
1220        }
1221    }    }
1222  SLJIT_ASSERT(cc == ccend);  *private_data_start = private_data_ptr;
 return localsize;  
1223  }  }
1224    
1225  static void copy_locals(compiler_common *common, uschar *cc, uschar *ccend,  /* Returns with a frame_types (always < 0) if no need for frame. */
1226    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)
1227  {  {
1228  DEFINE_COMPILER;  int length = 0;
1229  int srcw[2];  int possessive = 0;
1230  int count;  BOOL stack_restore = FALSE;
1231  BOOL tmp1next = TRUE;  BOOL setsom_found = recursive;
1232  BOOL tmp1empty = TRUE;  BOOL setmark_found = recursive;
1233  BOOL tmp2empty = TRUE;  /* The last capture is a local variable even for recursions. */
1234  uschar *alternative;  BOOL capture_last_found = FALSE;
1235  enum {  
1236    start,  #if defined DEBUG_FORCE_CONTROL_HEAD && DEBUG_FORCE_CONTROL_HEAD
1237    loop,  SLJIT_ASSERT(common->control_head_ptr != 0);
1238    end  *needs_control_head = TRUE;
1239  } status;  #else
1240    *needs_control_head = FALSE;
1241  status = save ? start : loop;  #endif
 stackptr = STACK(stackptr - 2);  
 stacktop = STACK(stacktop - 1);  
1242    
1243  if (!save)  if (ccend == NULL)
1244    {    {
1245    stackptr += sizeof(sljit_w);    ccend = bracketend(cc) - (1 + LINK_SIZE);
1246    if (stackptr < stacktop)    if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))
     {  
     OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);  
     stackptr += sizeof(sljit_w);  
     tmp1empty = FALSE;  
     }  
   if (stackptr < stacktop)  
1247      {      {
1248      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);      possessive = length = (common->capture_last_ptr != 0) ? 5 : 3;
1249      stackptr += sizeof(sljit_w);      /* This is correct regardless of common->capture_last_ptr. */
1250      tmp2empty = FALSE;      capture_last_found = TRUE;
1251      }      }
1252    /* The tmp1next must be TRUE in either way. */    cc = next_opcode(common, cc);
1253    }    }
1254    
1255  while (status != end)  SLJIT_ASSERT(cc != NULL);
1256    {  while (cc < ccend)
1257    count = 0;    switch(*cc)
   switch(status)  
1258      {      {
1259      case start:      case OP_SET_SOM:
1260      SLJIT_ASSERT(save);      SLJIT_ASSERT(common->has_set_som);
1261      count = 1;      stack_restore = TRUE;
1262      srcw[0] = RECURSIVE_HEAD;      if (!setsom_found)
1263      status = loop;        {
1264          length += 2;
1265          setsom_found = TRUE;
1266          }
1267        cc += 1;
1268      break;      break;
1269    
1270      case loop:      case OP_MARK:
1271      if (cc >= ccend)      case OP_PRUNE_ARG:
1272        case OP_THEN_ARG:
1273        SLJIT_ASSERT(common->mark_ptr != 0);
1274        stack_restore = TRUE;
1275        if (!setmark_found)
1276        {        {
1277        status = end;        length += 2;
1278        break;        setmark_found = TRUE;
1279        }        }
1280        if (common->control_head_ptr != 0)
1281          *needs_control_head = TRUE;
1282        cc += 1 + 2 + cc[1];
1283        break;
1284    
1285      switch(*cc)      case OP_RECURSE:
1286        stack_restore = TRUE;
1287        if (common->has_set_som && !setsom_found)
1288        {        {
1289        case OP_ASSERT:        length += 2;
1290        case OP_ASSERT_NOT:        setsom_found = TRUE;
1291        case OP_ASSERTBACK:        }
1292        case OP_ASSERTBACK_NOT:      if (common->mark_ptr != 0 && !setmark_found)
1293        case OP_ONCE:        {
1294        case OP_BRAPOS:        length += 2;
1295        case OP_SBRA:        setmark_found = TRUE;
1296        case OP_SBRAPOS:        }
1297        case OP_SCOND:      if (common->capture_last_ptr != 0 && !capture_last_found)
1298        count = 1;        {
1299        srcw[0] = PRIV(cc);        length += 2;
1300        SLJIT_ASSERT(srcw[0] != 0);        capture_last_found = TRUE;
       cc += 1 + LINK_SIZE;  
       break;  
   
       case OP_CBRA:  
       case OP_SCBRA:  
       count = 1;  
       srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));  
       cc += 1 + LINK_SIZE + 2;  
       break;  
   
       case OP_CBRAPOS:  
       case OP_SCBRAPOS:  
       count = 2;  
       srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));  
       srcw[0] = PRIV(cc);  
       SLJIT_ASSERT(srcw[0] != 0);  
       cc += 1 + LINK_SIZE + 2;  
       break;  
   
       case OP_COND:  
       /* Might be a hidden SCOND. */  
       alternative = cc + GET(cc, 1);  
       if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)  
         {  
         count = 1;  
         srcw[0] = PRIV(cc);  
         SLJIT_ASSERT(srcw[0] != 0);  
         }  
       cc += 1 + LINK_SIZE;  
       break;  
   
       default:  
       cc = next_opcode(common, cc);  
       SLJIT_ASSERT(cc != NULL);  
       break;  
1301        }        }
1302        cc += 1 + LINK_SIZE;
1303      break;      break;
1304    
1305      case end:      case OP_CBRA:
1306      SLJIT_ASSERT_STOP();      case OP_CBRAPOS:
1307        case OP_SCBRA:
1308        case OP_SCBRAPOS:
1309        stack_restore = TRUE;
1310        if (common->capture_last_ptr != 0 && !capture_last_found)
1311          {
1312          length += 2;
1313          capture_last_found = TRUE;
1314          }
1315        length += 3;
1316        cc += 1 + LINK_SIZE + IMM2_SIZE;
1317        break;
1318    
1319        default:
1320        stack_restore = TRUE;
1321        /* Fall through. */
1322    
1323        case OP_NOT_WORD_BOUNDARY:
1324        case OP_WORD_BOUNDARY:
1325        case OP_NOT_DIGIT:
1326        case OP_DIGIT:
1327        case OP_NOT_WHITESPACE:
1328        case OP_WHITESPACE:
1329        case OP_NOT_WORDCHAR:
1330        case OP_WORDCHAR:
1331        case OP_ANY:
1332        case OP_ALLANY:
1333        case OP_ANYBYTE:
1334        case OP_NOTPROP:
1335        case OP_PROP:
1336        case OP_ANYNL:
1337        case OP_NOT_HSPACE:
1338        case OP_HSPACE:
1339        case OP_NOT_VSPACE:
1340        case OP_VSPACE:
1341        case OP_EXTUNI:
1342        case OP_EODN:
1343        case OP_EOD:
1344        case OP_CIRC:
1345        case OP_CIRCM:
1346        case OP_DOLL:
1347        case OP_DOLLM:
1348        case OP_CHAR:
1349        case OP_CHARI:
1350        case OP_NOT:
1351        case OP_NOTI:
1352    
1353        case OP_EXACT:
1354        case OP_POSSTAR:
1355        case OP_POSPLUS:
1356        case OP_POSQUERY:
1357        case OP_POSUPTO:
1358    
1359        case OP_EXACTI:
1360        case OP_POSSTARI:
1361        case OP_POSPLUSI:
1362        case OP_POSQUERYI:
1363        case OP_POSUPTOI:
1364    
1365        case OP_NOTEXACT:
1366        case OP_NOTPOSSTAR:
1367        case OP_NOTPOSPLUS:
1368        case OP_NOTPOSQUERY:
1369        case OP_NOTPOSUPTO:
1370    
1371        case OP_NOTEXACTI:
1372        case OP_NOTPOSSTARI:
1373        case OP_NOTPOSPLUSI:
1374        case OP_NOTPOSQUERYI:
1375        case OP_NOTPOSUPTOI:
1376    
1377        case OP_TYPEEXACT:
1378        case OP_TYPEPOSSTAR:
1379        case OP_TYPEPOSPLUS:
1380        case OP_TYPEPOSQUERY:
1381        case OP_TYPEPOSUPTO:
1382    
1383        case OP_CLASS:
1384        case OP_NCLASS:
1385        case OP_XCLASS:
1386    
1387        cc = next_opcode(common, cc);
1388        SLJIT_ASSERT(cc != NULL);
1389      break;      break;
1390      }      }
1391    
1392    while (count > 0)  /* Possessive quantifiers can use a special case. */
1393    if (SLJIT_UNLIKELY(possessive == length))
1394      return stack_restore ? no_frame : no_stack;
1395    
1396    if (length > 0)
1397      return length + 1;
1398    return stack_restore ? no_frame : no_stack;
1399    }
1400    
1401    static void init_frame(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, int stackpos, int stacktop, BOOL recursive)
1402    {
1403    DEFINE_COMPILER;
1404    BOOL setsom_found = recursive;
1405    BOOL setmark_found = recursive;
1406    /* The last capture is a local variable even for recursions. */
1407    BOOL capture_last_found = FALSE;
1408    int offset;
1409    
1410    /* >= 1 + shortest item size (2) */
1411    SLJIT_UNUSED_ARG(stacktop);
1412    SLJIT_ASSERT(stackpos >= stacktop + 2);
1413    
1414    stackpos = STACK(stackpos);
1415    if (ccend == NULL)
1416      {
1417      ccend = bracketend(cc) - (1 + LINK_SIZE);
1418      if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))
1419        cc = next_opcode(common, cc);
1420      }
1421    
1422    SLJIT_ASSERT(cc != NULL);
1423    while (cc < ccend)
1424      switch(*cc)
1425      {      {
1426      count--;      case OP_SET_SOM:
1427      if (save)      SLJIT_ASSERT(common->has_set_som);
1428        if (!setsom_found)
1429        {        {
1430        if (tmp1next)        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(0));
1431          {        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1432          if (!tmp1empty)        stackpos += (int)sizeof(sljit_sw);
1433            {        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1434            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        stackpos += (int)sizeof(sljit_sw);
1435            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;  
         }  
1436        }        }
1437      else      cc += 1;
1438        break;
1439    
1440        case OP_MARK:
1441        case OP_PRUNE_ARG:
1442        case OP_THEN_ARG:
1443        SLJIT_ASSERT(common->mark_ptr != 0);
1444        if (!setmark_found)
1445          {
1446          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->mark_ptr);
1447          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
1448          stackpos += (int)sizeof(sljit_sw);
1449          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1450          stackpos += (int)sizeof(sljit_sw);
1451          setmark_found = TRUE;
1452          }
