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

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

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

revision 715 by zherczeg, Sat Oct 1 06:42:38 2011 UTC revision 1435 by zherczeg, Tue Jan 7 07:47:12 2014 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) (PUBL(malloc))(size)
56    #define SLJIT_FREE(ptr) (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;    sljit_uw *read_only_data[JIT_NUMBER_OF_COMPILE_MODES];
183      sljit_uw executable_sizes[JIT_NUMBER_OF_COMPILE_MODES];
184      PUBL(jit_callback) callback;
185    void *userdata;    void *userdata;
186  } executable_function;    pcre_uint32 top_bracket;
187      pcre_uint32 limit_match;
188    } executable_functions;
189    
190  typedef struct jump_list {  typedef struct jump_list {
191    struct sljit_jump *jump;    struct sljit_jump *jump;
192    struct jump_list *next;    struct jump_list *next;
193  } jump_list;  } jump_list;
194    
 enum stub_types { stack_alloc };  
   
195  typedef struct stub_list {  typedef struct stub_list {
   enum stub_types type;  
   int data;  
196    struct sljit_jump *start;    struct sljit_jump *start;
197    struct sljit_label *leave;    struct sljit_label *quit;
198    struct stub_list *next;    struct stub_list *next;
199  } stub_list;  } stub_list;
200    
201    typedef struct label_addr_list {
202      struct sljit_label *label;
203      sljit_uw *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;    /* This read-only data is available during runtime. */
326    int cbraptr;    sljit_uw *read_only_data;
327      /* The total size of the read-only data. */
328      sljit_uw read_only_data_size;
329      /* The next free entry of the read_only_data. */
330      sljit_uw *read_only_data_ptr;
331      /* Tells whether the capturing bracket is optimized. */
332      pcre_uint8 *optimized_cbracket;
333      /* Tells whether the starting offset is a target of then. */
334      pcre_uint8 *then_offsets;
335      /* Current position where a THEN must jump. */
336      then_trap_backtrack *then_trap;
337      /* Starting offset of private data for capturing brackets. */
338      int cbra_ptr;
339      /* Output vector starting point. Must be divisible by 2. */
340      int ovector_start;
341      /* Last known position of the requested byte. */
342      int req_char_ptr;
343      /* Head of the last recursion. */
344      int recursive_head_ptr;
345      /* First inspected character for partial matching. */
346      int start_used_ptr;
347      /* Starting pointer for partial soft matches. */
348      int hit_start;
349      /* End pointer of the first line. */
350      int first_line_end;
351      /* Points to the marked string. */
352      int mark_ptr;
353      /* Recursive control verb management chain. */
354      int control_head_ptr;
355      /* Points to the last matched capture block index. */
356      int capture_last_ptr;
357      /* Points to the starting position of the current match. */
358      int start_ptr;
359    
360      /* Flipped and lower case tables. */
361      const pcre_uint8 *fcc;
362      sljit_sw lcc;
363      /* Mode can be PCRE_STUDY_JIT_COMPILE and others. */
364      int mode;
365      /* TRUE, when minlength is greater than 0. */
366      BOOL might_be_empty;
367      /* \K is found in the pattern. */
368      BOOL has_set_som;
369      /* (*SKIP:arg) is found in the pattern. */
370      BOOL has_skip_arg;
371      /* (*THEN) is found in the pattern. */
372      BOOL has_then;
373      /* Needs to know the start position anytime. */
374      BOOL needs_start_ptr;
375      /* Currently in recurse or negative assert. */
