/[pcre]/code/trunk/pcre_jit_compile.c
ViewVC logotype

Diff of /code/trunk/pcre_jit_compile.c

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

revision 695 by zherczeg, Sat Sep 17 06:05:38 2011 UTC revision 1587 by zherczeg, Sat Aug 8 06:02:16 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 CALL_COUNT    SLJIT_GENERAL_EREG2  #define ARGUMENTS     SLJIT_S4
483  #define RETURN_ADDR   SLJIT_TEMPORARY_EREG1  #define RETURN_ADDR   SLJIT_R4
484    
485  /* Locals layout. */  /* Local space layout. */
486  /* These two locals can be used by the current opcode. */  /* These two locals can be used by the current opcode. */
487  #define LOCALS0          (0 * sizeof(sljit_w))  #define LOCALS0          (0 * sizeof(sljit_sw))
488  #define LOCALS1          (1 * sizeof(sljit_w))  #define LOCALS1          (1 * sizeof(sljit_sw))
489  /* Two local variables for possessive quantifiers (char1 cannot use them). */  /* Two local variables for possessive quantifiers (char1 cannot use them). */
490  #define POSSESSIVE0      (2 * sizeof(sljit_w))  #define POSSESSIVE0      (2 * sizeof(sljit_sw))
491  #define POSSESSIVE1      (3 * sizeof(sljit_w))  #define POSSESSIVE1      (3 * sizeof(sljit_sw))
 /* Head of the 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 MAX_INDEX        (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    #define READ_CHAR_MAX 0x7fffffff
543    
544  static uschar* bracketend(uschar* cc)  static pcre_uchar *bracketend(pcre_uchar *cc)
545  {  {
546  SLJIT_ASSERT((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND));  SLJIT_ASSERT((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND));
547  do cc += GET(cc, 1); while (*cc == OP_ALT);  do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 405  cc += 1 + LINK_SIZE; Line 550  cc += 1 + LINK_SIZE;
550  return cc;  return cc;
551  }  }
552    
553    static int no_alternatives(pcre_uchar *cc)
554    {
555    int count = 0;
556    SLJIT_ASSERT((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND));
557    do
558      {
559      cc += GET(cc, 1);
560      count++;
561      }
562    while (*cc == OP_ALT);
563    SLJIT_ASSERT(*cc >= OP_KET && *cc <= OP_KETRPOS);
564    return count;
565    }
566    
567    static int ones_in_half_byte[16] = {
568      /* 0 */ 0, 1, 1, 2, /* 4 */ 1, 2, 2, 3,
569      /* 8 */ 1, 2, 2, 3, /* 12 */ 2, 3, 3, 4
570    };
571    
572  /* Functions whose might need modification for all new supported opcodes:  /* Functions whose might need modification for all new supported opcodes:
573   next_opcode   next_opcode
574   get_localspace   check_opcode_types
575   set_localptrs   set_private_data_ptrs
576   get_framesize   get_framesize
577   init_frame   init_frame
578   get_localsize   get_private_data_copy_length
579   copy_locals   copy_private_data
580   compile_hotpath   compile_matchingpath
581   compile_fallbackpath   compile_backtrackingpath
582  */  */
583    
584  static uschar *next_opcode(compiler_common *common, uschar *cc)  static pcre_uchar *next_opcode(compiler_common *common, pcre_uchar *cc)
585  {  {
586  SLJIT_UNUSED_ARG(common);  SLJIT_UNUSED_ARG(common);
587  switch(*cc)  switch(*cc)
# Line 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    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  if (stackpos < stacktop)  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, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), MAX_INDEX);      case OP_NOTPLUS: \
1033        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setmaxindex);      case OP_NOTSTARI: \
1034        stackpos += (int)sizeof(sljit_w);      case OP_NOTPLUSI:
       OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);  
       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        break;
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;
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 OP_TYPEUPTO:
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_TYPEMINUPTO:
1172        space = 2;
1173        size = 1 + IMM2_SIZE;
1174        break;
1175    
1176        case OP_CLASS:
1177        case OP_NCLASS:
1178        size += 1 + 32 / sizeof(pcre_uchar);
1179        space = get_class_iterator_size(cc + size);
1180        break;
1181    
1182    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1183        case OP_XCLASS:
1184        size = GET(cc, 1);
1185        space = get_class_iterator_size(cc + size);
1186      break;      break;
1187    #endif
1188    
1189      default:      default:
1190      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
1191      SLJIT_ASSERT(cc != NULL);      SLJIT_ASSERT(cc != NULL);
1192      break;      break;
1193      }      }
1194    
1195      /* Character iterators, which are not inside a repeated bracket,
1196         gets a private slot instead of allocating it on the stack. */
1197      if (space > 0 && cc >= end)
1198        {
1199        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1200        private_data_ptr += sizeof(sljit_sw) * space;
1201        }
1202    
1203      if (size != 0)
1204        {
1205        if (size < 0)
1206          {
1207          cc += -size;
1208    #ifdef SUPPORT_UTF
1209          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1210    #endif
1211          }
1212        else
1213          cc += size;
1214        }
1215    
1216      if (bracketlen > 0)
1217        {
1218        if (cc >= end)
1219          {
1220          end = bracketend(cc);
1221          if (end[-1 - LINK_SIZE] == OP_KET)
1222            end = NULL;
1223          }
1224        cc += bracketlen;
1225        }
1226    }    }
1227  SLJIT_ASSERT(cc == ccend);  *private_data_start = private_data_ptr;
 return localsize;  
1228  }  }
1229    
1230  static void copy_locals(compiler_common *common, uschar *cc, uschar *ccend,  /* Returns with a frame_types (always < 0) if no need for frame. */
1231    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)
1232  {  {
1233  DEFINE_COMPILER;  int length = 0;
1234  int srcw[2];  int possessive = 0;
1235  int count;  BOOL stack_restore = FALSE;
1236  BOOL tmp1next = TRUE;  BOOL setsom_found = recursive;
1237  BOOL tmp1empty = TRUE;  BOOL setmark_found = recursive;
1238  BOOL tmp2empty = TRUE;  /* The last capture is a local variable even for recursions. */
1239  uschar *alternative;  BOOL capture_last_found = FALSE;
1240  enum {  
1241    start,  #if defined DEBUG_FORCE_CONTROL_HEAD && DEBUG_FORCE_CONTROL_HEAD
1242    loop,  SLJIT_ASSERT(common->control_head_ptr != 0);
1243    end  *needs_control_head = TRUE;
1244  } status;  #else
1245    *needs_control_head = FALSE;
1246  status = save ? start : loop;  #endif
 stackptr = STACK(stackptr - 2);  
 stacktop = STACK(stacktop - 1);  
1247    
1248  if (!save)  if (ccend == NULL)
1249    {    {
1250    stackptr += sizeof(sljit_w);    ccend = bracketend(cc) - (1 + LINK_SIZE);
1251    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)  
1252      {      {
1253      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);      possessive = length = (common->capture_last_ptr != 0) ? 5 : 3;
1254      stackptr += sizeof(sljit_w);      /* This is correct regardless of common->capture_last_ptr. */
1255      tmp2empty = FALSE;      capture_last_found = TRUE;
1256      }      }
1257    /* The tmp1next must be TRUE in either way. */    cc = next_opcode(common, cc);
1258    }    }
1259    
1260  while (status != end)  SLJIT_ASSERT(cc != NULL);
1261    {  while (cc < ccend)
1262    count = 0;    switch(*cc)
   switch(status)  
1263      {      {
1264      case start:      case OP_SET_SOM:
1265      SLJIT_ASSERT(save);      SLJIT_ASSERT(common->has_set_som);
1266      count = 1;      stack_restore = TRUE;
1267      srcw[0] = RECURSIVE_HEAD;      if (!setsom_found)
1268      status = loop;        {
1269          length += 2;
1270          setsom_found = TRUE;
1271          }
1272        cc += 1;
1273      break;      break;
1274    
1275      case loop:      case OP_MARK:
1276      if (cc >= ccend)      case OP_PRUNE_ARG:
1277        case OP_THEN_ARG:
1278        SLJIT_ASSERT(common->mark_ptr != 0);
1279        stack_restore = TRUE;
1280        if (!setmark_found)
1281        {        {
1282        status = end;        length += 2;
1283        break;        setmark_found = TRUE;
1284        }        }
1285        if (common->control_head_ptr != 0)
1286          *needs_control_head = TRUE;
1287        cc += 1 + 2 + cc[1];
1288        break;
1289    
1290      switch(*cc)      case OP_RECURSE:
1291        stack_restore = TRUE;
1292        if (common->has_set_som && !setsom_found)
1293        {        {
1294        case OP_ASSERT:        length += 2;
1295        case OP_ASSERT_NOT:        setsom_found = TRUE;
1296        case OP_ASSERTBACK:        }
1297        case OP_ASSERTBACK_NOT:      if (common->mark_ptr != 0 && !setmark_found)
1298        case OP_ONCE:        {
1299        case OP_BRAPOS:        length += 2;
1300        case OP_SBRA:        setmark_found = TRUE;
1301        case OP_SBRAPOS:        }
1302        case OP_SCOND:      if (common->capture_last_ptr != 0 && !capture_last_found)
1303        count = 1;        {
1304        srcw[0] = PRIV(cc);        length += 2;
1305        SLJIT_ASSERT(srcw[0] != 0);        capture_last_found = TRUE;
1306        cc += 1 + LINK_SIZE;        }
1307        break;      cc += 1 + LINK_SIZE;
1308        break;
1309    
1310        case OP_CBRA:      case OP_CBRA:
1311        case OP_SCBRA:      case OP_CBRAPOS:
1312        count = 1;      case OP_SCBRA:
1313        srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));      case OP_SCBRAPOS:
1314        cc += 1 + LINK_SIZE + 2;      stack_restore = TRUE;
1315        break;      if (common->capture_last_ptr != 0 && !capture_last_found)
1316          {
1317          length += 2;
1318          capture_last_found = TRUE;
1319          }
1320        length += 3;
1321        cc += 1 + LINK_SIZE + IMM2_SIZE;
1322        break;
1323    
1324        case OP_CBRAPOS:      case OP_THEN:
1325        case OP_SCBRAPOS:      stack_restore = TRUE;
1326        count = 2;      if (common->control_head_ptr != 0)
1327        srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));        *needs_control_head = TRUE;
1328        srcw[0] = PRIV(cc);      cc ++;
1329        SLJIT_ASSERT(srcw[0] != 0);      break;
       cc += 1 + LINK_SIZE + 2;  
       break;  
1330    
1331        case OP_COND:      default:
1332        /* Might be a hidden SCOND. */      stack_restore = TRUE;
1333        alternative = cc + GET(cc, 1);      /* Fall through. */
       if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)  
         {  
         count = 1;  
         srcw[0] = PRIV(cc);  
         SLJIT_ASSERT(srcw[0] != 0);  
         }  
       cc += 1 + LINK_SIZE;  
       break;  
1334    
1335        default:      case OP_NOT_WORD_BOUNDARY:
1336        cc = next_opcode(common, cc);      case OP_WORD_BOUNDARY:
1337        SLJIT_ASSERT(cc != NULL);      case OP_NOT_DIGIT:
1338        break;      case OP_DIGIT:
1339        }      case OP_NOT_WHITESPACE:
1340      break;      case OP_WHITESPACE:
1341        case OP_NOT_WORDCHAR:
1342        case OP_WORDCHAR:
1343        case OP_ANY:
1344        case OP_ALLANY:
1345        case OP_ANYBYTE:
1346        case OP_NOTPROP:
1347        case OP_PROP:
1348        case OP_ANYNL:
1349        case OP_NOT_HSPACE:
1350        case OP_HSPACE:
1351        case OP_NOT_VSPACE:
1352        case OP_VSPACE:
1353        case OP_EXTUNI:
1354        case OP_EODN:
1355        case OP_EOD:
1356        case OP_CIRC:
1357        case OP_CIRCM:
1358        case OP_DOLL:
1359        case OP_DOLLM:
1360        case OP_CHAR:
1361        case OP_CHARI:
1362        case OP_NOT:
1363        case OP_NOTI:
1364    
1365      case end:      case OP_EXACT:
1366      SLJIT_ASSERT_STOP();      case OP_POSSTAR:
1367        case OP_POSPLUS:
1368        case OP_POSQUERY:
1369        case OP_POSUPTO:
1370    
1371        case OP_EXACTI:
1372        case OP_POSSTARI:
1373        case OP_POSPLUSI:
1374        case OP_POSQUERYI:
1375        case OP_POSUPTOI:
1376    
1377        case OP_NOTEXACT:
1378        case OP_NOTPOSSTAR:
1379        case OP_NOTPOSPLUS:
1380        case OP_NOTPOSQUERY:
1381        case OP_NOTPOSUPTO:
1382    
1383        case OP_NOTEXACTI:
1384        case OP_NOTPOSSTARI:
1385        case OP_NOTPOSPLUSI:
1386        case OP_NOTPOSQUERYI:
1387        case OP_NOTPOSUPTOI:
1388    
1389        case OP_TYPEEXACT:
1390        case OP_TYPEPOSSTAR:
1391        case OP_TYPEPOSPLUS:
1392        case OP_TYPEPOSQUERY:
1393        case OP_TYPEPOSUPTO:
1394    
1395        case OP_CLASS:
1396        case OP_NCLASS:
1397        case OP_XCLASS:
1398    
1399        cc = next_opcode(common, cc);
1400        SLJIT_ASSERT(cc != NULL);
1401      break;      break;
1402      }      }
1403    
1404    while (count > 0)  /* Possessive quantifiers can use a special case. */
1405    if (SLJIT_UNLIKELY(possessive == length))
1406      return stack_restore ? no_frame : no_stack;
1407    
1408    if (length > 0)
1409      return length + 1;
1410    return stack_restore ? no_frame : no_stack;
1411    }
1412    
1413    static void init_frame(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, int stackpos, int stacktop, BOOL recursive)
1414    {
1415    DEFINE_COMPILER;
1416    BOOL setsom_found = recursive;
1417    BOOL setmark_found = recursive;
1418    /* The last capture is a local variable even for recursions. */
1419    BOOL capture_last_found = FALSE;
1420    int offset;
1421    
1422    /* >= 1 + shortest item size (2) */
1423    SLJIT_UNUSED_ARG(stacktop);
1424    SLJIT_ASSERT(stackpos >= stacktop + 2);
1425    
1426    stackpos = STACK(stackpos);
1427    if (ccend == NULL)
1428      {
1429      ccend = bracketend(cc) - (1 + LINK_SIZE);
1430      if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))
1431        cc = next_opcode(common, cc);
1432      }
1433    
1434    SLJIT_ASSERT(cc != NULL);
1435    while (cc < ccend)
1436      switch(*cc)
1437      {      {
1438      count--;      case OP_SET_SOM:
1439      if (save)      SLJIT_ASSERT(common->has_set_som);
1440        if (!setsom_found)
1441        {        {
1442        if (tmp1next)        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(0));
1443          {        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1444          if (!tmp1empty)        stackpos += (int)sizeof(sljit_sw);
1445            {        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1446            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        stackpos += (int)sizeof(sljit_sw);
1447            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;  
         }  
1448        }        }
1449      else      cc += 1;
1450        {      break;
1451        if (tmp1next)  
1452          {      case OP_MARK:
1453          SLJIT_ASSERT(!tmp1empty);      case OP_PRUNE_ARG:
1454          OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count], TMP1, 0);      case OP_THEN_ARG:
1455          tmp1empty = stackptr >= stacktop;      SLJIT_ASSERT(common->mark_ptr != 0);
1456          if (!tmp1empty)      if (!setmark_found)
1457            {        {
1458            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->mark_ptr);
1459            stackptr += sizeof(sljit_w);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
1460            }        stackpos += (int)sizeof(sljit_sw);
1461          tmp1next = FALSE;        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1462          }        stackpos += (int)sizeof(sljit_sw);
1463        else        setmark_found = TRUE;
         {  
         SLJIT_ASSERT(!tmp2empty);  
         OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count], TMP2, 0);  
         tmp2empty = stackptr >= stacktop;  
         if (!tmp2empty)  
           {  
           OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);  
           stackptr += sizeof(sljit_w);  
           }  
         tmp1next = TRUE;  
         }  
1464        }        }
1465      }      cc += 1 + 2 + cc[1];
1466    }      break;
1467    
1468  if (save)      case OP_RECURSE:
1469    {      if (common->has_set_som && !setsom_found)
   if (tmp1next)  
     {  
     if (!tmp1empty)  
1470        {        {
1471        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(0));
1472        stackptr += sizeof(sljit_w);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1473          stackpos += (int)sizeof(sljit_sw);
1474          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1475          stackpos += (int)sizeof(sljit_sw);
1476          setsom_found = TRUE;
1477        }        }
1478      if (!tmp2empty)      if (common->mark_ptr != 0 && !setmark_found)
1479        {        {
1480        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->mark_ptr);
1481        stackptr += sizeof(sljit_w);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
1482          stackpos += (int)sizeof(sljit_sw);
1483          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1484          stackpos += (int)sizeof(sljit_sw);
1485          setmark_found = TRUE;
1486        }        }
1487      }      if (common->capture_last_ptr != 0 && !capture_last_found)
   else  
     {  
     if (!tmp2empty)  
1488        {        {
1489        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->capture_last_ptr);
1490        stackptr += sizeof(sljit_w);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1491          stackpos += (int)sizeof(sljit_sw);
1492          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1493          stackpos += (int)sizeof(sljit_sw);
1494          capture_last_found = TRUE;
1495        }        }
1496      if (!tmp1empty)      cc += 1 + LINK_SIZE;
1497        break;
1498    
1499        case OP_CBRA:
1500        case OP_CBRAPOS:
1501        case OP_SCBRA:
1502        case OP_SCBRAPOS:
1503        if (common->capture_last_ptr != 0 && !capture_last_found)
1504        {        {
1505        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->capture_last_ptr);
1506        stackptr += sizeof(sljit_w);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1507          stackpos += (int)sizeof(sljit_sw);
1508          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1509          stackpos += (int)sizeof(sljit_sw);
1510          capture_last_found = TRUE;
1511        }        }
1512        offset = (GET2(cc, 1 + LINK_SIZE)) << 1;
1513        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));
1514        stackpos += (int)sizeof(sljit_sw);
1515        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(offset));
1516        OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(offset + 1));
1517        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1518        stackpos += (int)sizeof(sljit_sw);
1519        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);
1520        stackpos += (int)sizeof(sljit_sw);
1521    
1522        cc += 1 + LINK_SIZE + IMM2_SIZE;
1523        break;
1524    
1525        default:
1526        cc = next_opcode(common, cc);
1527        SLJIT_ASSERT(cc != NULL);
1528        break;
1529      }      }
   }  
 SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));  
 }  
1530    
1531  static SLJIT_INLINE BOOL ispowerof2(unsigned int value)  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, 0);
1532  {  SLJIT_ASSERT(stackpos == STACK(stacktop));
 return (value & (value - 1)) == 0;  
1533  }  }
1534    
1535  static SLJIT_INLINE void set_jumps(jump_list *list, struct sljit_label *label)  static SLJIT_INLINE int get_private_data_copy_length(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, BOOL needs_control_head)
1536  {  {
1537  while (list)  int private_data_length = needs_control_head ? 3 : 2;
1538    int size;
1539    pcre_uchar *alternative;
1540    /* Calculate the sum of the private machine words. */
1541    while (cc < ccend)
1542    {    {
1543    /* sljit_set_label is clever enough to do nothing    size = 0;
1544    if either the jump or the label is NULL */    switch(*cc)
1545    sljit_set_label(list->jump, label);      {
1546    list = list->next;      case OP_KET:
1547        if (PRIVATE_DATA(cc) != 0)
1548          {
1549          private_data_length++;
1550          SLJIT_ASSERT(PRIVATE_DATA(cc + 1) != 0);
1551          cc += PRIVATE_DATA(cc + 1);
1552          }
1553        cc += 1 + LINK_SIZE;
1554        break;
1555    
1556        case OP_ASSERT:
1557        case OP_ASSERT_NOT:
1558        case OP_ASSERTBACK:
1559        case OP_ASSERTBACK_NOT:
1560        case OP_ONCE:
1561        case OP_ONCE_NC:
1562        case OP_BRAPOS:
1563        case OP_SBRA:
1564        case OP_SBRAPOS:
1565        case OP_SCOND:
1566        private_data_length++;
1567        SLJIT_ASSERT(PRIVATE_DATA(cc) != 0);
1568        cc += 1 + LINK_SIZE;
1569        break;
1570    
1571        case OP_CBRA:
1572        case OP_SCBRA:
1573        if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1574          private_data_length++;
1575        cc += 1 + LINK_SIZE + IMM2_SIZE;
1576        break;
1577    
1578        case OP_CBRAPOS:
1579        case OP_SCBRAPOS:
1580        private_data_length += 2;
1581        cc += 1 + LINK_SIZE + IMM2_SIZE;
1582        break;
1583    
1584        case OP_COND:
1585        /* Might be a hidden SCOND. */
1586        alternative = cc + GET(cc, 1);
1587        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1588          private_data_length++;
1589        cc += 1 + LINK_SIZE;
1590        break;
1591    
1592        CASE_ITERATOR_PRIVATE_DATA_1
1593        if (PRIVATE_DATA(cc))
1594          private_data_length++;
1595        cc += 2;
1596    #ifdef SUPPORT_UTF
1597        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1598    #endif
1599        break;
1600    
1601        CASE_ITERATOR_PRIVATE_DATA_2A
1602        if (PRIVATE_DATA(cc))
1603          private_data_length += 2;
1604        cc += 2;
1605    #ifdef SUPPORT_UTF