1453        cc += 1 + 2 + cc[1];
1454        break;
1455    
1456        case OP_RECURSE:
1457        if (common->has_set_som && !setsom_found)
1458        {        {
1459        if (tmp1next)        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(0));
1460          {        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1461          SLJIT_ASSERT(!tmp1empty);        stackpos += (int)sizeof(sljit_sw);
1462          OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count], TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1463          tmp1empty = stackptr >= stacktop;        stackpos += (int)sizeof(sljit_sw);
1464          if (!tmp1empty)        setsom_found = TRUE;
1465            {        }
1466            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);      if (common->mark_ptr != 0 && !setmark_found)
1467            stackptr += sizeof(sljit_w);        {
1468            }        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->mark_ptr);
1469          tmp1next = FALSE;        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
1470          }        stackpos += (int)sizeof(sljit_sw);
1471        else        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1472          {        stackpos += (int)sizeof(sljit_sw);
1473          SLJIT_ASSERT(!tmp2empty);        setmark_found = TRUE;
1474          OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count], TMP2, 0);        }
1475          tmp2empty = stackptr >= stacktop;      if (common->capture_last_ptr != 0 && !capture_last_found)
1476          if (!tmp2empty)        {
1477            {        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->capture_last_ptr);
1478            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1479            stackptr += sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1480            }        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1481          tmp1next = TRUE;        stackpos += (int)sizeof(sljit_sw);
1482          }        capture_last_found = TRUE;
1483          }
1484        cc += 1 + LINK_SIZE;
1485        break;
1486    
1487        case OP_CBRA:
1488        case OP_CBRAPOS:
1489        case OP_SCBRA:
1490        case OP_SCBRAPOS:
1491        if (common->capture_last_ptr != 0 && !capture_last_found)
1492          {
1493          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->capture_last_ptr);
1494          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1495          stackpos += (int)sizeof(sljit_sw);
1496          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1497          stackpos += (int)sizeof(sljit_sw);
1498          capture_last_found = TRUE;
1499          }
1500        offset = (GET2(cc, 1 + LINK_SIZE)) << 1;
1501        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));
1502        stackpos += (int)sizeof(sljit_sw);
1503        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(offset));
1504        OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(offset + 1));
1505        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1506        stackpos += (int)sizeof(sljit_sw);
1507        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);
1508        stackpos += (int)sizeof(sljit_sw);
1509    
1510        cc += 1 + LINK_SIZE + IMM2_SIZE;
1511        break;
1512    
1513        default:
1514        cc = next_opcode(common, cc);
1515        SLJIT_ASSERT(cc != NULL);
1516        break;
1517        }
1518    
1519    OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, 0);
1520    SLJIT_ASSERT(stackpos == STACK(stacktop));
1521    }
1522    
1523    static SLJIT_INLINE int get_private_data_copy_length(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, BOOL needs_control_head)
1524    {
1525    int private_data_length = needs_control_head ? 3 : 2;
1526    int size;
1527    pcre_uchar *alternative;
1528    /* Calculate the sum of the private machine words. */
1529    while (cc < ccend)
1530      {
1531      size = 0;
1532      switch(*cc)
1533        {
1534        case OP_KET:
1535        if (PRIVATE_DATA(cc) != 0)
1536          {
1537          private_data_length++;
1538          SLJIT_ASSERT(PRIVATE_DATA(cc + 1) != 0);
1539          cc += PRIVATE_DATA(cc + 1);
1540        }        }
1541        cc += 1 + LINK_SIZE;
1542        break;
1543    
1544        case OP_ASSERT:
1545        case OP_ASSERT_NOT:
1546        case OP_ASSERTBACK:
1547        case OP_ASSERTBACK_NOT:
1548        case OP_ONCE:
1549        case OP_ONCE_NC:
1550        case OP_BRAPOS:
1551        case OP_SBRA:
1552        case OP_SBRAPOS:
1553        case OP_SCOND:
1554        private_data_length++;
1555        SLJIT_ASSERT(PRIVATE_DATA(cc) != 0);
1556        cc += 1 + LINK_SIZE;
1557        break;
1558    
1559        case OP_CBRA:
1560        case OP_SCBRA:
1561        if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1562          private_data_length++;
1563        cc += 1 + LINK_SIZE + IMM2_SIZE;
1564        break;
1565    
1566        case OP_CBRAPOS:
1567        case OP_SCBRAPOS:
1568        private_data_length += 2;
1569        cc += 1 + LINK_SIZE + IMM2_SIZE;
1570        break;
1571    
1572        case OP_COND:
1573        /* Might be a hidden SCOND. */
1574        alternative = cc + GET(cc, 1);
1575        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1576          private_data_length++;
1577        cc += 1 + LINK_SIZE;
1578        break;
1579    
1580        CASE_ITERATOR_PRIVATE_DATA_1
1581        if (PRIVATE_DATA(cc))
1582          private_data_length++;
1583        cc += 2;
1584    #ifdef SUPPORT_UTF
1585        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1586    #endif
1587        break;
1588    
1589        CASE_ITERATOR_PRIVATE_DATA_2A
1590        if (PRIVATE_DATA(cc))
1591          private_data_length += 2;
1592        cc += 2;
1593    #ifdef SUPPORT_UTF
1594        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1595    #endif
1596        break;
1597    
1598        CASE_ITERATOR_PRIVATE_DATA_2B
1599        if (PRIVATE_DATA(cc))
1600          private_data_length += 2;
1601        cc += 2 + IMM2_SIZE;
1602    #ifdef SUPPORT_UTF
1603        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1604    #endif
1605        break;
1606    
1607        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1608        if (PRIVATE_DATA(cc))
1609          private_data_length++;
1610        cc += 1;
1611        break;
1612    
1613        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1614        if (PRIVATE_DATA(cc))
1615          private_data_length += 2;
1616        cc += 1;
1617        break;
1618    
1619        CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1620        if (PRIVATE_DATA(cc))
1621          private_data_length += 2;
1622        cc += 1 + IMM2_SIZE;
1623        break;
1624    
1625        case OP_CLASS:
1626        case OP_NCLASS:
1627    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1628        case OP_XCLASS:
1629        size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
1630    #else
1631        size = 1 + 32 / (int)sizeof(pcre_uchar);
1632    #endif
1633        if (PRIVATE_DATA(cc))
1634          private_data_length += get_class_iterator_size(cc + size);
1635        cc += size;
1636        break;
1637    
1638        default:
1639        cc = next_opcode(common, cc);
1640        SLJIT_ASSERT(cc != NULL);
1641        break;
1642        }
1643      }
1644    SLJIT_ASSERT(cc == ccend);
1645    return private_data_length;
1646    }
1647    
1648    static void copy_private_data(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend,
1649      BOOL save, int stackptr, int stacktop, BOOL needs_control_head)
1650    {
1651    DEFINE_COMPILER;
1652    int srcw[2];
1653    int count, size;
1654    BOOL tmp1next = TRUE;
1655    BOOL tmp1empty = TRUE;
1656    BOOL tmp2empty = TRUE;
1657    pcre_uchar *alternative;
1658    enum {
1659      start,
1660      loop,
1661      end
1662    } status;
1663    
1664    status = save ? start : loop;
1665    stackptr = STACK(stackptr - 2);
1666    stacktop = STACK(stacktop - 1);
1667    
1668    if (!save)
1669      {
1670      stackptr += (needs_control_head ? 2 : 1) * sizeof(sljit_sw);
1671      if (stackptr < stacktop)
1672        {
1673        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1674        stackptr += sizeof(sljit_sw);
1675        tmp1empty = FALSE;
1676        }
1677      if (stackptr < stacktop)
1678        {
1679        OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1680        stackptr += sizeof(sljit_sw);
1681        tmp2empty = FALSE;
1682        }
1683      /* The tmp1next must be TRUE in either way. */
1684      }
1685    
1686    do
1687      {
1688      count = 0;
1689      switch(status)
1690        {
1691        case start:
1692        SLJIT_ASSERT(save && common->recursive_head_ptr != 0);
1693        count = 1;
1694        srcw[0] = common->recursive_head_ptr;
1695        if (needs_control_head)
1696          {
1697          SLJIT_ASSERT(common->control_head_ptr != 0);
1698          count = 2;
1699          srcw[1] = common->control_head_ptr;
1700          }
1701        status = loop;
1702        break;
1703    
1704        case loop:
1705        if (cc >= ccend)
1706          {
1707          status = end;
1708          break;
1709          }
1710    
1711        switch(*cc)
1712          {
1713          case OP_KET:
1714          if (PRIVATE_DATA(cc) != 0)
1715            {
1716            count = 1;
1717            srcw[0] = PRIVATE_DATA(cc);
1718            SLJIT_ASSERT(PRIVATE_DATA(cc + 1) != 0);
1719            cc += PRIVATE_DATA(cc + 1);
1720            }
1721          cc += 1 + LINK_SIZE;
1722          break;
1723    
1724          case OP_ASSERT:
1725          case OP_ASSERT_NOT:
1726          case OP_ASSERTBACK:
1727          case OP_ASSERTBACK_NOT:
1728          case OP_ONCE:
1729          case OP_ONCE_NC:
1730          case OP_BRAPOS:
1731          case OP_SBRA:
1732          case OP_SBRAPOS:
1733          case OP_SCOND:
1734          count = 1;
1735          srcw[0] = PRIVATE_DATA(cc);
1736          SLJIT_ASSERT(srcw[0] != 0);
1737          cc += 1 + LINK_SIZE;
1738          break;
1739    
1740          case OP_CBRA:
1741          case OP_SCBRA:
1742          if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1743            {
1744            count = 1;
1745            srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1746            }
1747          cc += 1 + LINK_SIZE + IMM2_SIZE;
1748          break;
1749    
1750          case OP_CBRAPOS:
1751          case OP_SCBRAPOS:
1752          count = 2;
1753          srcw[0] = PRIVATE_DATA(cc);
1754          srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1755          SLJIT_ASSERT(srcw[0] != 0 && srcw[1] != 0);
1756          cc += 1 + LINK_SIZE + IMM2_SIZE;
1757          break;
1758    
1759          case OP_COND:
1760          /* Might be a hidden SCOND. */
1761          alternative = cc + GET(cc, 1);
1762          if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1763            {
1764            count = 1;
1765            srcw[0] = PRIVATE_DATA(cc);
1766            SLJIT_ASSERT(srcw[0] != 0);
1767            }
1768          cc += 1 + LINK_SIZE;
1769          break;
1770    
1771          CASE_ITERATOR_PRIVATE_DATA_1
1772          if (PRIVATE_DATA(cc))
1773            {
1774            count = 1;
1775            srcw[0] = PRIVATE_DATA(cc);
1776            }
1777          cc += 2;
1778    #ifdef SUPPORT_UTF
1779          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1780    #endif
1781          break;
1782    
1783          CASE_ITERATOR_PRIVATE_DATA_2A
1784          if (PRIVATE_DATA(cc))
1785            {
1786            count = 2;
1787            srcw[0] = PRIVATE_DATA(cc);
1788            srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1789            }
1790          cc += 2;
1791    #ifdef SUPPORT_UTF
1792          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1793    #endif
1794          break;
1795    
1796          CASE_ITERATOR_PRIVATE_DATA_2B
1797          if (PRIVATE_DATA(cc))
1798            {
1799            count = 2;
1800            srcw[0] = PRIVATE_DATA(cc);
1801            srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1802            }
1803          cc += 2 + IMM2_SIZE;
1804    #ifdef SUPPORT_UTF
1805          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1806    #endif
1807          break;
1808    
1809          CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1810          if (PRIVATE_DATA(cc))
1811            {
1812            count = 1;
1813            srcw[0] = PRIVATE_DATA(cc);
1814            }
1815          cc += 1;
1816          break;
1817    
1818          CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1819          if (PRIVATE_DATA(cc))
1820            {
1821            count = 2;
1822            srcw[0] = PRIVATE_DATA(cc);
1823            srcw[1] = srcw[0] + sizeof(sljit_sw);
1824            }
1825          cc += 1;
1826          break;
1827    
1828          CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1829          if (PRIVATE_DATA(cc))
1830            {
1831            count = 2;
1832            srcw[0] = PRIVATE_DATA(cc);
1833            srcw[1] = srcw[0] + sizeof(sljit_sw);
1834            }
1835          cc += 1 + IMM2_SIZE;
1836          break;
1837    
1838          case OP_CLASS:
1839          case OP_NCLASS:
1840    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1841          case OP_XCLASS:
1842          size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
1843    #else
1844          size = 1 + 32 / (int)sizeof(pcre_uchar);
1845    #endif
1846          if (PRIVATE_DATA(cc))
1847            switch(get_class_iterator_size(cc + size))
1848              {
1849              case 1:
1850              count = 1;
1851              srcw[0] = PRIVATE_DATA(cc);
1852              break;
1853    
1854              case 2:
1855              count = 2;
1856              srcw[0] = PRIVATE_DATA(cc);
1857              srcw[1] = srcw[0] + sizeof(sljit_sw);
1858              break;
1859    
1860              default:
1861              SLJIT_ASSERT_STOP();
1862              break;
1863              }
1864          cc += size;
1865          break;
1866    
1867          default:
1868          cc = next_opcode(common, cc);
1869          SLJIT_ASSERT(cc != NULL);
1870          break;
1871          }
1872        break;
1873    
1874        case end:
1875        SLJIT_ASSERT_STOP();
1876        break;
1877        }
1878    
1879      while (count > 0)
1880        {
1881        count--;
1882        if (save)
1883          {
1884          if (tmp1next)
1885            {
1886            if (!tmp1empty)
1887              {
1888              OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1889              stackptr += sizeof(sljit_sw);
1890              }
1891            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), srcw[count]);
1892            tmp1empty = FALSE;
1893            tmp1next = FALSE;
1894            }
1895          else
1896            {
1897            if (!tmp2empty)
1898              {
1899              OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1900              stackptr += sizeof(sljit_sw);
1901              }
1902            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_SP), srcw[count]);
1903            tmp2empty = FALSE;
1904            tmp1next = TRUE;
1905            }
1906          }
1907        else
1908          {
1909          if (tmp1next)
1910            {
1911            SLJIT_ASSERT(!tmp1empty);
1912            OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), srcw[count], TMP1, 0);
1913            tmp1empty = stackptr >= stacktop;
1914            if (!tmp1empty)
1915              {
1916              OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1917              stackptr += sizeof(sljit_sw);
1918              }
1919            tmp1next = FALSE;
1920            }
1921          else
1922            {
1923            SLJIT_ASSERT(!tmp2empty);
1924            OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), srcw[count], TMP2, 0);
1925            tmp2empty = stackptr >= stacktop;
1926            if (!tmp2empty)
1927              {
1928              OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1929              stackptr += sizeof(sljit_sw);
1930              }
1931            tmp1next = TRUE;
1932            }
1933          }
1934        }
1935      }
1936    while (status != end);
1937    
1938    if (save)
1939      {
1940      if (tmp1next)
1941        {
1942        if (!tmp1empty)
1943          {
1944          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1945          stackptr += sizeof(sljit_sw);
1946          }
1947        if (!tmp2empty)
1948          {
1949          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1950          stackptr += sizeof(sljit_sw);
1951          }
1952        }
1953      else
1954        {
1955        if (!tmp2empty)
1956          {
1957          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1958          stackptr += sizeof(sljit_sw);
1959          }
1960        if (!tmp1empty)
1961          {
1962          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1963          stackptr += sizeof(sljit_sw);
1964          }
1965        }
1966      }
1967    SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));
1968    }
1969    
1970    static SLJIT_INLINE pcre_uchar *set_then_offsets(compiler_common *common, pcre_uchar *cc, pcre_uint8 *current_offset)
1971    {
1972    pcre_uchar *end = bracketend(cc);
1973    BOOL has_alternatives = cc[GET(cc, 1)] == OP_ALT;
1974    
1975    /* Assert captures then. */
1976    if (*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT)
1977      current_offset = NULL;
1978    /* Conditional block does not. */
1979    if (*cc == OP_COND || *cc == OP_SCOND)
1980      has_alternatives = FALSE;
1981    
1982    cc = next_opcode(common, cc);
1983    if (has_alternatives)
1984      current_offset = common->then_offsets + (cc - common->start);
1985    
1986    while (cc < end)
1987      {
1988      if ((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND))
1989        cc = set_then_offsets(common, cc, current_offset);
1990      else
1991        {
1992        if (*cc == OP_ALT && has_alternatives)
1993          current_offset = common->then_offsets + (cc + 1 + LINK_SIZE - common->start);
1994        if (*cc >= OP_THEN && *cc <= OP_THEN_ARG && current_offset != NULL)
1995          *current_offset = 1;
1996        cc = next_opcode(common, cc);
1997        }
1998      }
1999    
2000    return end;
2001    }
2002    
2003    #undef CASE_ITERATOR_PRIVATE_DATA_1
2004    #undef CASE_ITERATOR_PRIVATE_DATA_2A
2005    #undef CASE_ITERATOR_PRIVATE_DATA_2B
2006    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_1
2007    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
2008    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
2009    
2010    static SLJIT_INLINE BOOL is_powerof2(unsigned int value)
2011    {
2012    return (value & (value - 1)) == 0;
2013    }
2014    
2015    static SLJIT_INLINE void set_jumps(jump_list *list, struct sljit_label *label)
2016    {
2017    while (list)
2018      {
2019      /* sljit_set_label is clever enough to do nothing
2020      if either the jump or the label is NULL. */
2021      SET_LABEL(list->jump, label);
2022      list = list->next;
2023      }
2024    }
2025    
2026    static SLJIT_INLINE void add_jump(struct sljit_compiler *compiler, jump_list **list, struct sljit_jump *jump)
2027    {
2028    jump_list *list_item = sljit_alloc_memory(compiler, sizeof(jump_list));
2029    if (list_item)