376      BOOL local_exit;
377      /* Currently in a positive assert. */
378      BOOL positive_assert;
379      /* Newline control. */
380    int nltype;    int nltype;
381      pcre_uint32 nlmax;
382      pcre_uint32 nlmin;
383    int newline;    int newline;
384    int bsr_nltype;    int bsr_nltype;
385      pcre_uint32 bsr_nlmax;
386      pcre_uint32 bsr_nlmin;
387      /* Dollar endonly. */
388    int endonly;    int endonly;
389    sljit_w ctypes;    /* Tables. */
390    struct sljit_label *acceptlabel;    sljit_sw ctypes;
391      /* Named capturing brackets. */
392      pcre_uchar *name_table;
393      sljit_sw name_count;
394      sljit_sw name_entry_size;
395    
396      /* Labels and jump lists. */
397      struct sljit_label *partialmatchlabel;
398      struct sljit_label *quit_label;
399      struct sljit_label *forced_quit_label;
400      struct sljit_label *accept_label;
401    stub_list *stubs;    stub_list *stubs;
402      label_addr_list *label_addrs;
403    recurse_entry *entries;    recurse_entry *entries;
404    recurse_entry *currententry;    recurse_entry *currententry;
405      jump_list *partialmatch;
406      jump_list *quit;
407      jump_list *positive_assert_quit;
408      jump_list *forced_quit;
409    jump_list *accept;    jump_list *accept;
410    jump_list *calllimit;    jump_list *calllimit;
411    jump_list *stackalloc;    jump_list *stackalloc;
# Line 290  typedef struct compiler_common { Line 416  typedef struct compiler_common {
416    jump_list *vspace;    jump_list *vspace;
417    jump_list *casefulcmp;    jump_list *casefulcmp;
418    jump_list *caselesscmp;    jump_list *caselesscmp;
419      jump_list *reset_match;
420    BOOL jscript_compat;    BOOL jscript_compat;
421  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
422    BOOL utf8;    BOOL utf;
423  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
424    BOOL useucp;    BOOL use_ucp;
425  #endif  #endif
426    jump_list *utf8readchar;  #ifdef COMPILE_PCRE8
427    jump_list *utf8readtype8;    jump_list *utfreadchar;
428      jump_list *utfreadchar16;
429      jump_list *utfreadtype8;
430  #endif  #endif
431    #endif /* SUPPORT_UTF */
432  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
433    jump_list *getucd;    jump_list *getucd;
434  #endif  #endif
# Line 310  typedef struct compare_context { Line 440  typedef struct compare_context {
440    int length;    int length;
441    int sourcereg;    int sourcereg;
442  #if defined SLJIT_UNALIGNED && SLJIT_UNALIGNED  #if defined SLJIT_UNALIGNED && SLJIT_UNALIGNED
443    int byteptr;    int ucharptr;
444    union {    union {
445      int asint;      sljit_si asint;
446      short asshort;      sljit_uh asushort;
447    #if defined COMPILE_PCRE8
448      sljit_ub asbyte;      sljit_ub asbyte;
449      sljit_ub asbytes[4];      sljit_ub asuchars[4];
450    #elif defined COMPILE_PCRE16
451        sljit_uh asuchars[2];
452    #elif defined COMPILE_PCRE32
453        sljit_ui asuchars[1];
454    #endif
455    } c;    } c;
456    union {    union {
457      int asint;      sljit_si asint;
458      short asshort;      sljit_uh asushort;
459    #if defined COMPILE_PCRE8
460      sljit_ub asbyte;      sljit_ub asbyte;
461      sljit_ub asbytes[4];      sljit_ub asuchars[4];
462    #elif defined COMPILE_PCRE16
463        sljit_uh asuchars[2];
464    #elif defined COMPILE_PCRE32
465        sljit_ui asuchars[1];
466    #endif
467    } oc;    } oc;
468  #endif  #endif
469  } compare_context;  } compare_context;
470    
471  enum {  /* Undefine sljit macros. */
472    frame_end = 0,  #undef CMP
   frame_setstrbegin = -1  