1606        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1607    #endif
1608        break;
1609    
1610        CASE_ITERATOR_PRIVATE_DATA_2B
1611        if (PRIVATE_DATA(cc))
1612          private_data_length += 2;
1613        cc += 2 + IMM2_SIZE;
1614    #ifdef SUPPORT_UTF
1615        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1616    #endif
1617        break;
1618    
1619        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1620        if (PRIVATE_DATA(cc))
1621          private_data_length++;
1622        cc += 1;
1623        break;
1624    
1625        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1626        if (PRIVATE_DATA(cc))
1627          private_data_length += 2;
1628        cc += 1;
1629        break;
1630    
1631        CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1632        if (PRIVATE_DATA(cc))
1633          private_data_length += 2;
1634        cc += 1 + IMM2_SIZE;
1635        break;
1636    
1637        case OP_CLASS:
1638        case OP_NCLASS:
1639    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1640        case OP_XCLASS:
1641        size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
1642    #else
1643        size = 1 + 32 / (int)sizeof(pcre_uchar);
1644    #endif
1645        if (PRIVATE_DATA(cc))
1646          private_data_length += get_class_iterator_size(cc + size);
1647        cc += size;
1648        break;
1649    
1650        default:
1651        cc = next_opcode(common, cc);
1652        SLJIT_ASSERT(cc != NULL);
1653        break;
1654        }
1655    }    }
1656    SLJIT_ASSERT(cc == ccend);
1657    return private_data_length;
1658  }  }
1659    
1660  static SLJIT_INLINE void add_jump(struct sljit_compiler *compiler, jump_list **list, struct sljit_jump* jump)  static void copy_private_data(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend,
1661      BOOL save, int stackptr, int stacktop, BOOL needs_control_head)
1662    {
1663    DEFINE_COMPILER;
1664    int srcw[2];
1665    int count, size;
1666    BOOL tmp1next = TRUE;
1667    BOOL tmp1empty = TRUE;
1668    BOOL tmp2empty = TRUE;
1669    pcre_uchar *alternative;
1670    enum {
1671      start,
1672      loop,
1673      end
1674    } status;
1675    
1676    status = save ? start : loop;
1677    stackptr = STACK(stackptr - 2);
1678    stacktop = STACK(stacktop - 1);
1679    
1680    if (!save)
1681      {
1682      stackptr += (needs_control_head ? 2 : 1) * sizeof(sljit_sw);
1683      if (stackptr < stacktop)
1684        {
1685        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1686        stackptr += sizeof(sljit_sw);
1687        tmp1empty = FALSE;
1688        }
1689      if (stackptr < stacktop)
1690        {
1691        OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1692        stackptr += sizeof(sljit_sw);
1693        tmp2empty = FALSE;
1694        }
1695      /* The tmp1next must be TRUE in either way. */
1696      }
1697    
1698    do
1699      {
1700      count = 0;
1701      switch(status)
1702        {
1703        case start:
1704        SLJIT_ASSERT(save && common->recursive_head_ptr != 0);
1705        count = 1;
1706        srcw[0] = common->recursive_head_ptr;
1707        if (needs_control_head)
1708          {
1709          SLJIT_ASSERT(common->control_head_ptr != 0);
1710          count = 2;
1711          srcw[1] = common->control_head_ptr;
1712          }
1713        status = loop;
1714        break;
1715    
1716        case loop:
1717        if (cc >= ccend)
1718          {
1719          status = end;
1720          break;
1721          }
1722    
1723        switch(*cc)
1724          {
1725          case OP_KET:
1726          if (PRIVATE_DATA(cc) != 0)
1727            {
1728            count = 1;
1729            srcw[0] = PRIVATE_DATA(cc);
1730            SLJIT_ASSERT(PRIVATE_DATA(cc + 1) != 0);
1731            cc += PRIVATE_DATA(cc + 1);
1732            }
1733          cc += 1 + LINK_SIZE;
1734          break;
1735    
1736          case OP_ASSERT:
1737          case OP_ASSERT_NOT:
1738          case OP_ASSERTBACK:
1739          case OP_ASSERTBACK_NOT:
1740          case OP_ONCE:
1741          case OP_ONCE_NC:
1742          case OP_BRAPOS:
1743          case OP_SBRA:
1744          case OP_SBRAPOS:
1745          case OP_SCOND:
1746          count = 1;
1747          srcw[0] = PRIVATE_DATA(cc);
1748          SLJIT_ASSERT(srcw[0] != 0);
1749          cc += 1 + LINK_SIZE;
1750          break;
1751    
1752          case OP_CBRA:
1753          case OP_SCBRA:
1754          if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1755            {
1756            count = 1;
1757            srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1758            }
1759          cc += 1 + LINK_SIZE + IMM2_SIZE;
1760          break;
1761    
1762          case OP_CBRAPOS:
1763          case OP_SCBRAPOS:
1764          count = 2;
1765          srcw[0] = PRIVATE_DATA(cc);
1766          srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1767          SLJIT_ASSERT(srcw[0] != 0 && srcw[1] != 0);
1768          cc += 1 + LINK_SIZE + IMM2_SIZE;
1769          break;
1770    
1771          case OP_COND:
1772          /* Might be a hidden SCOND. */
1773          alternative = cc + GET(cc, 1);
1774          if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1775            {
1776            count = 1;
1777            srcw[0] = PRIVATE_DATA(cc);
1778            SLJIT_ASSERT(srcw[0] != 0);
1779            }
1780          cc += 1 + LINK_SIZE;
1781          break;
1782    
1783          CASE_ITERATOR_PRIVATE_DATA_1
1784          if (PRIVATE_DATA(cc))
1785            {
1786            count = 1;
1787            srcw[0] = PRIVATE_DATA(cc);
1788            }
1789          cc += 2;
1790    #ifdef SUPPORT_UTF
1791          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1792    #endif
1793          break;
1794    
1795          CASE_ITERATOR_PRIVATE_DATA_2A
1796          if (PRIVATE_DATA(cc))
1797            {
1798            count = 2;
1799            srcw[0] = PRIVATE_DATA(cc);
1800            srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1801            }
1802          cc += 2;
1803    #ifdef SUPPORT_UTF
1804          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1805    #endif
1806          break;
1807    
1808          CASE_ITERATOR_PRIVATE_DATA_2B
1809          if (PRIVATE_DATA(cc))
1810            {
1811            count = 2;
1812            srcw[0] = PRIVATE_DATA(cc);
1813            srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1814            }
1815          cc += 2 + IMM2_SIZE;
1816    #ifdef SUPPORT_UTF
1817          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1818    #endif
1819          break;
1820    
1821          CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1822          if (PRIVATE_DATA(cc))
1823            {
1824            count = 1;
1825            srcw[0] = PRIVATE_DATA(cc);
1826            }
1827          cc += 1;
1828          break;
1829    
1830          CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1831          if (PRIVATE_DATA(cc))
1832            {
1833            count = 2;
1834            srcw[0] = PRIVATE_DATA(cc);
1835            srcw[1] = srcw[0] + sizeof(sljit_sw);
1836            }
1837          cc += 1;
1838          break;
1839    
1840          CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1841          if (PRIVATE_DATA(cc))
1842            {
1843            count = 2;
1844            srcw[0] = PRIVATE_DATA(cc);
1845            srcw[1] = srcw[0] + sizeof(sljit_sw);
1846            }
1847          cc += 1 + IMM2_SIZE;
1848          break;
1849    
1850          case OP_CLASS:
1851          case OP_NCLASS:
1852    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1853          case OP_XCLASS:
1854          size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
1855    #else
1856          size = 1 + 32 / (int)sizeof(pcre_uchar);
1857    #endif
1858          if (PRIVATE_DATA(cc))
1859            switch(get_class_iterator_size(cc + size))
1860              {
1861              case 1:
1862              count = 1;
1863              srcw[0] = PRIVATE_DATA(cc);
1864              break;
1865    
1866              case 2:
1867              count = 2;
1868              srcw[0] = PRIVATE_DATA(cc);
1869              srcw[1] = srcw[0] + sizeof(sljit_sw);
1870              break;
1871    
1872              default:
1873              SLJIT_ASSERT_STOP();
1874              break;
1875              }
1876          cc += size;
1877          break;
1878    
1879          default:
1880          cc = next_opcode(common, cc);
1881          SLJIT_ASSERT(cc != NULL);
1882          break;
1883          }
1884        break;
1885    
1886        case end:
1887        SLJIT_ASSERT_STOP();
1888        break;
1889        }
1890    
1891      while (count > 0)
1892        {
1893        count--;
1894        if (save)
1895          {
1896          if (tmp1next)
1897            {
1898            if (!tmp1empty)
1899              {
1900              OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1901              stackptr += sizeof(sljit_sw);
1902              }
1903            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), srcw[count]);
1904            tmp1empty = FALSE;
1905            tmp1next = FALSE;
1906            }
1907          else
1908            {
1909            if (!tmp2empty)
1910              {
1911              OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1912              stackptr += sizeof(sljit_sw);
1913              }
1914            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_SP), srcw[count]);
1915            tmp2empty = FALSE;
1916            tmp1next = TRUE;
1917            }
1918          }
1919        else
1920          {
1921          if (tmp1next)
1922            {
1923            SLJIT_ASSERT(!tmp1empty);
1924            OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), srcw[count], TMP1, 0);
1925            tmp1empty = stackptr >= stacktop;
1926            if (!tmp1empty)
1927              {
1928              OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1929              stackptr += sizeof(sljit_sw);
1930              }
1931            tmp1next = FALSE;
1932            }
1933          else
1934            {
1935            SLJIT_ASSERT(!tmp2empty);
1936            OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), srcw[count], TMP2, 0);
1937            tmp2empty = stackptr >= stacktop;
1938            if (!tmp2empty)
1939              {
1940              OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1941              stackptr += sizeof(sljit_sw);
1942              }
1943            tmp1next = TRUE;
1944            }
1945          }
1946        }
1947      }
1948    while (status != end);
1949    
1950    if (save)
1951      {
1952      if (tmp1next)
1953        {
1954        if (!tmp1empty)
1955          {