2030      {
2031      list_item->next = *list;
2032      list_item->jump = jump;
2033      *list = list_item;
2034      }
2035    }
2036    
2037    static void add_stub(compiler_common *common, struct sljit_jump *start)
2038    {
2039    DEFINE_COMPILER;
2040    stub_list *list_item = sljit_alloc_memory(compiler, sizeof(stub_list));
2041    
2042    if (list_item)
2043      {
2044      list_item->start = start;
2045      list_item->quit = LABEL();
2046      list_item->next = common->stubs;
2047      common->stubs = list_item;
2048      }
2049    }
2050    
2051    static void flush_stubs(compiler_common *common)
2052    {
2053    DEFINE_COMPILER;
2054    stub_list *list_item = common->stubs;
2055    
2056    while (list_item)
2057      {
2058      JUMPHERE(list_item->start);
2059      add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));
2060      JUMPTO(SLJIT_JUMP, list_item->quit);
2061      list_item = list_item->next;
2062      }
2063    common->stubs = NULL;
2064    }
2065    
2066    static void add_label_addr(compiler_common *common, sljit_uw *update_addr)
2067    {
2068    DEFINE_COMPILER;
2069    label_addr_list *label_addr;
2070    
2071    label_addr = sljit_alloc_memory(compiler, sizeof(label_addr_list));
2072    if (label_addr == NULL)
2073      return;
2074    label_addr->label = LABEL();
2075    label_addr->update_addr = update_addr;
2076    label_addr->next = common->label_addrs;
2077    common->label_addrs = label_addr;
2078    }
2079    
2080    static SLJIT_INLINE void count_match(compiler_common *common)
2081    {
2082    DEFINE_COMPILER;
2083    
2084    OP2(SLJIT_SUB | SLJIT_SET_E, COUNT_MATCH, 0, COUNT_MATCH, 0, SLJIT_IMM, 1);
2085    add_jump(compiler, &common->calllimit, JUMP(SLJIT_ZERO));
2086    }
2087    
2088    static SLJIT_INLINE void allocate_stack(compiler_common *common, int size)
2089    {
2090    /* May destroy all locals and registers except TMP2. */
2091    DEFINE_COMPILER;
2092    
2093    OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_sw));
2094    #ifdef DESTROY_REGISTERS
2095    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);
2096    OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);
2097    OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);
2098    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), LOCALS0, TMP1, 0);
2099    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), LOCALS1, TMP1, 0);
2100    #endif
2101    add_stub(common, CMP(SLJIT_GREATER, STACK_TOP, 0, STACK_LIMIT, 0));
2102    }
2103    
2104    static SLJIT_INLINE void free_stack(compiler_common *common, int size)
2105    {
2106    DEFINE_COMPILER;
2107    OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_sw));
2108    }
2109    
2110    static sljit_uw * allocate_read_only_data(compiler_common *common, sljit_uw size)
2111    {
2112    DEFINE_COMPILER;
2113    sljit_uw *result;
2114    
2115    if (SLJIT_UNLIKELY(sljit_get_compiler_error(compiler)))
2116      return NULL;
2117    
2118    result = (sljit_uw *)SLJIT_MALLOC(size + sizeof(sljit_uw), compiler->allocator_data);
2119    if (SLJIT_UNLIKELY(result == NULL))
2120      {
2121      sljit_set_compiler_memory_error(compiler);
2122      return NULL;
2123      }
2124    
2125    *(void**)result = common->read_only_data_head;
2126    common->read_only_data_head = (void *)result;
2127    return result + 1;
2128    }
2129    
2130    static void free_read_only_data(void *current, void *allocator_data)
2131    {
2132    void *next;
2133    
2134    SLJIT_UNUSED_ARG(allocator_data);
2135    
2136    while (current != NULL)
2137      {
2138      next = *(void**)current;
2139      SLJIT_FREE(current, allocator_data);
2140      current = next;
2141      }
2142    }
2143    
2144    static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)
2145    {
2146    DEFINE_COMPILER;
2147    struct sljit_label *loop;
2148    int i;
2149    
2150    /* At this point we can freely use all temporary registers. */
2151    SLJIT_ASSERT(length > 1);
2152    /* TMP1 returns with begin - 1. */
2153    OP2(SLJIT_SUB, SLJIT_R0, 0, SLJIT_MEM1(SLJIT_S0), SLJIT_OFFSETOF(jit_arguments, begin), SLJIT_IMM, IN_UCHARS(1));
2154    if (length < 8)
2155      {
2156      for (i = 1; i < length; i++)
2157        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), OVECTOR(i), SLJIT_R0, 0);
2158      }
2159    else
2160      {
2161      GET_LOCAL_BASE(SLJIT_R1, 0, OVECTOR_START);
2162      OP1(SLJIT_MOV, SLJIT_R2, 0, SLJIT_IMM, length - 1);
2163      loop = LABEL();
2164      OP1(SLJIT_MOVU, SLJIT_MEM1(SLJIT_R1), sizeof(sljit_sw), SLJIT_R0, 0);
2165      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_IMM, 1);
2166      JUMPTO(SLJIT_NOT_ZERO, loop);
2167      }
2168    }
2169    
2170    static SLJIT_INLINE void do_reset_match(compiler_common *common, int length)
2171    {
2172    DEFINE_COMPILER;
2173    struct sljit_label *loop;
2174    int i;
2175    
2176    SLJIT_ASSERT(length > 1);
2177    /* OVECTOR(1) contains the "string begin - 1" constant. */
2178    if (length > 2)
2179      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(1));
2180    if (length < 8)
2181      {
2182      for (i = 2; i < length; i++)
2183        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), OVECTOR(i), TMP1, 0);
2184      }
2185    else
2186      {
2187      GET_LOCAL_BASE(TMP2, 0, OVECTOR_START + sizeof(sljit_sw));
2188      OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_IMM, length - 2);
2189      loop = LABEL();
2190      OP1(SLJIT_MOVU, SLJIT_MEM1(TMP2), sizeof(sljit_sw), TMP1, 0);
2191      OP2(SLJIT_SUB | SLJIT_SET_E, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, 1);
2192      JUMPTO(SLJIT_NOT_ZERO, loop);
2193      }
2194    
2195    OP1(SLJIT_MOV, STACK_TOP, 0, ARGUMENTS, 0);
2196    if (common->mark_ptr != 0)
2197      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->mark_ptr, SLJIT_IMM, 0);
2198    if (common->control_head_ptr != 0)
2199      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->control_head_ptr, SLJIT_IMM, 0);
2200    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(jit_arguments, stack));
2201    OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->start_ptr);
2202    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(struct sljit_stack, base));
2203    }
2204    
2205    static sljit_sw SLJIT_CALL do_search_mark(sljit_sw *current, const pcre_uchar *skip_arg)
2206    {
2207    while (current != NULL)
2208      {
2209      switch (current[-2])
2210        {
2211        case type_then_trap:
2212        break;
2213    
2214        case type_mark:
2215        if (STRCMP_UC_UC(skip_arg, (pcre_uchar *)current[-3]) == 0)
2216          return current[-4];
2217        break;
2218    
2219        default:
2220        SLJIT_ASSERT_STOP();
2221        break;
2222        }
2223      current = (sljit_sw*)current[-1];
2224      }
2225    return -1;
2226    }
2227    
2228    static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)
2229    {
2230    DEFINE_COMPILER;
2231    struct sljit_label *loop;
2232    struct sljit_jump *early_quit;
2233    
2234    /* At this point we can freely use all registers. */
2235    OP1(SLJIT_MOV, SLJIT_S2, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(1));
2236    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), OVECTOR(1), STR_PTR, 0);
2237    
2238    OP1(SLJIT_MOV, SLJIT_R0, 0, ARGUMENTS, 0);
2239    if (common->mark_ptr != 0)
2240      OP1(SLJIT_MOV, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_SP), common->mark_ptr);
2241    OP1(SLJIT_MOV_SI, SLJIT_R1, 0, SLJIT_MEM1(SLJIT_R0), SLJIT_OFFSETOF(jit_arguments, offset_count));
2242    if (common->mark_ptr != 0)
2243      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_R0), SLJIT_OFFSETOF(jit_arguments, mark_ptr), SLJIT_R2, 0);
2244    OP2(SLJIT_SUB, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_R0), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));
2245    OP1(SLJIT_MOV, SLJIT_R0, 0, SLJIT_MEM1(SLJIT_R0), SLJIT_OFFSETOF(jit_arguments, begin));
2246    GET_LOCAL_BASE(SLJIT_S0, 0, OVECTOR_START);
2247    /* Unlikely, but possible */
2248    early_quit = CMP(SLJIT_EQUAL, SLJIT_R1, 0, SLJIT_IMM, 0);
2249    loop = LABEL();
2250    OP2(SLJIT_SUB, SLJIT_S1, 0, SLJIT_MEM1(SLJIT_S0), 0, SLJIT_R0, 0);
2251    OP2(SLJIT_ADD, SLJIT_S0, 0, SLJIT_S0, 0, SLJIT_IMM, sizeof(sljit_sw));
2252    /* Copy the integer value to the output buffer */
2253    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2254    OP2(SLJIT_ASHR, SLJIT_S1, 0, SLJIT_S1, 0, SLJIT_IMM, UCHAR_SHIFT);
2255    #endif
2256    OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_R2), sizeof(int), SLJIT_S1, 0);
2257    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_R1, 0, SLJIT_R1, 0, SLJIT_IMM, 1);
2258    JUMPTO(SLJIT_NOT_ZERO, loop);
2259    JUMPHERE(early_quit);
2260    
2261    /* Calculate the return value, which is the maximum ovector value. */
2262    if (topbracket > 1)
2263      {
2264      GET_LOCAL_BASE(SLJIT_R0, 0, OVECTOR_START + topbracket * 2 * sizeof(sljit_sw));
2265      OP1(SLJIT_MOV, SLJIT_R1, 0, SLJIT_IMM, topbracket + 1);
2266    
2267      /* OVECTOR(0) is never equal to SLJIT_S2. */
2268      loop = LABEL();
2269      OP1(SLJIT_MOVU, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_R0), -(2 * (sljit_sw)sizeof(sljit_sw)));
2270      OP2(SLJIT_SUB, SLJIT_R1, 0, SLJIT_R1, 0, SLJIT_IMM, 1);
2271      CMPTO(SLJIT_EQUAL, SLJIT_R2, 0, SLJIT_S2, 0, loop);
2272      OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_R1, 0);
2273      }
2274    else
2275      OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, 1);
2276    }
2277    
2278    static SLJIT_INLINE void return_with_partial_match(compiler_common *common, struct sljit_label *quit)
2279    {
2280    DEFINE_COMPILER;
2281    struct sljit_jump *jump;
2282    