 };  
473    
474  /* Used for accessing the elements of the stack. */  /* Used for accessing the elements of the stack. */
475  #define STACK(i)      ((-(i) - 1) * (int)sizeof(sljit_w))  #define STACK(i)      ((-(i) - 1) * (int)sizeof(sljit_sw))
476    
477  #define TMP1          SLJIT_TEMPORARY_REG1  #define TMP1          SLJIT_SCRATCH_REG1
478  #define TMP2          SLJIT_TEMPORARY_REG3  #define TMP2          SLJIT_SCRATCH_REG3
479  #define TMP3          SLJIT_TEMPORARY_EREG2  #define TMP3          SLJIT_TEMPORARY_EREG2
480  #define STR_PTR       SLJIT_GENERAL_REG1  #define STR_PTR       SLJIT_SAVED_REG1
481  #define STR_END       SLJIT_GENERAL_REG2  #define STR_END       SLJIT_SAVED_REG2
482  #define STACK_TOP     SLJIT_TEMPORARY_REG2  #define STACK_TOP     SLJIT_SCRATCH_REG2
483  #define STACK_LIMIT   SLJIT_GENERAL_REG3  #define STACK_LIMIT   SLJIT_SAVED_REG3
484  #define ARGUMENTS     SLJIT_GENERAL_EREG1  #define ARGUMENTS     SLJIT_SAVED_EREG1
485  #define CALL_COUNT    SLJIT_GENERAL_EREG2  #define COUNT_MATCH   SLJIT_SAVED_EREG2
486  #define RETURN_ADDR   SLJIT_TEMPORARY_EREG1  #define RETURN_ADDR   SLJIT_TEMPORARY_EREG1
487    
488  /* Locals layout. */  /* Local space layout. */
489  /* These two locals can be used by the current opcode. */  /* These two locals can be used by the current opcode. */
490  #define LOCALS0          (0 * sizeof(sljit_w))  #define LOCALS0          (0 * sizeof(sljit_sw))
491  #define LOCALS1          (1 * sizeof(sljit_w))  #define LOCALS1          (1 * sizeof(sljit_sw))
492  /* Two local variables for possessive quantifiers (char1 cannot use them). */  /* Two local variables for possessive quantifiers (char1 cannot use them). */
493  #define POSSESSIVE0      (2 * sizeof(sljit_w))  #define POSSESSIVE0      (2 * sizeof(sljit_sw))
494  #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))  
495  /* Max limit of recursions. */  /* Max limit of recursions. */
496  #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))  
497  /* The output vector is stored on the stack, and contains pointers  /* The output vector is stored on the stack, and contains pointers
498  to characters. The vector data is divided into two groups: the first  to characters. The vector data is divided into two groups: the first
499  group contains the start / end character pointers, and the second is  group contains the start / end character pointers, and the second is
500  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. */
501  #define OVECTOR_START    (10 * sizeof(sljit_w))  #define OVECTOR_START    (common->ovector_start)
502  #define OVECTOR(i)       (OVECTOR_START + (i) * sizeof(sljit_w))  #define OVECTOR(i)       (OVECTOR_START + (i) * (sljit_sw)sizeof(sljit_sw))
503  #define OVECTOR_PRIV(i)  (common->cbraptr + (i) * sizeof(sljit_w))  #define OVECTOR_PRIV(i)  (common->cbra_ptr + (i) * (sljit_sw)sizeof(sljit_sw))
504  #define PRIV(cc)         (common->localptrs[(cc) - common->start])  #define PRIVATE_DATA(cc) (common->private_data_ptrs[(cc) - common->start])
505    
506    #if defined COMPILE_PCRE8
507    #define MOV_UCHAR  SLJIT_MOV_UB
508    #define MOVU_UCHAR SLJIT_MOVU_UB
509    #elif defined COMPILE_PCRE16
510    #define MOV_UCHAR  SLJIT_MOV_UH
511    #define MOVU_UCHAR SLJIT_MOVU_UH
512    #elif defined COMPILE_PCRE32
513    #define MOV_UCHAR  SLJIT_MOV_UI
514    #define MOVU_UCHAR SLJIT_MOVU_UI
515    #else
516    #error Unsupported compiling mode
517    #endif
518    
519  /* Shortcuts. */  /* Shortcuts. */
520  #define DEFINE_COMPILER \  #define DEFINE_COMPILER \
# Line 386  the start pointers when the end of the c Line 531  the start pointers when the end of the c
531    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))