1956          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1957          stackptr += sizeof(sljit_sw);
1958          }
1959        if (!tmp2empty)
1960          {
1961          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1962          stackptr += sizeof(sljit_sw);
1963          }
1964        }
1965      else
1966        {
1967        if (!tmp2empty)
1968          {
1969          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1970          stackptr += sizeof(sljit_sw);
1971          }
1972        if (!tmp1empty)
1973          {
1974          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1975          stackptr += sizeof(sljit_sw);
1976          }
1977        }
1978      }
1979    SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));
1980    }
1981    
1982    static SLJIT_INLINE pcre_uchar *set_then_offsets(compiler_common *common, pcre_uchar *cc, pcre_uint8 *current_offset)
1983    {
1984    pcre_uchar *end = bracketend(cc);
1985    BOOL has_alternatives = cc[GET(cc, 1)] == OP_ALT;
1986    
1987    /* Assert captures then. */
1988    if (*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT)
1989      current_offset = NULL;
1990    /* Conditional block does not. */
1991    if (*cc == OP_COND || *cc == OP_SCOND)
1992      has_alternatives = FALSE;
1993    
1994    cc = next_opcode(common, cc);
1995    if (has_alternatives)
1996      current_offset = common->then_offsets + (cc - common->start);
1997    
1998    while (cc < end)
1999      {
2000      if ((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND))
2001        cc = set_then_offsets(common, cc, current_offset);
2002      else
2003        {
2004        if (*cc == OP_ALT && has_alternatives)
2005          current_offset = common->then_offsets + (cc + 1 + LINK_SIZE - common->start);
2006        if (*cc >= OP_THEN && *cc <= OP_THEN_ARG && current_offset != NULL)
2007          *current_offset = 1;
2008        cc = next_opcode(common, cc);
2009        }
2010      }
2011    
2012    return end;
2013    }
2014    
2015    #undef CASE_ITERATOR_PRIVATE_DATA_1
2016    #undef CASE_ITERATOR_PRIVATE_DATA_2A
2017    #undef CASE_ITERATOR_PRIVATE_DATA_2B
2018    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_1
2019    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
2020    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
2021    
2022    static SLJIT_INLINE BOOL is_powerof2(unsigned int value)
2023    {
2024    return (value & (value - 1)) == 0;
2025    }
2026    
2027    static SLJIT_INLINE void set_jumps(jump_list *list, struct sljit_label *label)
2028    {
2029    while (list)
2030      {
2031      /* sljit_set_label is clever enough to do nothing
2032      if either the jump or the label is NULL. */
2033      SET_LABEL(list->jump, label);
2034      list = list->next;
2035      }
2036    }
2037    
2038    static SLJIT_INLINE void add_jump(struct sljit_compiler *compiler, jump_list **list, struct sljit_jump *jump)
2039  {  {
2040  jump_list *list_item = sljit_alloc_memory(compiler, sizeof(jump_list));  jump_list *list_item = sljit_alloc_memory(compiler, sizeof(jump_list));
2041  if (list_item)  if (list_item)
# Line 1142  if (list_item) Line 2046  if (list_item)
2046    }    }
2047  }  }
2048    
2049  static void add_stub(compiler_common *common, enum stub_types type, int data, struct sljit_jump *start)  static void add_stub(compiler_common *common, struct sljit_jump *start)
2050  {  {
2051  DEFINE_COMPILER;  DEFINE_COMPILER;
2052  stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));  stub_list *list_item = sljit_alloc_memory(compiler, sizeof(stub_list));
2053    
2054  if (list_item)  if (list_item)
2055    {    {
   list_item->type = type;  
   list_item->data = data;  
2056    list_item->start = start;    list_item->start = start;
2057    list_item->leave = LABEL();    list_item->quit = LABEL();
2058    list_item->next = common->stubs;    list_item->next = common->stubs;
2059    common->stubs = list_item;    common->stubs = list_item;
2060    }    }
2061  }  }
2062    
2063    static void flush_stubs(compiler_common *common)
2064    {
2065    DEFINE_COMPILER;
2066    stub_list *list_item = common->stubs;
2067    
2068    while (list_item)
2069      {
2070      JUMPHERE(list_item->start);
2071      add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));
2072      JUMPTO(SLJIT_JUMP, list_item->quit);
2073      list_item = list_item->next;
2074      }
2075    common->stubs = NULL;
2076    }
2077    
2078    static void add_label_addr(compiler_common *common, sljit_uw *update_addr)
2079    {
2080    DEFINE_COMPILER;
2081    label_addr_list *label_addr;
2082    
2083    label_addr = sljit_alloc_memory(compiler, sizeof(label_addr_list));
2084    if (label_addr == NULL)
2085      return;
2086    label_addr->label = LABEL();
2087    label_addr->update_addr = update_addr;
2088    label_addr->next = common->label_addrs;
2089    common->label_addrs = label_addr;
2090    }
2091    
2092    static SLJIT_INLINE void count_match(compiler_common *common)
2093    {
2094    DEFINE_COMPILER;
2095    
2096    OP2(SLJIT_SUB | SLJIT_SET_E, COUNT_MATCH, 0, COUNT_MATCH, 0, SLJIT_IMM, 1);
2097    add_jump(compiler, &common->calllimit, JUMP(SLJIT_ZERO));
2098    }
2099    
2100    static SLJIT_INLINE void allocate_stack(compiler_common *common, int size)
2101    {
2102    /* May destroy all locals and registers except TMP2. */
2103    DEFINE_COMPILER;
2104    
2105    OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_sw));
2106    #ifdef DESTROY_REGISTERS
2107    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);
2108    OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);
2109    OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);
2110    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), LOCALS0, TMP1, 0);
2111    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), LOCALS1, TMP1, 0);
2112    #endif
2113    add_stub(common, CMP(SLJIT_GREATER, STACK_TOP, 0, STACK_LIMIT, 0));
2114    }
2115    
2116    static SLJIT_INLINE void free_stack(compiler_common *common, int size)
2117    {
2118    DEFINE_COMPILER;
2119    OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_sw));
2120    }
2121    
2122    static sljit_uw * allocate_read_only_data(compiler_common *common, sljit_uw size)
2123    {
2124    DEFINE_COMPILER;
2125    sljit_uw *result;
2126    
2127    if (SLJIT_UNLIKELY(sljit_get_compiler_error(compiler)))
2128      return NULL;
2129    
2130    result = (sljit_uw *)SLJIT_MALLOC(size + sizeof(sljit_uw), compiler->allocator_data);
2131    if (SLJIT_UNLIKELY(result == NULL))
2132      {
2133      sljit_set_compiler_memory_error(compiler);
2134      return NULL;
2135      }
2136    
2137    *(void**)result = common->read_only_data_head;
2138    common->read_only_data_head = (void *)result;
2139    return result + 1;
2140    }
2141    
2142    static void free_read_only_data(void *current, void *allocator_data)
2143    {
2144    void *next;
2145    
2146    SLJIT_UNUSED_ARG(allocator_data);
2147    
2148    while (current != NULL)
2149      {
2150      next = *(void**)current;
2151      SLJIT_FREE(current, allocator_data);
2152      current = next;
2153      }
2154    }
2155    
2156    static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)
2157    {
2158    DEFINE_COMPILER;
2159    struct sljit_label *loop;
2160    int i;
2161    
2162    /* At this point we can freely use all temporary registers. */
2163    SLJIT_ASSERT(length > 1);
2164    /* TMP1 returns with begin - 1. */
2165    OP2(SLJIT_SUB, SLJIT_R0, 0, SLJIT_MEM1(SLJIT_S0), SLJIT_OFFSETOF(jit_arguments, begin), SLJIT_IMM, IN_UCHARS(1));
2166    if (length < 8)
2167      {
2168      for (i = 1; i < length; i++)
2169        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), OVECTOR(i), SLJIT_R0, 0);
2170      }
2171    else
2172      {
2173      GET_LOCAL_BASE(SLJIT_R1, 0, OVECTOR_START);
2174      OP1(SLJIT_MOV, SLJIT_R2, 0, SLJIT_IMM, length - 1);
2175      loop = LABEL();
2176      OP1(SLJIT_MOVU, SLJIT_MEM1(SLJIT_R1), sizeof(sljit_sw), SLJIT_R0, 0);
2177      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_IMM, 1);
2178      JUMPTO(SLJIT_NOT_ZERO, loop);
2179      }
2180    }
2181    
2182    static SLJIT_INLINE void do_reset_match(compiler_common *common, int length)
2183    {
2184    DEFINE_COMPILER;
2185    struct sljit_label *loop;
2186    int i;
2187    
2188    SLJIT_ASSERT(length > 1);
2189    /* OVECTOR(1) contains the "string begin - 1" constant. */
2190    if (length > 2)
2191      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(1));
2192    if (length < 8)
2193      {
2194      for (i = 2; i < length; i++)
2195        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), OVECTOR(i), TMP1, 0);
2196      }
2197    else
2198      {
2199      GET_LOCAL_BASE(TMP2, 0, OVECTOR_START + sizeof(sljit_sw));
2200      OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_IMM, length - 2);
2201      loop = LABEL();
2202      OP1(SLJIT_MOVU, SLJIT_MEM1(TMP2), sizeof(sljit_sw), TMP1, 0);
2203      OP2(SLJIT_SUB | SLJIT_SET_E, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, 1);
2204      JUMPTO(SLJIT_NOT_ZERO, loop);
2205      }
2206    
2207    OP1(SLJIT_MOV, STACK_TOP, 0, ARGUMENTS, 0);
2208    if (common->mark_ptr != 0)
2209      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->mark_ptr, SLJIT_IMM, 0);
2210    if (common->control_head_ptr != 0)
2211      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->control_head_ptr, SLJIT_IMM, 0);
2212    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(jit_arguments, stack));
2213    OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->start_ptr);
2214    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(struct sljit_stack, base));
2215    }
2216    
2217    static sljit_sw SLJIT_CALL do_search_mark(sljit_sw *current, const pcre_uchar *skip_arg)
2218    {
2219    while (current != NULL)
2220      {
2221      switch (current[-2])
2222        {
2223        case type_then_trap:
2224        break;
2225    
2226        case type_mark:
2227        if (STRCMP_UC_UC(skip_arg, (pcre_uchar *)current[-3]) == 0)
2228          return current[-4];
2229        break;
2230    
2231        default:
2232        SLJIT_ASSERT_STOP();
2233        break;
2234        }
2235      current = (sljit_sw*)current[-1];
2236      }
2237    return -1;
2238    }
2239    
2240    static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)
2241    {
2242    DEFINE_COMPILER;
2243    struct sljit_label *loop;
2244    struct sljit_jump *early_quit;
2245    
2246    /* At this point we can freely use all registers. */
2247    OP1(SLJIT_MOV, SLJIT_S2, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(1));
2248    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), OVECTOR(1), STR_PTR, 0);
2249    
2250    OP1(SLJIT_MOV, SLJIT_R0, 0, ARGUMENTS, 0);
2251    if (common->mark_ptr != 0)
2252      OP1(SLJIT_MOV, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_SP), common->mark_ptr);