2283    SLJIT_COMPILE_ASSERT(STR_END == SLJIT_S1, str_end_must_be_saved_reg2);
2284    SLJIT_ASSERT(common->start_used_ptr != 0 && common->start_ptr != 0
2285      && (common->mode == JIT_PARTIAL_SOFT_COMPILE ? common->hit_start != 0 : common->hit_start == 0));
2286    
2287    OP1(SLJIT_MOV, SLJIT_R1, 0, ARGUMENTS, 0);
2288    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, PCRE_ERROR_PARTIAL);
2289    OP1(SLJIT_MOV_SI, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_R1), SLJIT_OFFSETOF(jit_arguments, real_offset_count));
2290    CMPTO(SLJIT_SIG_LESS, SLJIT_R2, 0, SLJIT_IMM, 2, quit);
2291    
2292    /* Store match begin and end. */
2293    OP1(SLJIT_MOV, SLJIT_S0, 0, SLJIT_MEM1(SLJIT_R1), SLJIT_OFFSETOF(jit_arguments, begin));
2294    OP1(SLJIT_MOV, SLJIT_R1, 0, SLJIT_MEM1(SLJIT_R1), SLJIT_OFFSETOF(jit_arguments, offsets));
2295    
2296    jump = CMP(SLJIT_SIG_LESS, SLJIT_R2, 0, SLJIT_IMM, 3);
2297    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);
2298    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2299    OP2(SLJIT_ASHR, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_IMM, UCHAR_SHIFT);
2300    #endif
2301    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_R1), 2 * sizeof(int), SLJIT_R2, 0);
2302    JUMPHERE(jump);
2303    
2304    OP1(SLJIT_MOV, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_SP), common->mode == JIT_PARTIAL_HARD_COMPILE ? common->start_used_ptr : common->hit_start);
2305    OP2(SLJIT_SUB, SLJIT_S1, 0, STR_END, 0, SLJIT_S0, 0);
2306    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2307    OP2(SLJIT_ASHR, SLJIT_S1, 0, SLJIT_S1, 0, SLJIT_IMM, UCHAR_SHIFT);
2308    #endif
2309    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_R1), sizeof(int), SLJIT_S1, 0);
2310    
2311    OP2(SLJIT_SUB, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_S0, 0);
2312    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2313    OP2(SLJIT_ASHR, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_IMM, UCHAR_SHIFT);
2314    #endif
2315    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_R1), 0, SLJIT_R2, 0);
2316    
2317    JUMPTO(SLJIT_JUMP, quit);
2318    }
2319    
2320    static SLJIT_INLINE void check_start_used_ptr(compiler_common *common)
2321    {
2322    /* May destroy TMP1. */
2323    DEFINE_COMPILER;
2324    struct sljit_jump *jump;
2325    
2326    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2327      {
2328      /* The value of -1 must be kept for start_used_ptr! */
2329      OP2(SLJIT_ADD, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, SLJIT_IMM, 1);
2330      /* Jumps if start_used_ptr < STR_PTR, or start_used_ptr == -1. Although overwriting
2331      is not necessary if start_used_ptr == STR_PTR, it does not hurt as well. */
2332      jump = CMP(SLJIT_LESS_EQUAL, TMP1, 0, STR_PTR, 0);
2333      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0);
2334      JUMPHERE(jump);
2335      }
2336    else if (common->mode == JIT_PARTIAL_HARD_COMPILE)
2337      {
2338      jump = CMP(SLJIT_LESS_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0);
2339      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0);
2340      JUMPHERE(jump);
2341      }
2342    }
2343    
2344    static SLJIT_INLINE BOOL char_has_othercase(compiler_common *common, pcre_uchar *cc)
2345    {
2346    /* Detects if the character has an othercase. */
2347    unsigned int c;
2348    
2349    #ifdef SUPPORT_UTF
2350    if (common->utf)
2351      {
2352      GETCHAR(c, cc);
2353      if (c > 127)
2354        {
2355    #ifdef SUPPORT_UCP
2356        return c != UCD_OTHERCASE(c);
2357    #else
2358        return FALSE;
2359    #endif
2360        }
2361    #ifndef COMPILE_PCRE8
2362      return common->fcc[c] != c;
2363    #endif
2364      }
2365    else
2366    #endif
2367      c = *cc;
2368    return MAX_255(c) ? common->fcc[c] != c : FALSE;
2369    }
2370    
2371    static SLJIT_INLINE unsigned int char_othercase(compiler_common *common, unsigned int c)
2372    {
2373    /* Returns with the othercase. */
2374    #ifdef SUPPORT_UTF
2375    if (common->utf && c > 127)
2376      {
2377    #ifdef SUPPORT_UCP
2378      return UCD_OTHERCASE(c);
2379    #else
2380      return c;
2381    #endif
2382      }
2383    #endif
2384    return TABLE_GET(c, common->fcc, c);
2385    }
2386    
2387    static unsigned int char_get_othercase_bit(compiler_common *common, pcre_uchar *cc)
2388    {
2389    /* Detects if the character and its othercase has only 1 bit difference. */
2390    unsigned int c, oc, bit;
2391    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2392    int n;
2393    #endif
2394    
2395    #ifdef SUPPORT_UTF
2396    if (common->utf)
2397      {
2398      GETCHAR(c, cc);
2399      if (c <= 127)
2400        oc = common->fcc[c];
2401      else
2402        {
2403    #ifdef SUPPORT_UCP
2404        oc = UCD_OTHERCASE(c);
2405    #else
2406        oc = c;
2407    #endif
2408        }
2409      }
2410    else
2411      {
2412      c = *cc;
2413      oc = TABLE_GET(c, common->fcc, c);
2414      }
2415    #else
2416    c = *cc;
2417    oc = TABLE_GET(c, common->fcc, c);
2418    #endif
2419    
2420    SLJIT_ASSERT(c != oc);
2421    
2422    bit = c ^ oc;
2423    /* Optimized for English alphabet. */
2424    if (c <= 127 && bit == 0x20)
2425      return (0 << 8) | 0x20;
2426    
2427    /* Since c != oc, they must have at least 1 bit difference. */
2428    if (!is_powerof2(bit))
2429      return 0;
2430    
2431    #if defined COMPILE_PCRE8
2432    
2433    #ifdef SUPPORT_UTF
2434    if (common->utf && c > 127)
2435      {
2436      n = GET_EXTRALEN(*cc);
2437      while ((bit & 0x3f) == 0)
2438        {
2439        n--;
2440        bit >>= 6;
2441        }
2442      return (n << 8) | bit;
2443      }
2444    #endif /* SUPPORT_UTF */
2445    return (0 << 8) | bit;
2446    
2447    #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2448    
2449    #ifdef SUPPORT_UTF
2450    if (common->utf && c > 65535)
2451      {
2452      if (bit >= (1 << 10))
2453        bit >>= 10;
2454      else
2455        return (bit < 256) ? ((2 << 8) | bit) : ((3 << 8) | (bit >> 8));
2456      }
2457    #endif /* SUPPORT_UTF */
2458    return (bit < 256) ? ((0 << 8) | bit) : ((1 << 8) | (bit >> 8));
2459    
2460    #endif /* COMPILE_PCRE[8|16|32] */
2461    }
2462    
2463    static void check_partial(compiler_common *common, BOOL force)
2464    {
2465    /* Checks whether a partial matching is occurred. Does not modify registers. */
2466    DEFINE_COMPILER;
2467    struct sljit_jump *jump = NULL;
2468    
2469    SLJIT_ASSERT(!force || common->mode != JIT_COMPILE);
2470    
2471    if (common->mode == JIT_COMPILE)
2472      return;
2473    
2474    if (!force)
2475      jump = CMP(SLJIT_GREATER_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0);
2476    else if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2477      jump = CMP(SLJIT_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, SLJIT_IMM, -1);
2478    
2479    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2480      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->hit_start, SLJIT_IMM, 0);
2481    else
2482      {
2483      if (common->partialmatchlabel != NULL)
2484        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2485      else
2486        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2487      }
2488    
2489    if (jump != NULL)
2490      JUMPHERE(jump);
2491    }
2492    
2493    static void check_str_end(compiler_common *common, jump_list **end_reached)
2494    {
2495    /* Does not affect registers. Usually used in a tight spot. */
2496    DEFINE_COMPILER;
2497    struct sljit_jump *jump;
2498    
2499    if (common->mode == JIT_COMPILE)
2500      {
2501      add_jump(compiler, end_reached, CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2502      return;
2503      }
2504    
2505    jump = CMP(SLJIT_LESS, STR_PTR, 0, STR_END, 0);
2506    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2507      {
2508      add_jump(compiler, end_reached, CMP(SLJIT_GREATER_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0));
2509      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->hit_start, SLJIT_IMM, 0);
2510      add_jump(compiler, end_reached, JUMP(SLJIT_JUMP));
2511      }
2512    else
2513      {
2514      add_jump(compiler, end_reached, CMP(SLJIT_GREATER_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0));
2515      if (common->partialmatchlabel != NULL)
2516        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2517      else
2518        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2519      }
2520    JUMPHERE(jump);
2521    }
2522    
2523    static void detect_partial_match(compiler_common *common, jump_list **backtracks)
2524    {
2525    DEFINE_COMPILER;
2526    struct sljit_jump *jump;
2527    
2528    if (common->mode == JIT_COMPILE)
2529      {
2530      add_jump(compiler, backtracks, CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2531      return;
2532      }
2533    
2534    /* Partial matching mode. */
2535    jump = CMP(SLJIT_LESS, STR_PTR, 0, STR_END, 0);
2536    add_jump(compiler, backtracks, CMP(SLJIT_GREATER_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0));
2537    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2538      {
2539      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->hit_start, SLJIT_IMM, 0);
2540      add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));
2541      }
2542    else
2543      {
2544      if (common->partialmatchlabel != NULL)
2545        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2546      else
2547        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2548      }
2549    JUMPHERE(jump);
2550    }
2551    
2552    static void peek_char(compiler_common *common, pcre_uint32 max)
2553    {
2554    /* Reads the character into TMP1, keeps STR_PTR.