532  #define JUMPHERE(jump) \  #define JUMPHERE(jump) \
533    sljit_set_label((jump), sljit_emit_label(compiler))    sljit_set_label((jump), sljit_emit_label(compiler))
534    #define SET_LABEL(jump, label) \
535      sljit_set_label((jump), (label))
536  #define CMP(type, src1, src1w, src2, src2w) \  #define CMP(type, src1, src1w, src2, src2w) \
537    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))
538  #define CMPTO(type, src1, src1w, src2, src2w, label) \  #define CMPTO(type, src1, src1w, src2, src2w, label) \
539    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))
540  #define COND_VALUE(op, dst, dstw, type) \  #define OP_FLAGS(op, dst, dstw, src, srcw, type) \
541    sljit_emit_cond_value(compiler, (op), (dst), (dstw), (type))    sljit_emit_op_flags(compiler, (op), (dst), (dstw), (src), (srcw), (type))
542    #define GET_LOCAL_BASE(dst, dstw, offset) \
543      sljit_get_local_base(compiler, (dst), (dstw), (offset))
544    
545    #define READ_CHAR_MAX 0x7fffffff
546    
547  static uschar* bracketend(uschar* cc)  static pcre_uchar* bracketend(pcre_uchar* cc)
548  {  {
549  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));
550  do cc += GET(cc, 1); while (*cc == OP_ALT);  do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 402  cc += 1 + LINK_SIZE; Line 553  cc += 1 + LINK_SIZE;
553  return cc;  return cc;
554  }  }
555    
556    static int no_alternatives(pcre_uchar* cc)
557    {
558    int count = 0;
559    SLJIT_ASSERT((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND));
560    do
561      {
562      cc += GET(cc, 1);
563      count++;
564      }
565    while (*cc == OP_ALT);
566    SLJIT_ASSERT(*cc >= OP_KET && *cc <= OP_KETRPOS);
567    return count;
568    }
569    
570    static int ones_in_half_byte[16] = {
571      /* 0 */ 0, 1, 1, 2, /* 4 */ 1, 2, 2, 3,
572      /* 8 */ 1, 2, 2, 3, /* 12 */ 2, 3, 3, 4
573    };
574    
575  /* Functions whose might need modification for all new supported opcodes:  /* Functions whose might need modification for all new supported opcodes:
576   next_opcode   next_opcode
577   get_localspace   check_opcode_types
578   set_localptrs   set_private_data_ptrs
579   get_framesize   get_framesize
580   init_frame   init_frame
581   get_localsize   get_private_data_copy_length
582   copy_locals   copy_private_data
583   compile_hotpath   compile_matchingpath
584   compile_fallbackpath   compile_backtrackingpath
585  */  */
586    
587  static uschar *next_opcode(compiler_common *common, uschar *cc)  static pcre_uchar *next_opcode(compiler_common *common, pcre_uchar *cc)
588  {  {
589  SLJIT_UNUSED_ARG(common);  SLJIT_UNUSED_ARG(common);
590  switch(*cc)  switch(*cc)
# Line 432  switch(*cc) Line 602  switch(*cc)
602    case OP_WORDCHAR:    case OP_WORDCHAR:
603    case OP_ANY:    case OP_ANY:
604    case OP_ALLANY:    case OP_ALLANY:
605      case OP_NOTPROP:
606      case OP_PROP:
607    case OP_ANYNL:    case OP_ANYNL:
608    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
609    case OP_HSPACE:    case OP_HSPACE:
# Line 444  switch(*cc) Line 616  switch(*cc)
616    case OP_CIRCM:    case OP_CIRCM:
617    case OP_DOLL:    case OP_DOLL:
618    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:  
619    case OP_CRSTAR:    case OP_CRSTAR:
620    case OP_CRMINSTAR:    case OP_CRMINSTAR:
621    case OP_CRPLUS:    case OP_CRPLUS:
622    case OP_CRMINPLUS:    case OP_CRMINPLUS:
623    case OP_CRQUERY:    case OP_CRQUERY:
624    case OP_CRMINQUERY:    case OP_CRMINQUERY:
625      case OP_CRRANGE:
626      case OP_CRMINRANGE:
627      case OP_CRPOSSTAR:
628      case OP_CRPOSPLUS:
629      case OP_CRPOSQUERY:
630      case OP_CRPOSRANGE:
631      case OP_CLASS:
632      case OP_NCLASS:
633      case OP_REF:
634      case OP_REFI:
635      case OP_DNREF:
636      case OP_DNREFI:
637      case OP_RECURSE:
638      case OP_CALLOUT:
639      case OP_ALT:
640      case OP_KET:
641      case OP_KETRMAX:
642      case OP_KETRMIN:
643      case OP_KETRPOS:
644      case OP_REVERSE:
645      case OP_ASSERT:
646      case OP_ASSERT_NOT:
647      case OP_ASSERTBACK:
648      case OP_ASSERTBACK_NOT:
649      case OP_ONCE:
650      case OP_ONCE_NC:
651      case OP_BRA:
652      case OP_BRAPOS:
653      case OP_CBRA:
654      case OP_CBRAPOS:
655      case OP_COND:
656      case OP_SBRA:
657      case OP_SBRAPOS:
658      case OP_SCBRA:
659      case OP_SCBRAPOS:
660      case OP_SCOND:
661      case OP_CREF:
662      case OP_DNCREF:
663      case OP_RREF:
664      case OP_DNRREF:
665    case OP_DEF:    case OP_DEF:
666    case OP_BRAZERO:    case OP_BRAZERO:
667    case OP_BRAMINZERO:    case OP_BRAMINZERO:
668    case OP_BRAPOSZERO:    case OP_BRAPOSZERO:
669      case OP_PRUNE:
670      case OP_SKIP:
671      case OP_THEN:
672      case OP_COMMIT:
673    case OP_FAIL:    case OP_FAIL:
674    case OP_ACCEPT:    case OP_ACCEPT:
675    case OP_ASSERT_ACCEPT:    case OP_ASSERT_ACCEPT:
676      case OP_CLOSE:
677    case OP_SKIPZERO:    case OP_SKIPZERO:
678    return cc + 1;    return cc + PRIV(OP_lengths)[*cc];
679    
680    case OP_CHAR:    case OP_CHAR:
681    case OP_CHARI:    case OP_CHARI:
682    case OP_NOT:    case OP_NOT:
683    case OP_NOTI:    case OP_NOTI:
   
684    case OP_STAR:    case OP_STAR:
685    case OP_MINSTAR:    case OP_MINSTAR:
686    case OP_PLUS:    case OP_PLUS:
687    case OP_MINPLUS:    case OP_MINPLUS:
688    case OP_QUERY:    case OP_QUERY:
689    case OP_MINQUERY:    case OP_MINQUERY:
690      case OP_UPTO:
691      case OP_MINUPTO:
692      case OP_EXACT:
693    case OP_POSSTAR:    case OP_POSSTAR:
694    case OP_POSPLUS:    case OP_POSPLUS:
695    case OP_POSQUERY:    case OP_POSQUERY:
696      case OP_POSUPTO:
697    case OP_STARI:    case OP_STARI:
698    case OP_MINSTARI:    case OP_MINSTARI:
699    case OP_PLUSI:    case OP_PLUSI:
700    case OP_MINPLUSI:    case OP_MINPLUSI:
701    case OP_QUERYI:    case OP_QUERYI:
702    case OP_MINQUERYI:    case OP_MINQUERYI:
703      case OP_UPTOI:
704      case OP_MINUPTOI:
705      case OP_EXACTI:
706    case OP_POSSTARI:    case OP_POSSTARI:
707    case OP_POSPLUSI:    case OP_POSPLUSI:
708    case OP_POSQUERYI:    case OP_POSQUERYI:
709      case OP_POSUPTOI:
710    case OP_NOTSTAR:    case OP_NOTSTAR:
711    case OP_NOTMINSTAR:    case OP_NOTMINSTAR:
712    case OP_NOTPLUS:    case OP_NOTPLUS:
713    case OP_NOTMINPLUS:    case OP_NOTMINPLUS:
714    case OP_NOTQUERY:    case OP_NOTQUERY:
715    case OP_NOTMINQUERY:    case OP_NOTMINQUERY:
716      case OP_NOTUPTO:
717      case OP_NOTMINUPTO:
718      case OP_NOTEXACT:
719    case OP_NOTPOSSTAR:    case OP_NOTPOSSTAR:
720    case OP_NOTPOSPLUS:    case OP_NOTPOSPLUS:
721    case OP_NOTPOSQUERY:    case OP_NOTPOSQUERY:
722      case OP_NOTPOSUPTO:
723    case OP_NOTSTARI:    case OP_NOTSTARI:
724    case OP_NOTMINSTARI:    case OP_NOTMINSTARI:
725    case OP_NOTPLUSI:    case OP_NOTPLUSI:
726    case OP_NOTMINPLUSI:    case OP_NOTMINPLUSI:
727    case OP_NOTQUERYI:    case OP_NOTQUERYI:
728    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:  
729    case OP_NOTUPTOI:    case OP_NOTUPTOI:
730    case OP_NOTMINUPTOI:    case OP_NOTMINUPTOI:
731    case OP_NOTEXACTI:    case OP_NOTEXACTI:
732      case OP_NOTPOSSTARI:
733      case OP_NOTPOSPLUSI:
734      case OP_NOTPOSQUERYI:
735    case OP_NOTPOSUPTOI:    case OP_NOTPOSUPTOI:
736    cc += 4;    cc += PRIV(OP_lengths)[*cc];
737  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
738    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]);
739  #endif  #endif
740    return cc;    return cc;
741    
742    case OP_NOTPROP:    /* Special cases. */
743    case OP_PROP:    case OP_TYPESTAR:
744      case OP_TYPEMINSTAR:
745      case OP_TYPEPLUS:
746      case OP_TYPEMINPLUS:
747      case OP_TYPEQUERY:
748      case OP_TYPEMINQUERY:
749    case OP_TYPEUPTO:    case OP_TYPEUPTO:
750    case OP_TYPEMINUPTO:    case OP_TYPEMINUPTO:
751    case OP_TYPEEXACT:    case OP_TYPEEXACT:
752      case OP_TYPEPOSSTAR:
753      case OP_TYPEPOSPLUS:
754      case OP_TYPEPOSQUERY:
755    case OP_TYPEPOSUPTO:    case OP_TYPEPOSUPTO:
756    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;  
757    
758    case OP_CLASS:    case OP_ANYBYTE:
759    case OP_NCLASS:  #ifdef SUPPORT_UTF
760    return cc + 33;    if (common->utf) return NULL;
761    #endif
762      return cc + 1;
763    
764  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
765    case OP_XCLASS:    case OP_XCLASS:
766    return cc + GET(cc, 1);    return cc + GET(cc, 1);
767  #endif  #endif
768    
769    case OP_RECURSE:    case OP_MARK:
770    case OP_ASSERT:    case OP_PRUNE_ARG:
771    case OP_ASSERT_NOT:    case OP_SKIP_ARG:
772    case OP_ASSERTBACK:    case OP_THEN_ARG:
773    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;  
774    
775    default:    default:
776      /* All opcodes are supported now! */
777      SLJIT_ASSERT_STOP();
778    return NULL;    return NULL;
779    }    }
780  }  }
781    
782  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)
783  {  {
784  int localspace = 0;  int count;
785  uschar *alternative;  pcre_uchar *slot;
786    
787  /* 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. */
788  while (cc < ccend)  while (cc < ccend)
789    {    {
790    switch(*cc)    switch(*cc)
791      {      {
792      case OP_ASSERT:      case OP_SET_SOM:
793      case OP_ASSERT_NOT:      common->has_set_som = TRUE;
794      case OP_ASSERTBACK:      common->might_be_empty = TRUE;
795      case OP_ASSERTBACK_NOT:      cc += 1;
     case OP_ONCE:  
     case OP_BRAPOS:  
     case OP_SBRA:  
     case OP_SBRAPOS:  
     case OP_SCOND:  
     localspace += sizeof(sljit_w);  
     cc += 1 + LINK_SIZE;  
796      break;      break;
797    
798      case OP_CBRAPOS:      case OP_REF:
799      case OP_SCBRAPOS:      case OP_REFI:
800      localspace += sizeof(sljit_w);      common->optimized_cbracket[GET2(cc, 1)] = 0;
801      cc += 1 + LINK_SIZE + 2;      cc += 1 + IMM2_SIZE;
802      break;      break;
803    
804      case OP_COND:      case OP_BRA:
805      /* Might be a hidden SCOND. */      case OP_CBRA:
806      alternative = cc + GET(cc, 1);      case OP_SBRA:
807      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)      case OP_SCBRA:
808        localspace += sizeof(sljit_w);      count = no_alternatives(cc);
809      cc += 1 + LINK_SIZE;      if (count > 4)
810          common->read_only_data_size += count * sizeof(sljit_uw);
811        cc += 1 + LINK_SIZE + (*cc == OP_CBRA || *cc == OP_SCBRA ? IMM2_SIZE : 0);
812      break;      break;
813    
814      default:      case OP_CBRAPOS:
815      cc = next_opcode(common, cc);      case OP_SCBRAPOS:
816      if (cc == NULL)      common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] = 0;
817        return -1;      cc += 1 + LINK_SIZE + IMM2_SIZE;
818      break;      break;
     }  