2253    OP1(SLJIT_MOV_SI, SLJIT_R1, 0, SLJIT_MEM1(SLJIT_R0), SLJIT_OFFSETOF(jit_arguments, offset_count));
2254    if (common->mark_ptr != 0)
2255      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_R0), SLJIT_OFFSETOF(jit_arguments, mark_ptr), SLJIT_R2, 0);
2256    OP2(SLJIT_SUB, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_R0), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));
2257    OP1(SLJIT_MOV, SLJIT_R0, 0, SLJIT_MEM1(SLJIT_R0), SLJIT_OFFSETOF(jit_arguments, begin));
2258    GET_LOCAL_BASE(SLJIT_S0, 0, OVECTOR_START);
2259    /* Unlikely, but possible */
2260    early_quit = CMP(SLJIT_EQUAL, SLJIT_R1, 0, SLJIT_IMM, 0);
2261    loop = LABEL();
2262    OP2(SLJIT_SUB, SLJIT_S1, 0, SLJIT_MEM1(SLJIT_S0), 0, SLJIT_R0, 0);
2263    OP2(SLJIT_ADD, SLJIT_S0, 0, SLJIT_S0, 0, SLJIT_IMM, sizeof(sljit_sw));
2264    /* Copy the integer value to the output buffer */
2265    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2266    OP2(SLJIT_ASHR, SLJIT_S1, 0, SLJIT_S1, 0, SLJIT_IMM, UCHAR_SHIFT);
2267    #endif
2268    OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_R2), sizeof(int), SLJIT_S1, 0);
2269    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_R1, 0, SLJIT_R1, 0, SLJIT_IMM, 1);
2270    JUMPTO(SLJIT_NOT_ZERO, loop);
2271    JUMPHERE(early_quit);
2272    
2273    /* Calculate the return value, which is the maximum ovector value. */
2274    if (topbracket > 1)
2275      {
2276      GET_LOCAL_BASE(SLJIT_R0, 0, OVECTOR_START + topbracket * 2 * sizeof(sljit_sw));
2277      OP1(SLJIT_MOV, SLJIT_R1, 0, SLJIT_IMM, topbracket + 1);
2278    
2279      /* OVECTOR(0) is never equal to SLJIT_S2. */
2280      loop = LABEL();
2281      OP1(SLJIT_MOVU, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_R0), -(2 * (sljit_sw)sizeof(sljit_sw)));
2282      OP2(SLJIT_SUB, SLJIT_R1, 0, SLJIT_R1, 0, SLJIT_IMM, 1);
2283      CMPTO(SLJIT_EQUAL, SLJIT_R2, 0, SLJIT_S2, 0, loop);
2284      OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_R1, 0);
2285      }
2286    else
2287      OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, 1);
2288    }
2289    
2290    static SLJIT_INLINE void return_with_partial_match(compiler_common *common, struct sljit_label *quit)
2291    {
2292    DEFINE_COMPILER;
2293    struct sljit_jump *jump;
2294    
2295    SLJIT_COMPILE_ASSERT(STR_END == SLJIT_S1, str_end_must_be_saved_reg2);
2296    SLJIT_ASSERT(common->start_used_ptr != 0 && common->start_ptr != 0
2297      && (common->mode == JIT_PARTIAL_SOFT_COMPILE ? common->hit_start != 0 : common->hit_start == 0));
2298    
2299    OP1(SLJIT_MOV, SLJIT_R1, 0, ARGUMENTS, 0);
2300    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, PCRE_ERROR_PARTIAL);
2301    OP1(SLJIT_MOV_SI, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_R1), SLJIT_OFFSETOF(jit_arguments, real_offset_count));
2302    CMPTO(SLJIT_SIG_LESS, SLJIT_R2, 0, SLJIT_IMM, 2, quit);
2303    
2304    /* Store match begin and end. */
2305    OP1(SLJIT_MOV, SLJIT_S0, 0, SLJIT_MEM1(SLJIT_R1), SLJIT_OFFSETOF(jit_arguments, begin));
2306    OP1(SLJIT_MOV, SLJIT_R1, 0, SLJIT_MEM1(SLJIT_R1), SLJIT_OFFSETOF(jit_arguments, offsets));
2307    
2308    jump = CMP(SLJIT_SIG_LESS, SLJIT_R2, 0, SLJIT_IMM, 3);
2309    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);
2310    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2311    OP2(SLJIT_ASHR, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_IMM, UCHAR_SHIFT);
2312    #endif
2313    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_R1), 2 * sizeof(int), SLJIT_R2, 0);
2314    JUMPHERE(jump);
2315    
2316    OP1(SLJIT_MOV, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_SP), common->mode == JIT_PARTIAL_HARD_COMPILE ? common->start_used_ptr : common->hit_start);
2317    OP2(SLJIT_SUB, SLJIT_S1, 0, STR_END, 0, SLJIT_S0, 0);
2318    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2319    OP2(SLJIT_ASHR, SLJIT_S1, 0, SLJIT_S1, 0, SLJIT_IMM, UCHAR_SHIFT);
2320    #endif
2321    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_R1), sizeof(int), SLJIT_S1, 0);
2322    
2323    OP2(SLJIT_SUB, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_S0, 0);
2324    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2325    OP2(SLJIT_ASHR, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_IMM, UCHAR_SHIFT);
2326    #endif
2327    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_R1), 0, SLJIT_R2, 0);
2328    
2329    JUMPTO(SLJIT_JUMP, quit);
2330    }
2331    
2332    static SLJIT_INLINE void check_start_used_ptr(compiler_common *common)
2333    {
2334    /* May destroy TMP1. */
2335    DEFINE_COMPILER;
2336    struct sljit_jump *jump;
2337    
2338    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2339      {
2340      /* The value of -1 must be kept for start_used_ptr! */
2341      OP2(SLJIT_ADD, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, SLJIT_IMM, 1);
2342      /* Jumps if start_used_ptr < STR_PTR, or start_used_ptr == -1. Although overwriting
2343      is not necessary if start_used_ptr == STR_PTR, it does not hurt as well. */
2344      jump = CMP(SLJIT_LESS_EQUAL, TMP1, 0, STR_PTR, 0);
2345      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0);
2346      JUMPHERE(jump);
2347      }
2348    else if (common->mode == JIT_PARTIAL_HARD_COMPILE)
2349      {
2350      jump = CMP(SLJIT_LESS_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0);
2351      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0);
2352      JUMPHERE(jump);
2353      }
2354    }
2355    
2356    static SLJIT_INLINE BOOL char_has_othercase(compiler_common *common, pcre_uchar *cc)
2357    {
2358    /* Detects if the character has an othercase. */
2359    unsigned int c;
2360    
2361    #ifdef SUPPORT_UTF
2362    if (common->utf)
2363      {
2364      GETCHAR(c, cc);
2365      if (c > 127)
2366        {
2367    #ifdef SUPPORT_UCP
2368        return c != UCD_OTHERCASE(c);
2369    #else
2370        return FALSE;
2371    #endif
2372        }
2373    #ifndef COMPILE_PCRE8
2374      return common->fcc[c] != c;
2375    #endif
2376      }
2377    else
2378    #endif
2379      c = *cc;
2380    return MAX_255(c) ? common->fcc[c] != c : FALSE;
2381    }
2382    
2383    static SLJIT_INLINE unsigned int char_othercase(compiler_common *common, unsigned int c)
2384    {
2385    /* Returns with the othercase. */
2386    #ifdef SUPPORT_UTF
2387    if (common->utf && c > 127)
2388      {
2389    #ifdef SUPPORT_UCP
2390      return UCD_OTHERCASE(c);
2391    #else
2392      return c;
2393    #endif
2394      }
2395    #endif
2396    return TABLE_GET(c, common->fcc, c);
2397    }
2398    
2399    static unsigned int char_get_othercase_bit(compiler_common *common, pcre_uchar *cc)
2400    {
2401    /* Detects if the character and its othercase has only 1 bit difference. */
2402    unsigned int c, oc, bit;
2403    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2404    int n;
2405    #endif
2406    
2407    #ifdef SUPPORT_UTF
2408    if (common->utf)
2409      {
2410      GETCHAR(c, cc);
2411      if (c <= 127)
2412        oc = common->fcc[c];
2413      else
2414        {
2415    #ifdef SUPPORT_UCP
2416        oc = UCD_OTHERCASE(c);
2417    #else
2418        oc = c;
2419    #endif
2420        }
2421      }
2422    else
2423      {
2424      c = *cc;
2425      oc = TABLE_GET(c, common->fcc, c);
2426      }
2427    #else
2428    c = *cc;
2429    oc = TABLE_GET(c, common->fcc, c);
2430    #endif
2431    
2432    SLJIT_ASSERT(c != oc);
2433    
2434    bit = c ^ oc;
2435    /* Optimized for English alphabet. */
2436    if (c <= 127 && bit == 0x20)
2437      return (0 << 8) | 0x20;
2438    
2439    /* Since c != oc, they must have at least 1 bit difference. */
2440    if (!is_powerof2(bit))
2441      return 0;
2442    
2443    #if defined COMPILE_PCRE8
2444    
2445    #ifdef SUPPORT_UTF
2446    if (common->utf && c > 127)
2447      {
2448      n = GET_EXTRALEN(*cc);
2449      while ((bit & 0x3f) == 0)
2450        {
2451        n--;
2452        bit >>= 6;
2453        }
2454      return (n << 8) | bit;
2455      }
2456    #endif /* SUPPORT_UTF */
2457    return (0 << 8) | bit;
2458    
2459    #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2460    
2461    #ifdef SUPPORT_UTF
2462    if (common->utf && c > 65535)
2463      {
2464      if (bit >= (1 << 10))
2465        bit >>= 10;
2466      else
2467        return (bit < 256) ? ((2 << 8) | bit) : ((3 << 8) | (bit >> 8));
2468      }
2469    #endif /* SUPPORT_UTF */
2470    return (bit < 256) ? ((0 << 8) | bit) : ((1 << 8) | (bit >> 8));
2471    
2472    #endif /* COMPILE_PCRE[8|16|32] */
2473    }
2474    
2475    static void check_partial(compiler_common *common, BOOL force)
2476    {
2477    /* Checks whether a partial matching is occurred. Does not modify registers. */
2478    DEFINE_COMPILER;
2479    struct sljit_jump *jump = NULL;
2480    
2481    SLJIT_ASSERT(!force || common->mode != JIT_COMPILE);
2482    
2483    if (common->mode == JIT_COMPILE)
2484      return;
2485    
2486    if (!force)
2487      jump = CMP(SLJIT_GREATER_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0);
2488    else if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2489      jump = CMP(SLJIT_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, SLJIT_IMM, -1);
2490    
2491    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2492      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->hit_start, SLJIT_IMM, 0);
2493    else
2494      {
2495      if (common->partialmatchlabel != NULL)
2496        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2497      else
2498        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2499      }
2500    
2501    if (jump != NULL)
2502      JUMPHERE(jump);
2503    }
2504    
2505    static void check_str_end(compiler_common *common, jump_list **end_reached)
2506    {
2507    /* Does not affect registers. Usually used in a tight spot. */
2508    DEFINE_COMPILER;
2509    struct sljit_jump *jump;
2510    
2511    if (common->mode == JIT_COMPILE)
2512      {
2513      add_jump(compiler, end_reached, CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2514      return;
2515      }
2516    
2517    jump = CMP(SLJIT_LESS, STR_PTR, 0, STR_END, 0);
2518    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2519      {
2520      add_jump(compiler, end_reached, CMP(SLJIT_GREATER_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0));
2521      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->hit_start, SLJIT_IMM, 0);
2522      add_jump(compiler, end_reached, JUMP(SLJIT_JUMP));
2523      }
2524    else
2525      {
2526      add_jump(compiler, end_reached, CMP(SLJIT_GREATER_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0));
2527      if (common->partialmatchlabel != NULL)
2528        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2529      else
2530        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2531      }
2532    JUMPHERE(jump);
2533    }
2534    