2555    Does not check STR_END. TMP2 Destroyed. */
2556    DEFINE_COMPILER;
2557    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2558    struct sljit_jump *jump;
2559    #endif
2560    
2561    SLJIT_UNUSED_ARG(max);
2562    
2563    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2564    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2565    if (common->utf)
2566      {
2567      if (max < 128) return;
2568    
2569      jump = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2570      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2571      add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2572      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2573      JUMPHERE(jump);
2574      }
2575    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2576    
2577    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2578    if (common->utf)
2579      {
2580      if (max < 0xd800) return;
2581    
2582      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2583      jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2584      /* TMP2 contains the high surrogate. */
2585      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2586      OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2587      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2588      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2589      OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2590      JUMPHERE(jump);
2591      }
2592    #endif
2593    }
2594    
2595    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2596    
2597    static BOOL is_char7_bitset(const pcre_uint8 *bitset, BOOL nclass)
2598    {
2599    /* Tells whether the character codes below 128 are enough
2600    to determine a match. */
2601    const pcre_uint8 value = nclass ? 0xff : 0;
2602    const pcre_uint8 *end = bitset + 32;
2603    
2604    bitset += 16;
2605    do
2606      {
2607      if (*bitset++ != value)
2608        return FALSE;
2609      }
2610    while (bitset < end);
2611    return TRUE;
2612    }
2613    
2614    static void read_char7_type(compiler_common *common, BOOL full_read)
2615    {
2616    /* Reads the precise character type of a character into TMP1, if the character
2617    is less than 128. Otherwise it returns with zero. Does not check STR_END. The
2618    full_read argument tells whether characters above max are accepted or not. */
2619    DEFINE_COMPILER;
2620    struct sljit_jump *jump;
2621    
2622    SLJIT_ASSERT(common->utf);
2623    
2624    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2625    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2626    
2627    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2628    
2629    if (full_read)
2630      {
2631      jump = CMP(SLJIT_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2632      OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2633      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2634      JUMPHERE(jump);
2635      }
2636    }
2637    
2638    #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2639    
2640    static void read_char_range(compiler_common *common, pcre_uint32 min, pcre_uint32 max, BOOL update_str_ptr)
2641    {
2642    /* Reads the precise value of a character into TMP1, if the character is
2643    between min and max (c >= min && c <= max). Otherwise it returns with a value
2644    outside the range. Does not check STR_END. */
2645    DEFINE_COMPILER;
2646    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2647    struct sljit_jump *jump;
2648    #endif
2649    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2650    struct sljit_jump *jump2;
2651    #endif
2652    
2653    SLJIT_UNUSED_ARG(update_str_ptr);
2654    SLJIT_UNUSED_ARG(min);
2655    SLJIT_UNUSED_ARG(max);
2656    SLJIT_ASSERT(min <= max);
2657    
2658    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2659    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2660    
2661    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2662    if (common->utf)
2663      {
2664      if (max < 128 && !update_str_ptr) return;
2665    
2666      jump = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2667      if (min >= 0x10000)
2668        {
2669        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xf0);
2670        if (update_str_ptr)
2671          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2672        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2673        jump2 = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0x7);
2674        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2675        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2676        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2677        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2678        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2679        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2680        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2681        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2682        if (!update_str_ptr)
2683          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2684        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2685        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2686        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2687        JUMPHERE(jump2);
2688        if (update_str_ptr)
2689          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2690        }
2691      else if (min >= 0x800 && max <= 0xffff)
2692        {
2693        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xe0);
2694        if (update_str_ptr)
2695          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2696        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2697        jump2 = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xf);
2698        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2699        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2700        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2701        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2702        if (!update_str_ptr)
2703          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2704        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2705        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2706        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2707        JUMPHERE(jump2);
2708        if (update_str_ptr)
2709          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2710        }
2711      else if (max >= 0x800)
2712        add_jump(compiler, (max < 0x10000) ? &common->utfreadchar16 : &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2713      else if (max < 128)
2714        {
2715        OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2716        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2717        }
2718      else
2719        {
2720        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2721        if (!update_str_ptr)
2722          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2723        else
2724          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2725        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2726        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2727        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2728        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2729        if (update_str_ptr)
2730          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2731        }
2732      JUMPHERE(jump);
2733      }
2734    #endif
2735    
2736    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2737    if (common->utf)
2738      {
2739      if (max >= 0x10000)
2740        {
2741        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2742        jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2743        /* TMP2 contains the high surrogate. */
2744        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2745        OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2746        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2747        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2748        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2749        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2750        JUMPHERE(jump);
2751        return;
2752        }
2753    
2754      if (max < 0xd800 && !update_str_ptr) return;
2755    
2756      /* Skip low surrogate if necessary. */
2757      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2758      jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2759      if (update_str_ptr)
2760        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2761      if (max >= 0xd800)
2762        OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0x10000);
2763      JUMPHERE(jump);
2764      }
2765    #endif
2766    }
2767    
2768    static SLJIT_INLINE void read_char(compiler_common *common)
2769    {
2770    read_char_range(common, 0, READ_CHAR_MAX, TRUE);
2771    }
2772    
2773    static void read_char8_type(compiler_common *common, BOOL update_str_ptr)
2774    {
2775    /* Reads the character type into TMP1, updates STR_PTR. Does not check STR_END. */
2776    DEFINE_COMPILER;
2777    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2778    struct sljit_jump *jump;
2779    #endif
2780    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2781    struct sljit_jump *jump2;
2782    #endif
2783    
2784    SLJIT_UNUSED_ARG(update_str_ptr);
2785    
2786    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2787    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2788    
2789    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2790    if (common->utf)
2791      {
2792      /* This can be an extra read in some situations, but hopefully
2793      it is needed in most cases. */
2794      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2795      jump = CMP(SLJIT_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2796      if (!update_str_ptr)
2797        {
2798        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2799        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2800        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2801        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2802        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2803        OP2(SLJIT_OR, TMP2, 0, TMP2, 0, TMP1, 0);
2804        OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2805        jump2 = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 255);
2806        OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2807        JUMPHERE(jump2);
2808      }      }
2809      else
2810        add_jump(compiler, &common->utfreadtype8, JUMP(SLJIT_FAST_CALL));
2811      JUMPHERE(jump);
2812      return;
2813    }    }
2814    #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2815    
2816  if (save)  #if !defined COMPILE_PCRE8
2817    /* The ctypes array contains only 256 values. */
2818    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2819    jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 255);
2820    #endif
2821    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2822    #if !defined COMPILE_PCRE8
2823    JUMPHERE(jump);
2824    #endif
2825    
2826    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2827    if (common->utf && update_str_ptr)
2828    {    {
2829    if (tmp1next)    /* Skip low surrogate if necessary. */
2830      OP2(SLJIT_SUB, TMP2, 0, TMP2, 0, SLJIT_IMM, 0xd800);
2831      jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2832      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2833      JUMPHERE(jump);
2834      }
2835    #endif /* SUPPORT_UTF && COMPILE_PCRE16 */
2836    }
2837    
2838    static void skip_char_back(compiler_common *common)
2839    {
2840    /* Goes one character back. Affects STR_PTR and TMP1. Does not check begin. */
2841    DEFINE_COMPILER;
2842    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2843    #if defined COMPILE_PCRE8
2844    struct sljit_label *label;
2845    
2846    if (common->utf)
2847      {
2848      label = LABEL();
2849      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
2850      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2851      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xc0);
2852      CMPTO(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, 0x80, label);
2853      return;
2854      }
2855    #elif defined COMPILE_PCRE16
2856    if (common->utf)
2857      {
2858      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
2859      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2860      /* Skip low surrogate if necessary. */
2861      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
2862      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xdc00);
2863      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
2864      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
2865      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2866      return;
2867      }
2868    #endif /* COMPILE_PCRE[8|16] */
2869    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2870    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2871    }
2872    
2873    static void check_newlinechar(compiler_common *common, int nltype, jump_list **backtracks, BOOL jumpifmatch)
2874    {
2875    /* Character comes in TMP1. Checks if it is a newline. TMP2 may be destroyed. */
2876    DEFINE_COMPILER;
2877    struct sljit_jump *jump;
2878    
2879    if (nltype == NLTYPE_ANY)
2880      {
2881      add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));
2882      add_jump(compiler, backtracks, JUMP(jumpifmatch ? SLJIT_NOT_ZERO : SLJIT_ZERO));
2883      }
2884    else if (nltype == NLTYPE_ANYCRLF)
2885      {
2886      if (jumpifmatch)
2887      {      {
2888      if (!tmp1empty)      add_jump(compiler, backtracks, CMP(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR));
2889        {      add_jump(compiler, backtracks, CMP(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL));
       OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);  
       stackptr += sizeof(sljit_w);  
       }  
     if (!tmp2empty)  
       {  
       OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);  
       stackptr += sizeof(sljit_w);  
       }  
2890      }      }
2891    else    else
2892      {      {
2893      if (!tmp2empty)      jump = CMP(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR);
2894        {      add_jump(compiler, backtracks, CMP(SLJIT_NOT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL));
2895        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);      JUMPHERE(jump);
       stackptr += sizeof(sljit_w);  
       }  
     if (!tmp1empty)  
       {  
       OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);  
       stackptr += sizeof(sljit_w);  
       }  
2896      }      }
2897    }    }
2898  SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));  else
2899      {
2900      SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline < 256);
2901      add_jump(compiler, backtracks, CMP(jumpifmatch ? SLJIT_EQUAL : SLJIT_NOT_EQUAL, TMP1, 0, SLJIT_IMM, common->newline));
2902      }
2903  }  }
2904    
2905  static SLJIT_INLINE BOOL ispowerof2(unsigned int value)  #ifdef SUPPORT_UTF
2906    
2907    #if defined COMPILE_PCRE8
2908    static void do_utfreadchar(compiler_common *common)
2909  {  {
2910  return (value & (value - 1)) == 0;  /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2911    of the character (>= 0xc0). Return char value in TMP1, length in TMP2. */
2912    DEFINE_COMPILER;
2913    struct sljit_jump *jump;
2914    
2915    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2916    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2917    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2918    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2919    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2920    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2921    
2922    /* Searching for the first zero. */
2923    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
2924    jump = JUMP(SLJIT_NOT_ZERO);
2925    /* Two byte sequence. */
2926    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2927    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(2));
2928    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2929    
2930    JUMPHERE(jump);
2931    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2932    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
2933    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2934    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2935    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2936    
2937    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2938    jump = JUMP(SLJIT_NOT_ZERO);
2939    /* Three byte sequence. */
2940    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2941    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(3));
2942    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2943    
2944    /* Four byte sequence. */
2945    JUMPHERE(jump);
2946    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2947    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2948    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2949    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2950    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2951    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2952    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(4));
2953    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2954  }  }
2955    
2956  static SLJIT_INLINE void set_jumps(jump_list *list, struct sljit_label *label)  static void do_utfreadchar16(compiler_common *common)
2957  {  {
2958  while (list)  /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2959    {  of the character (>= 0xc0). Return value in TMP1. */
2960    /* sljit_set_label is clever enough to do nothing  DEFINE_COMPILER;
2961    if either the jump or the label is NULL */  struct sljit_jump *jump;
2962    sljit_set_label(list->jump, label);  
2963    list = list->next;  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2964    }  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2965    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2966    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2967    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2968    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2969    
2970    /* Searching for the first zero. */
2971    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
2972    jump = JUMP(SLJIT_NOT_ZERO);
2973    /* Two byte sequence. */
2974    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2975    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2976    
2977    JUMPHERE(jump);
2978    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x400);
2979    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_NOT_ZERO);
2980    /* This code runs only in 8 bit mode. No need to shift the value. */
2981    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2982    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2983    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
2984    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2985    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2986    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2987    /* Three byte sequence. */
2988    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2989    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2990  }  }
2991    
2992  static SLJIT_INLINE void add_jump(struct sljit_compiler *compiler, jump_list **list, struct sljit_jump* jump)  static void do_utfreadtype8(compiler_common *common)
2993  {  {
2994  jump_list *list_item = sljit_alloc_memory(compiler, sizeof(jump_list));  /* Fast decoding a UTF-8 character type. TMP2 contains the first byte
2995  if (list_item)  of the character (>= 0xc0). Return value in TMP1. */
2996    DEFINE_COMPILER;
2997    struct sljit_jump *jump;
2998    struct sljit_jump *compare;
2999    
3000    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
3001    
3002    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0x20);
3003    jump = JUMP(SLJIT_NOT_ZERO);
3004    /* Two byte sequence. */
3005    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
3006    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3007    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);
3008    /* The upper 5 bits are known at this point. */
3009    compare = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0x3);
3010    OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
3011    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
3012    OP2(SLJIT_OR, TMP2, 0, TMP2, 0, TMP1, 0);
3013    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
3014    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3015    
3016    JUMPHERE(compare);
3017    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
3018    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3019    
3020    /* We only have types for characters less than 256. */
3021    JUMPHERE(jump);
3022    OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
3023    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
3024    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
3025    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3026    }
3027    
3028    #endif /* COMPILE_PCRE8 */
3029    
3030    #endif /* SUPPORT_UTF */
3031    
3032    #ifdef SUPPORT_UCP
3033    
3034    /* UCD_BLOCK_SIZE must be 128 (see the assert below). */
3035    #define UCD_BLOCK_MASK 127
3036    #define UCD_BLOCK_SHIFT 7
3037    
3038    static void do_getucd(compiler_common *common)
3039    {
3040    /* Search the UCD record for the character comes in TMP1.