2535    static void detect_partial_match(compiler_common *common, jump_list **backtracks)
2536    {
2537    DEFINE_COMPILER;
2538    struct sljit_jump *jump;
2539    
2540    if (common->mode == JIT_COMPILE)
2541      {
2542      add_jump(compiler, backtracks, CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2543      return;
2544      }
2545    
2546    /* Partial matching mode. */
2547    jump = CMP(SLJIT_LESS, STR_PTR, 0, STR_END, 0);
2548    add_jump(compiler, backtracks, CMP(SLJIT_GREATER_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0));
2549    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2550      {
2551      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->hit_start, SLJIT_IMM, 0);
2552      add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));
2553      }
2554    else
2555      {
2556      if (common->partialmatchlabel != NULL)
2557        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2558      else
2559        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2560      }
2561    JUMPHERE(jump);
2562    }
2563    
2564    static void peek_char(compiler_common *common, pcre_uint32 max)
2565    {
2566    /* Reads the character into TMP1, keeps STR_PTR.
2567    Does not check STR_END. TMP2 Destroyed. */
2568    DEFINE_COMPILER;
2569    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2570    struct sljit_jump *jump;
2571    #endif
2572    
2573    SLJIT_UNUSED_ARG(max);
2574    
2575    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2576    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2577    if (common->utf)
2578      {
2579      if (max < 128) return;
2580    
2581      jump = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2582      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2583      add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2584      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2585      JUMPHERE(jump);
2586      }
2587    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2588    
2589    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2590    if (common->utf)
2591      {
2592      if (max < 0xd800) return;
2593    
2594      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2595      jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2596      /* TMP2 contains the high surrogate. */
2597      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2598      OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2599      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2600      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2601      OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2602      JUMPHERE(jump);
2603      }
2604    #endif
2605    }
2606    
2607    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2608    
2609    static BOOL is_char7_bitset(const pcre_uint8 *bitset, BOOL nclass)
2610    {
2611    /* Tells whether the character codes below 128 are enough
2612    to determine a match. */
2613    const pcre_uint8 value = nclass ? 0xff : 0;
2614    const pcre_uint8 *end = bitset + 32;
2615    
2616    bitset += 16;
2617    do
2618      {
2619      if (*bitset++ != value)
2620        return FALSE;
2621      }
2622    while (bitset < end);
2623    return TRUE;
2624    }
2625    
2626    static void read_char7_type(compiler_common *common, BOOL full_read)
2627    {
2628    /* Reads the precise character type of a character into TMP1, if the character
2629    is less than 128. Otherwise it returns with zero. Does not check STR_END. The
2630    full_read argument tells whether characters above max are accepted or not. */
2631    DEFINE_COMPILER;
2632    struct sljit_jump *jump;
2633    
2634    SLJIT_ASSERT(common->utf);
2635    
2636    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2637    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2638    
2639    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2640    
2641    if (full_read)
2642      {
2643      jump = CMP(SLJIT_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2644      OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2645      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2646      JUMPHERE(jump);
2647      }
2648    }
2649    
2650    #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2651    
2652    static void read_char_range(compiler_common *common, pcre_uint32 min, pcre_uint32 max, BOOL update_str_ptr)
2653    {
2654    /* Reads the precise value of a character into TMP1, if the character is
2655    between min and max (c >= min && c <= max). Otherwise it returns with a value
2656    outside the range. Does not check STR_END. */
2657    DEFINE_COMPILER;
2658    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2659    struct sljit_jump *jump;
2660    #endif
2661    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2662    struct sljit_jump *jump2;
2663    #endif
2664    
2665    SLJIT_UNUSED_ARG(update_str_ptr);
2666    SLJIT_UNUSED_ARG(min);
2667    SLJIT_UNUSED_ARG(max);
2668    SLJIT_ASSERT(min <= max);
2669    
2670    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2671    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2672    
2673    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2674    if (common->utf)
2675      {
2676      if (max < 128 && !update_str_ptr) return;
2677    
2678      jump = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2679      if (min >= 0x10000)
2680        {
2681        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xf0);
2682        if (update_str_ptr)
2683          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2684        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2685        jump2 = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0x7);
2686        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2687        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2688        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2689        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2690        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2691        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2692        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2693        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2694        if (!update_str_ptr)
2695          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2696        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2697        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2698        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2699        JUMPHERE(jump2);
2700        if (update_str_ptr)
2701          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2702        }
2703      else if (min >= 0x800 && max <= 0xffff)
2704        {
2705        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xe0);
2706        if (update_str_ptr)
2707          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2708        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2709        jump2 = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xf);
2710        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2711        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2712        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2713        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2714        if (!update_str_ptr)
2715          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2716        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2717        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2718        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2719        JUMPHERE(jump2);
2720        if (update_str_ptr)
2721          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2722        }
2723      else if (max >= 0x800)
2724        add_jump(compiler, (max < 0x10000) ? &common->utfreadchar16 : &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2725      else if (max < 128)
2726        {
2727        OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2728        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2729        }
2730      else
2731        {
2732        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2733        if (!update_str_ptr)
2734          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2735        else
2736          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2737        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2738        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2739        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2740        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2741        if (update_str_ptr)
2742          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2743        }
2744      JUMPHERE(jump);
2745      }
2746    #endif
2747    
2748    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2749    if (common->utf)
2750      {
2751      if (max >= 0x10000)
2752        {
2753        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2754        jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2755        /* TMP2 contains the high surrogate. */
2756        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2757        OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2758        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2759        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2760        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2761        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2762        JUMPHERE(jump);
2763        return;
2764        }
2765    
2766      if (max < 0xd800 && !update_str_ptr) return;
2767    
2768      /* Skip low surrogate if necessary. */
2769      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2770      jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2771      if (update_str_ptr)
2772        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2773      if (max >= 0xd800)
2774        OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0x10000);
2775      JUMPHERE(jump);
2776      }
2777    #endif
2778    }
2779    
2780    static SLJIT_INLINE void read_char(compiler_common *common)
2781    {
2782    read_char_range(common, 0, READ_CHAR_MAX, TRUE);
2783    }
2784    
2785    static void read_char8_type(compiler_common *common, BOOL update_str_ptr)
2786    {
2787    /* Reads the character type into TMP1, updates STR_PTR. Does not check STR_END. */
2788    DEFINE_COMPILER;
2789    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2790    struct sljit_jump *jump;
2791    #endif
2792    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2793    struct sljit_jump *jump2;
2794    #endif
2795    
2796    SLJIT_UNUSED_ARG(update_str_ptr);
2797    
2798    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2799    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2800    
2801    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2802    if (common->utf)
2803      {
2804      /* This can be an extra read in some situations, but hopefully
2805      it is needed in most cases. */
2806      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2807      jump = CMP(SLJIT_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2808      if (!update_str_ptr)
2809        {
2810        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2811        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2812        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2813        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2814        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2815        OP2(SLJIT_OR, TMP2, 0, TMP2, 0, TMP1, 0);
2816        OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2817        jump2 = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 255);
2818        OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2819        JUMPHERE(jump2);
2820        }
2821      else
2822        add_jump(compiler, &common->utfreadtype8, JUMP(SLJIT_FAST_CALL));
2823      JUMPHERE(jump);
2824      return;
2825      }
2826    #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2827    
2828    #if !defined COMPILE_PCRE8
2829    /* The ctypes array contains only 256 values. */
2830    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2831    jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 255);
2832    #endif
2833    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2834    #if !defined COMPILE_PCRE8
2835    JUMPHERE(jump);
2836    #endif
2837    
2838    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2839    if (common->utf && update_str_ptr)
2840      {
2841      /* Skip low surrogate if necessary. */
2842      OP2(SLJIT_SUB, TMP2, 0, TMP2, 0, SLJIT_IMM, 0xd800);
2843      jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2844      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2845      JUMPHERE(jump);
2846      }
2847    #endif /* SUPPORT_UTF && COMPILE_PCRE16 */
2848    }
2849    
2850    static void skip_char_back(compiler_common *common)
2851    {
2852    /* Goes one character back. Affects STR_PTR and TMP1. Does not check begin. */
2853    DEFINE_COMPILER;
2854    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2855    #if defined COMPILE_PCRE8
2856    struct sljit_label *label;
2857    
2858    if (common->utf)
2859      {
2860      label = LABEL();
2861      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
2862      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2863      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xc0);
2864      CMPTO(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, 0x80, label);
2865      return;
2866      }
2867    #elif defined COMPILE_PCRE16
2868    if (common->utf)
2869      {
2870      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
2871      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2872      /* Skip low surrogate if necessary. */
2873      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
2874      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xdc00);
2875      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
2876      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
2877      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2878      return;
2879      }
2880    #endif /* COMPILE_PCRE[8|16] */
2881    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2882    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2883    }
2884    
2885    static void check_newlinechar(compiler_common *common, int nltype, jump_list **backtracks, BOOL jumpifmatch)
2886    {
2887    /* Character comes in TMP1. Checks if it is a newline. TMP2 may be destroyed. */
2888    DEFINE_COMPILER;
2889    struct sljit_jump *jump;
2890    
2891    if (nltype == NLTYPE_ANY)
2892      {
2893      add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));
2894      add_jump(compiler, backtracks, JUMP(jumpifmatch ? SLJIT_NOT_ZERO : SLJIT_ZERO));
2895      }
2896    else if (nltype == NLTYPE_ANYCRLF)
2897      {
2898      if (jumpifmatch)
2899        {
2900        add_jump(compiler, backtracks, CMP(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR));
2901        add_jump(compiler, backtracks, CMP(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL));
2902        }
2903      else
2904        {
2905        jump = CMP(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR);
2906        add_jump(compiler, backtracks, CMP(SLJIT_NOT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL));
2907        JUMPHERE(jump);
2908        }
2909      }
2910    else
2911      {
2912      SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline < 256);
2913      add_jump(compiler, backtracks, CMP(jumpifmatch ? SLJIT_EQUAL : SLJIT_NOT_EQUAL, TMP1, 0, SLJIT_IMM, common->newline));
2914      }
2915    }
2916    
2917    #ifdef SUPPORT_UTF
2918    
2919    #if defined COMPILE_PCRE8
2920    static void do_utfreadchar(compiler_common *common)
2921    {
2922    /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2923    of the character (>= 0xc0). Return char value in TMP1, length in TMP2. */
2924    DEFINE_COMPILER;
2925    struct sljit_jump *jump;
2926    
2927    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2928    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2929    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2930    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2931    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2932    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2933    
2934    /* Searching for the first zero. */
2935    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
2936    jump = JUMP(SLJIT_NOT_ZERO);
2937    /* Two byte sequence. */
2938    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2939    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(2));
2940    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2941    
2942    JUMPHERE(jump);
2943    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2944    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
2945    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2946    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2947    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2948    
2949    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2950    jump = JUMP(SLJIT_NOT_ZERO);
2951    /* Three byte sequence. */
2952    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2953    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(3));
2954    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2955    
2956    /* Four byte sequence. */
2957    JUMPHERE(jump);
2958    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2959    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2960    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2961    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2962    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2963    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2964    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(4));
2965    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2966    }
2967    
2968    static void do_utfreadchar16(compiler_common *common)
2969    {
2970    /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2971    of the character (>= 0xc0). Return value in TMP1. */
2972    DEFINE_COMPILER;
2973    struct sljit_jump *jump;
2974    
2975    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2976    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2977    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2978    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2979    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2980    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2981    
2982    /* Searching for the first zero. */
2983    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
2984    jump = JUMP(SLJIT_NOT_ZERO);
2985    /* Two byte sequence. */
2986    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2987    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2988    
2989    JUMPHERE(jump);
2990    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x400);
2991    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_NOT_ZERO);
2992    /* This code runs only in 8 bit mode. No need to shift the value. */
2993    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2994    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2995    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
2996    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2997    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2998    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2999    /* Three byte sequence. */
3000    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
3001    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3002    }
3003    
3004    static void do_utfreadtype8(compiler_common *common)
3005    {
3006    /* Fast decoding a UTF-8 character type. TMP2 contains the first byte
3007    of the character (>= 0xc0). Return value in TMP1. */
3008    DEFINE_COMPILER;
3009    struct sljit_jump *jump;
3010    struct sljit_jump *compare;
3011    
3012    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
3013    
3014    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0x20);
3015    jump = JUMP(SLJIT_NOT_ZERO);
3016    /* Two byte sequence. */
3017    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
3018    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3019    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);
3020    /* The upper 5 bits are known at this point. */
3021    compare = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0x3);
3022    OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
3023    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
3024    OP2(SLJIT_OR, TMP2, 0, TMP2, 0, TMP1, 0);
3025    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
3026    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3027    
3028    JUMPHERE(compare);
3029    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
3030    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3031    
3032    /* We only have types for characters less than 256. */
3033    JUMPHERE(jump);
3034    OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
3035    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
3036    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
3037    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3038    }
3039    
3040    #endif /* COMPILE_PCRE8 */
3041    
3042    #endif /* SUPPORT_UTF */
3043    
3044    #ifdef SUPPORT_UCP
3045    
3046    /* UCD_BLOCK_SIZE must be 128 (see the assert below). */
3047    #define UCD_BLOCK_MASK 127
3048    #define UCD_BLOCK_SHIFT 7
3049    
3050    static void do_getucd(compiler_common *common)
3051    {
3052    /* Search the UCD record for the character comes in TMP1.