3041    Returns chartype in TMP1 and UCD offset in TMP2. */
3042    DEFINE_COMPILER;
3043    
3044    SLJIT_ASSERT(UCD_BLOCK_SIZE == 128 && sizeof(ucd_record) == 8);
3045    
3046    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
3047    OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
3048    OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(ucd_stage1));
3049    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);
3050    OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
3051    OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, TMP2, 0);
3052    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_stage2));
3053    OP1(SLJIT_MOV_UH, TMP2, 0, SLJIT_MEM2(TMP2, TMP1), 1);
3054    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_records) + SLJIT_OFFSETOF(ucd_record, chartype));
3055    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM2(TMP1, TMP2), 3);
3056    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3057    }
3058    #endif
3059    
3060    static SLJIT_INLINE struct sljit_label *mainloop_entry(compiler_common *common, BOOL hascrorlf, BOOL firstline)
3061    {
3062    DEFINE_COMPILER;
3063    struct sljit_label *mainloop;
3064    struct sljit_label *newlinelabel = NULL;
3065    struct sljit_jump *start;
3066    struct sljit_jump *end = NULL;
3067    struct sljit_jump *nl = NULL;
3068    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3069    struct sljit_jump *singlechar;
3070    #endif
3071    jump_list *newline = NULL;
3072    BOOL newlinecheck = FALSE;
3073    BOOL readuchar = FALSE;
3074    
3075    if (!(hascrorlf || firstline) && (common->nltype == NLTYPE_ANY ||
3076        common->nltype == NLTYPE_ANYCRLF || common->newline > 255))
3077      newlinecheck = TRUE;
3078    
3079    if (firstline)
3080    {    {
3081    list_item->next = *list;    /* Search for the end of the first line. */
3082    list_item->jump = jump;    SLJIT_ASSERT(common->first_line_end != 0);
3083    *list = list_item;    OP1(SLJIT_MOV, TMP3, 0, STR_PTR, 0);
3084    
3085      if (common->nltype == NLTYPE_FIXED && common->newline > 255)
3086        {
3087        mainloop = LABEL();
3088        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3089        end = CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3090        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(-1));
3091        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
3092        CMPTO(SLJIT_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, mainloop);
3093        CMPTO(SLJIT_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, mainloop);
3094        JUMPHERE(end);
3095        OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_SP), common->first_line_end, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3096        }
3097      else
3098        {
3099        end = CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3100        mainloop = LABEL();
3101        /* Continual stores does not cause data dependency. */
3102        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->first_line_end, STR_PTR, 0);
3103        read_char_range(common, common->nlmin, common->nlmax, TRUE);
3104        check_newlinechar(common, common->nltype, &newline, TRUE);
3105        CMPTO(SLJIT_LESS, STR_PTR, 0, STR_END, 0, mainloop);
3106        JUMPHERE(end);
3107        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->first_line_end, STR_PTR, 0);
3108        set_jumps(newline, LABEL());
3109        }
3110    
3111      OP1(SLJIT_MOV, STR_PTR, 0, TMP3, 0);
3112      }
3113    
3114    start = JUMP(SLJIT_JUMP);
3115    
3116    if (newlinecheck)
3117      {
3118      newlinelabel = LABEL();
3119      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3120      end = CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3121      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3122      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, common->newline & 0xff);
3123      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
3124    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
3125      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, UCHAR_SHIFT);
3126    #endif
3127      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3128      nl = JUMP(SLJIT_JUMP);
3129      }
3130    
3131    mainloop = LABEL();
3132    
3133    /* Increasing the STR_PTR here requires one less jump in the most common case. */
3134    #ifdef SUPPORT_UTF
3135    if (common->utf) readuchar = TRUE;
3136    #endif
3137    if (newlinecheck) readuchar = TRUE;
3138    
3139    if (readuchar)
3140      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3141    
3142    if (newlinecheck)
3143      CMPTO(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, newlinelabel);
3144    
3145    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3146    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3147    #if defined COMPILE_PCRE8
3148    if (common->utf)
3149      {
3150      singlechar = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
3151      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
3152      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3153      JUMPHERE(singlechar);
3154      }
3155    #elif defined COMPILE_PCRE16
3156    if (common->utf)
3157      {
3158      singlechar = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xd800);
3159      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
3160      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);
3161      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
3162      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
3163      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3164      JUMPHERE(singlechar);
3165      }
3166    #endif /* COMPILE_PCRE[8|16] */
3167    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
3168    JUMPHERE(start);
3169    
3170    if (newlinecheck)
3171      {
3172      JUMPHERE(end);
3173      JUMPHERE(nl);
3174    }    }
3175    
3176    return mainloop;
3177  }  }
3178    
3179  static void add_stub(compiler_common *common, enum stub_types type, int data, struct sljit_jump *start)  #define MAX_N_CHARS 16
3180    #define MAX_N_BYTES 8
3181    
3182    static SLJIT_INLINE void add_prefix_byte(pcre_uint8 byte, pcre_uint8 *bytes)
3183  {  {
3184  DEFINE_COMPILER;  pcre_uint8 len = bytes[0];
3185  stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));  int i;
3186    
3187  if (list_item)  if (len == 255)
3188      return;
3189    
3190    if (len == 0)
3191    {    {
3192    list_item->type = type;    bytes[0] = 1;
3193    list_item->data = data;    bytes[1] = byte;
3194    list_item->start = start;    return;
3195    list_item->leave = LABEL();    }
3196    list_item->next = common->stubs;  
3197    common->stubs = list_item;  for (i = len; i > 0; i--)
3198      if (bytes[i] == byte)
3199        return;
3200    
3201    if (len >= MAX_N_BYTES - 1)
3202      {
3203      bytes[0] = 255;
3204      return;
3205    }    }
 }  
3206    
3207  static void flush_stubs(compiler_common *common)  len++;
3208  {  bytes[len] = byte;
3209  DEFINE_COMPILER;  bytes[0] = len;
3210  stub_list* list_item = common->stubs;  }
3211    
3212    static int scan_prefix(compiler_common *common, pcre_uchar *cc, pcre_uint32 *chars, pcre_uint8 *bytes, int max_chars)
3213    {
3214    /* Recursive function, which scans prefix literals. */
3215    BOOL last, any, caseless;
3216    int len, repeat, len_save, consumed = 0;
3217    pcre_uint32 chr, mask;
3218    pcre_uchar *alternative, *cc_save, *oc;
3219    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3220    pcre_uchar othercase[8];
3221    #elif defined SUPPORT_UTF && defined COMPILE_PCRE16
3222    pcre_uchar othercase[2];
3223    #else
3224    pcre_uchar othercase[1];
3225    #endif
3226    
3227  while (list_item)  repeat = 1;
3228    while (TRUE)
3229    {    {
3230    JUMPHERE(list_item->start);    last = TRUE;
3231    switch(list_item->type)    any = FALSE;
3232      caseless = FALSE;
3233      switch (*cc)
3234      {      {
3235      case stack_alloc:      case OP_CHARI:
3236      add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));      caseless = TRUE;
3237        case OP_CHAR:
3238        last = FALSE;
3239        cc++;
3240      break;      break;
3241    
3242      case max_index:      case OP_SOD:
3243      OP1(SLJIT_MOV, MAX_INDEX, 0, SLJIT_IMM, list_item->data);      case OP_SOM:
3244      break;      case OP_SET_SOM:
3245      }      case OP_NOT_WORD_BOUNDARY:
3246    JUMPTO(SLJIT_JUMP, list_item->leave);      case OP_WORD_BOUNDARY:
3247    list_item = list_item->next;      case OP_EODN:
3248    }      case OP_EOD:
3249  common->stubs = NULL;      case OP_CIRC:
3250  }      case OP_CIRCM:
3251        case OP_DOLL:
3252        case OP_DOLLM:
3253        /* Zero width assertions. */
3254        cc++;
3255        continue;