3053    Returns chartype in TMP1 and UCD offset in TMP2. */
3054    DEFINE_COMPILER;
3055    
3056    SLJIT_ASSERT(UCD_BLOCK_SIZE == 128 && sizeof(ucd_record) == 8);
3057    
3058    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
3059    OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
3060    OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(ucd_stage1));
3061    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);
3062    OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
3063    OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, TMP2, 0);
3064    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_stage2));
3065    OP1(SLJIT_MOV_UH, TMP2, 0, SLJIT_MEM2(TMP2, TMP1), 1);
3066    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_records) + SLJIT_OFFSETOF(ucd_record, chartype));
3067    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM2(TMP1, TMP2), 3);
3068    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3069    }
3070    #endif
3071    
3072    static SLJIT_INLINE struct sljit_label *mainloop_entry(compiler_common *common, BOOL hascrorlf, BOOL firstline)
3073    {
3074    DEFINE_COMPILER;
3075    struct sljit_label *mainloop;
3076    struct sljit_label *newlinelabel = NULL;
3077    struct sljit_jump *start;
3078    struct sljit_jump *end = NULL;
3079    struct sljit_jump *nl = NULL;
3080    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3081    struct sljit_jump *singlechar;
3082    #endif
3083    jump_list *newline = NULL;
3084    BOOL newlinecheck = FALSE;
3085    BOOL readuchar = FALSE;
3086    
3087    if (!(hascrorlf || firstline) && (common->nltype == NLTYPE_ANY ||
3088        common->nltype == NLTYPE_ANYCRLF || common->newline > 255))
3089      newlinecheck = TRUE;
3090    
3091    if (firstline)
3092      {
3093      /* Search for the end of the first line. */
3094      SLJIT_ASSERT(common->first_line_end != 0);
3095      OP1(SLJIT_MOV, TMP3, 0, STR_PTR, 0);
3096    
3097      if (common->nltype == NLTYPE_FIXED && common->newline > 255)
3098        {
3099        mainloop = LABEL();
3100        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3101        end = CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3102        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(-1));
3103        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
3104        CMPTO(SLJIT_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, mainloop);
3105        CMPTO(SLJIT_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, mainloop);
3106        JUMPHERE(end);
3107        OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_SP), common->first_line_end, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3108        }
3109      else
3110        {
3111        end = CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3112        mainloop = LABEL();
3113        /* Continual stores does not cause data dependency. */
3114        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->first_line_end, STR_PTR, 0);
3115        read_char_range(common, common->nlmin, common->nlmax, TRUE);
3116        check_newlinechar(common, common->nltype, &newline, TRUE);
3117        CMPTO(SLJIT_LESS, STR_PTR, 0, STR_END, 0, mainloop);
3118        JUMPHERE(end);
3119        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->first_line_end, STR_PTR, 0);
3120        set_jumps(newline, LABEL());
3121        }
3122    
3123      OP1(SLJIT_MOV, STR_PTR, 0, TMP3, 0);
3124      }
3125    
3126    start = JUMP(SLJIT_JUMP);
3127    
3128    if (newlinecheck)
3129      {
3130      newlinelabel = LABEL();
3131      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3132      end = CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3133      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3134      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, common->newline & 0xff);
3135      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
3136    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
3137      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, UCHAR_SHIFT);
3138    #endif
3139      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3140      nl = JUMP(SLJIT_JUMP);
3141      }
3142    
3143    mainloop = LABEL();
3144    
3145    /* Increasing the STR_PTR here requires one less jump in the most common case. */
3146    #ifdef SUPPORT_UTF
3147    if (common->utf) readuchar = TRUE;
3148    #endif
3149    if (newlinecheck) readuchar = TRUE;
3150    
3151    if (readuchar)
3152      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3153    
3154    if (newlinecheck)
3155      CMPTO(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, newlinelabel);
3156    
3157    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3158    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3159    #if defined COMPILE_PCRE8
3160    if (common->utf)
3161      {
3162      singlechar = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
3163      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
3164      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3165      JUMPHERE(singlechar);
3166      }
3167    #elif defined COMPILE_PCRE16
3168    if (common->utf)
3169      {
3170      singlechar = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xd800);
3171      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
3172      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);
3173      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
3174      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
3175      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3176      JUMPHERE(singlechar);
3177      }
3178    #endif /* COMPILE_PCRE[8|16] */
3179    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
3180    JUMPHERE(start);
3181    
3182    if (newlinecheck)
3183      {
3184      JUMPHERE(end);
3185      JUMPHERE(nl);
3186      }
3187    
3188    return mainloop;
3189    }
3190    
3191    #define MAX_N_CHARS 16
3192    #define MAX_N_BYTES 8
3193    
3194    static SLJIT_INLINE void add_prefix_byte(pcre_uint8 byte, pcre_uint8 *bytes)
3195    {
3196    pcre_uint8 len = bytes[0];
3197    int i;
3198    
3199    if (len == 255)
3200      return;
3201    
3202    if (len == 0)
3203      {
3204      bytes[0] = 1;
3205      bytes[1] = byte;
3206      return;
3207      }
3208    
3209    for (i = len; i > 0; i--)
3210      if (bytes[i] == byte)
3211        return;
3212    
3213    if (len >= MAX_N_BYTES - 1)
3214      {
3215      bytes[0] = 255;
3216      return;
3217      }
3218    
3219  static void flush_stubs(compiler_common *common)  len++;
3220  {  bytes[len] = byte;
3221  DEFINE_COMPILER;  bytes[0] = len;
3222  stub_list* list_item = common->stubs;  }
3223    
3224    static int scan_prefix(compiler_common *common, pcre_uchar *cc, pcre_uint32 *chars, pcre_uint8 *bytes, int max_chars, pcre_uint32 *rec_count)
3225    {
3226    /* Recursive function, which scans prefix literals. */
3227    BOOL last, any, caseless;
3228    int len, repeat, len_save, consumed = 0;
3229    pcre_uint32 chr, mask;
3230    pcre_uchar *alternative, *cc_save, *oc;
3231    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3232    pcre_uchar othercase[8];
3233    #elif defined SUPPORT_UTF && defined COMPILE_PCRE16
3234    pcre_uchar othercase[2];
3235    #else
3236    pcre_uchar othercase[1];
3237    #endif
3238    
3239  while (list_item)  repeat = 1;
3240    while (TRUE)
3241    {    {
3242    JUMPHERE(list_item->start);    if (*rec_count == 0)
3243    switch(list_item->type)      return 0;
3244      (*rec_count)--;
3245    
3246      last = TRUE;
3247      any = FALSE;
3248      caseless = FALSE;
3249    
3250      switch (*cc)
3251      {      {
3252      case stack_alloc:      case OP_CHARI:
3253      add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));      caseless = TRUE;
3254        case OP_CHAR:
3255        last = FALSE;
3256        cc++;
3257      break;      break;
3258    
3259      case max_index:      case OP_SOD:
3260      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), MAX_INDEX, SLJIT_IMM, list_item->data);      case OP_SOM:
3261      break;      case OP_SET_SOM:
3262      }      case OP_NOT_WORD_BOUNDARY:
3263    JUMPTO(SLJIT_JUMP, list_item->leave);      case OP_WORD_BOUNDARY:
3264    list_item = list_item->next;      case OP_EODN:
3265    }      case OP_EOD:
3266  common->stubs = NULL;      case OP_CIRC:
3267  }      case OP_CIRCM:
3268        case OP_DOLL:
3269        case OP_DOLLM:
3270        /* Zero width assertions. */
3271        cc++;
3272        continue;
3273    
3274  static SLJIT_INLINE void decrease_call_count(compiler_common *common)      case OP_ASSERT:
3275  {      case OP_ASSERT_NOT:
3276  DEFINE_COMPILER;      case OP_ASSERTBACK:
3277        case OP_ASSERTBACK_NOT:
3278        cc = bracketend(cc);
3279        continue;
3280    
3281  OP2(SLJIT_SUB | SLJIT_SET_E, CALL_COUNT, 0, CALL_COUNT, 0, SLJIT_IMM, 1);      case OP_PLUSI:
3282  add_jump(compiler, &common->calllimit, JUMP(SLJIT_C_ZERO));      case OP_MINPLUSI:
3283  }      case OP_POSPLUSI:
3284        caseless = TRUE;
3285        case OP_PLUS:
3286        case OP_MINPLUS:
3287        case OP_POSPLUS:
3288        cc++;
3289        break;
3290    
3291  static SLJIT_INLINE void allocate_stack(compiler_common *common, int size)      case OP_EXACTI:
3292  {      caseless = TRUE;
3293  /* May destroy all locals and registers except TMP2. */      case OP_EXACT:
3294  DEFINE_COMPILER;      repeat = GET2(cc, 1);
3295        last = FALSE;
3296        cc += 1 + IMM2_SIZE;
3297        break;
3298    
3299  OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_w));      case OP_QUERYI:
3300  #ifdef DESTROY_REGISTERS      case OP_MINQUERYI:
3301  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);      case OP_POSQUERYI:
3302  OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);      caseless = TRUE;
3303  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);      case OP_QUERY:
3304  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);      case OP_MINQUERY:
3305  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);      case OP_POSQUERY:
3306        len = 1;
3307        cc++;
3308    #ifdef SUPPORT_UTF
3309        if (common->utf && HAS_EXTRALEN(*cc)) len += GET_EXTRALEN(*cc);
3310  #endif  #endif
3311  add_stub(common, stack_alloc, 0, CMP(SLJIT_C_GREATER, STACK_TOP, 0, STACK_LIMIT, 0));      max_chars = scan_prefix(common, cc + len, chars, bytes, max_chars, rec_count);
3312  }      if (max_chars == 0)
3313          return consumed;
3314        last = FALSE;
3315        break;
3316    
3317  static SLJIT_INLINE void free_stack(compiler_common *common, int size)      case OP_KET:
3318  {      cc += 1 + LINK_SIZE;
3319  DEFINE_COMPILER;      continue;
 OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_w));  
 }  
3320    
3321  static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)      case OP_ALT:
3322  {      cc += GET(cc, 1);
3323  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, SLJIT_MEM1(SLJIT_LOCALS_REG), MAX_INDEX, 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);  
   }  
 }  
3324    
3325  static SLJIT_INLINE void copy_ovector(compiler_common *common)      case OP_ONCE:
3326  {      case OP_ONCE_NC:
3327  DEFINE_COMPILER;      case OP_BRA:
3328  struct sljit_label *loop;      case OP_BRAPOS:
3329  struct sljit_jump *earlyexit;      case OP_CBRA:
3330        case OP_CBRAPOS:
3331        alternative = cc + GET(cc, 1);
3332        while (*alternative == OP_ALT)
3333          {
3334          max_chars = scan_prefix(common, alternative + 1 + LINK_SIZE, chars, bytes, max_chars, rec_count);
3335          if (max_chars == 0)
3336            return consumed;
3337          alternative += GET(alternative, 1);
3338          }
3339    
3340  /* At this point we can freely use all registers. */      if (*cc == OP_CBRA || *cc == OP_CBRAPOS)
3341  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, ARGUMENTS, 0);        cc += IMM2_SIZE;
3342  OP1(SLJIT_MOV_SI, SLJIT_TEMPORARY_REG2, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, offsetcount));      cc += 1 + LINK_SIZE;
3343  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);  
 }  
3344    
3345  static SLJIT_INLINE BOOL char_has_othercase(compiler_common *common, uschar* cc)      case OP_CLASS:
3346  {  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3347  /* Detects if the character has an othercase. */      if (common->utf && !is_char7_bitset((const pcre_uint8 *)(cc + 1), FALSE)) return consumed;
3348  unsigned int c;  #endif
3349        any = TRUE;
3350        cc += 1 + 32 / sizeof(pcre_uchar);
3351        break;
3352    
3353  #ifdef SUPPORT_UTF8      case OP_NCLASS:
3354  if (common->utf8)  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3355    {      if (common->utf) return consumed;
   GETCHAR(c, cc);  
   if (c > 127)  
     {  
 #ifdef SUPPORT_UCP  
     return c != UCD_OTHERCASE(c);  
 #else  
     return FALSE;  
3356  #endif  #endif
3357      }      any = TRUE;
3358    }      cc += 1 + 32 / sizeof(pcre_uchar);
3359  else      break;
3360    
3361    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3362     &