3256    
3257  static SLJIT_INLINE void decrease_call_count(compiler_common *common)      case OP_ASSERT:
3258  {      case OP_ASSERT_NOT:
3259  DEFINE_COMPILER;      case OP_ASSERTBACK:
3260        case OP_ASSERTBACK_NOT:
3261        cc = bracketend(cc);
3262        continue;
3263    
3264  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_MEM1(SLJIT_LOCALS_REG), CALL_COUNT, SLJIT_MEM1(SLJIT_LOCALS_REG), CALL_COUNT, SLJIT_IMM, 1);      case OP_PLUSI:
3265  add_jump(compiler, &common->calllimit, JUMP(SLJIT_C_ZERO));      case OP_MINPLUSI:
3266  }      case OP_POSPLUSI:
3267        caseless = TRUE;
3268        case OP_PLUS:
3269        case OP_MINPLUS:
3270        case OP_POSPLUS:
3271        cc++;
3272        break;
3273    
3274  static SLJIT_INLINE void allocate_stack(compiler_common *common, int size)      case OP_EXACTI:
3275  {      caseless = TRUE;
3276  /* May destroy all locals and registers except TMP2. */      case OP_EXACT:
3277  DEFINE_COMPILER;      repeat = GET2(cc, 1);
3278        last = FALSE;
3279        cc += 1 + IMM2_SIZE;
3280        break;
3281    
3282  OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_w));      case OP_QUERYI:
3283  #ifdef DESTROY_REGISTERS      case OP_MINQUERYI:
3284  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);      case OP_POSQUERYI:
3285  OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);      caseless = TRUE;
3286  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);      case OP_QUERY:
3287  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);      case OP_MINQUERY:
3288  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);      case OP_POSQUERY:
3289        len = 1;
3290        cc++;
3291    #ifdef SUPPORT_UTF
3292        if (common->utf && HAS_EXTRALEN(*cc)) len += GET_EXTRALEN(*cc);
3293  #endif  #endif
3294  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);
3295  }      if (max_chars == 0)
3296          return consumed;
3297        last = FALSE;
3298        break;
3299    
3300  static SLJIT_INLINE void free_stack(compiler_common *common, int size)      case OP_KET:
3301  {      cc += 1 + LINK_SIZE;
3302  DEFINE_COMPILER;      continue;
 OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_w));  
 }  
3303    
3304  static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)      case OP_ALT:
3305  {      cc += GET(cc, 1);
3306  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);  
   }  
 }  
3307    
3308  static SLJIT_INLINE void copy_ovector(compiler_common *common)      case OP_ONCE:
3309  {      case OP_ONCE_NC:
3310  DEFINE_COMPILER;      case OP_BRA:
3311  struct sljit_label *loop;      case OP_BRAPOS:
3312  struct sljit_jump *earlyexit;      case OP_CBRA:
3313        case OP_CBRAPOS:
3314        alternative = cc + GET(cc, 1);
3315        while (*alternative == OP_ALT)
3316          {
3317          max_chars = scan_prefix(common, alternative + 1 + LINK_SIZE, chars, bytes, max_chars);
3318          if (max_chars == 0)
3319            return consumed;
3320          alternative += GET(alternative, 1);
3321          }
3322    
3323  /* At this point we can freely use all registers. */      if (*cc == OP_CBRA || *cc == OP_CBRAPOS)
3324  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, ARGUMENTS, 0);        cc += IMM2_SIZE;
3325  OP1(SLJIT_MOV_SI, SLJIT_TEMPORARY_REG2, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, offsetcount));      cc += 1 + LINK_SIZE;
3326  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);  
 }  
3327    
3328  static SLJIT_INLINE BOOL char_has_othercase(compiler_common *common, uschar* cc)      case OP_CLASS:
3329  {  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3330  /* Detects if the character has an othercase. */      if (common->utf && !is_char7_bitset((const pcre_uint8 *)(cc + 1), FALSE)) return consumed;
3331  unsigned int c;  #endif
3332        any = TRUE;
3333        cc += 1 + 32 / sizeof(pcre_uchar);
3334        break;
3335    
3336  #ifdef SUPPORT_UTF8      case OP_NCLASS:
3337  if (common->utf8)  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3338    {      if (common->utf) return consumed;
   GETCHAR(c, cc);  
   if (c > 127)  
     {  
 #ifdef SUPPORT_UCP  
     return c != UCD_OTHERCASE(c);  
 #else  
     return FALSE;  
3339  #endif  #endif
3340      }      any = TRUE;
3341    }      cc += 1 + 32 / sizeof(pcre_uchar);
3342  else      break;
3343    
3344    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3345        case OP_XCLASS:
3346    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3347        if (common->utf) return consumed;
3348    #endif
3349        any = TRUE;
3350        cc += GET(cc, 1);
3351        break;
3352  #endif  #endif
   c = *cc;  
 return common->fcc[c] != c;  
 }  
3353    
3354  static SLJIT_INLINE unsigned int char_othercase(compiler_common *common, unsigned int c)      case OP_DIGIT:
3355  {  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3356  /* Returns with the othercase. */      if (common->utf && !is_char7_bitset((const pcre_uint8 *)common->ctypes - cbit_length + cbit_digit, FALSE))
3357  #ifdef SUPPORT_UTF8        return consumed;
 if (common->utf8 && c > 127)  
   {  
 #ifdef SUPPORT_UCP  
   return UCD_OTHERCASE(c);  
 #else  
   return c;  
3358  #endif  #endif
3359    }      any = TRUE;
3360        cc++;
3361        break;
3362    
3363        case OP_WHITESPACE:
3364    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3365        if (common->utf && !is_char7_bitset((const pcre_uint8 *)common->ctypes - cbit_length + cbit_space, FALSE))
3366          return consumed;
3367  #endif  #endif
3368  return common->fcc[c];      any = TRUE;
3369  }      cc++;
3370        break;
3371    
3372  static unsigned int char_get_othercase_bit(compiler_common *common, uschar* cc)      case OP_WORDCHAR:
3373  {  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3374  /* Detects if the character and its othercase has only 1 bit difference. */      if (common->utf && !is_char7_bitset((const pcre_uint8 *)common->ctypes - cbit_length + cbit_word, FALSE))
3375  unsigned int c, oc, bit;        return consumed;
 #ifdef SUPPORT_UTF8  
 int n;  
3376  #endif  #endif
3377        any = TRUE;
3378        cc++;
3379        break;
3380    
3381        case OP_NOT:
3382        case OP_NOTI:
3383        cc++;
3384        /* Fall through. */
3385        case OP_NOT_DIGIT:
3386        case OP_NOT_WHITESPACE:
3387        case OP_NOT_WORDCHAR:
3388        case OP_ANY:
3389        case OP_ALLANY:
3390    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3391        if (common->utf) return consumed;
3392    #endif
3393        any = TRUE;
3394        cc++;
3395        break;
3396    
 #ifdef SUPPORT_UTF8  
 if (common->utf8)  
   {  
   GETCHAR(c, cc);  
   if (c <= 127)  
     oc = common->fcc[c];  
   else  
     {  
3397  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3398      oc = UCD_OTHERCASE(c);      case OP_NOTPROP:
3399  #else      case OP_PROP:
3400      oc = c;  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3401        if (common->utf) return consumed;
3402  #endif  #endif
3403      }      any = TRUE;
3404    }      cc += 1 + 2;
3405  else      break;
   {  
   c = *cc;  
   oc = common->fcc[c];  
   }  
 #else  
 c = *cc;  
 oc = common->fcc[c];  
3406  #endif  #endif
3407    
3408  SLJIT_ASSERT(c != oc);      case OP_TYPEEXACT:
3409        repeat = GET2(cc, 1);
3410        cc += 1 + IMM2_SIZE;
3411        continue;
3412    
3413  bit = c ^ oc;      case OP_NOTEXACT:
3414  /* Optimized for English alphabet. */      case OP_NOTEXACTI:
3415  if (c <= 127 && bit == 0x20)  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3416    return (0 << 8) | 0x20;      if (common->utf) return consumed;
3417    #endif
3418        any = TRUE;
3419        repeat = GET2(cc, 1);
3420        cc += 1 + IMM2_SIZE + 1;
3421        break;
3422    
3423  /* Since c != oc, they must have at least 1 bit difference. */      default:
3424  if (!ispowerof2(bit))      return consumed;
3425    return 0;      }
3426    
3427  #ifdef SUPPORT_UTF8    if (any)
 if (common->utf8 && c > 127)  
   {  
   n = _pcre_utf8_table4[*cc & 0x3f];  
   while ((bit & 0x3f) == 0)  
3428      {      {
3429      n--;  #if defined COMPILE_PCRE8
3430      bit >>= 6;      mask = 0xff;
3431      }  #elif defined COMPILE_PCRE16
3432    return (n << 8) | bit;      mask = 0xffff;
3433    }  #elif defined COMPILE_PCRE32
3434        mask = 0xffffffff;
3435    #else
3436        SLJIT_ASSERT_STOP();
3437  #endif  #endif
 return (0 << 8) | bit;  
 }  
3438    
3439  static SLJIT_INLINE void check_input_end(compiler_common *common, jump_list **fallbacks)      do
3440  {        {
3441  DEFINE_COMPILER;        chars[0] = mask;
3442  add_jump(compiler, fallbacks, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));        chars[1] = mask;
3443  }        bytes[0] = 255;