   }  
 return localspace;  
 }  
819    
820  static void set_localptrs(compiler_common *common, int localptr, uschar *ccend)      case OP_COND:
 {  
 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:  
821      case OP_SCOND:      case OP_SCOND:
822      common->localptrs[cc - common->start] = localptr;      /* Only AUTO_CALLOUT can insert this opcode. We do
823      localptr += sizeof(sljit_w);         not intend to support this case. */
824        if (cc[1 + LINK_SIZE] == OP_CALLOUT)
825          return FALSE;
826      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
827      break;      break;
828    
829      case OP_CBRAPOS:      case OP_CREF:
830      case OP_SCBRAPOS:      common->optimized_cbracket[GET2(cc, 1)] = 0;
831      common->localptrs[cc - common->start] = localptr;      cc += 1 + IMM2_SIZE;
832      localptr += sizeof(sljit_w);      break;
833      cc += 1 + LINK_SIZE + 2;  
834        case OP_DNREF:
835        case OP_DNREFI:
836        case OP_DNCREF:
837        count = GET2(cc, 1 + IMM2_SIZE);
838        slot = common->name_table + GET2(cc, 1) * common->name_entry_size;
839        while (count-- > 0)
840          {
841          common->optimized_cbracket[GET2(slot, 0)] = 0;
842          slot += common->name_entry_size;
843          }
844        cc += 1 + 2 * IMM2_SIZE;
845      break;      break;
846    
847      case OP_COND:      case OP_RECURSE:
848      /* Might be a hidden SCOND. */      /* Set its value only once. */
849      alternative = cc + GET(cc, 1);      if (common->recursive_head_ptr == 0)
     if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)  
850        {        {
851        common->localptrs[cc - common->start] = localptr;        common->recursive_head_ptr = common->ovector_start;
852        localptr += sizeof(sljit_w);        common->ovector_start += sizeof(sljit_sw);
853        }        }
854      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
855      break;      break;
856    
857      default:      case OP_CALLOUT:
858      cc = next_opcode(common, cc);      if (common->capture_last_ptr == 0)
859      SLJIT_ASSERT(cc != NULL);        {
860          common->capture_last_ptr = common->ovector_start;
861          common->ovector_start += sizeof(sljit_sw);
862          }
863        cc += 2 + 2 * LINK_SIZE;
864      break;      break;
     }  
   }  
 }  
865    
866  /* Returns with -1 if no need for frame. */      case OP_THEN_ARG:
867  static int get_framesize(compiler_common *common, uschar *cc, BOOL recursive)      common->has_then = TRUE;
868  {      common->control_head_ptr = 1;
869  uschar *ccend = bracketend(cc);      /* Fall through. */
 uschar *end;  
 int length = 0;  
 BOOL possessive = FALSE;  
 BOOL needs_frame = FALSE;  
 BOOL setsom_found = FALSE;  
870    
871  if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))      case OP_PRUNE_ARG:
872    {      common->needs_start_ptr = TRUE;
873    length = 3;      /* Fall through. */
   possessive = TRUE;  
   }  
874    
875  cc = next_opcode(common, cc);      case OP_MARK:
876  SLJIT_ASSERT(cc != NULL);      if (common->mark_ptr == 0)
 while (cc < ccend)  
   switch(*cc)  
     {  
     case OP_SET_SOM:  
     case OP_RECURSE:  
     if (!setsom_found)  
877        {        {
878        length += 2;        common->mark_ptr = common->ovector_start;
879        setsom_found = TRUE;        common->ovector_start += sizeof(sljit_sw);
880        }        }
881      cc += (*cc == OP_SET_SOM) ? 1 : 1 + LINK_SIZE;      cc += 1 + 2 + cc[1];
882      break;      break;
883    
884      case OP_ASSERT:      case OP_THEN:
885      case OP_ASSERT_NOT:      common->has_then = TRUE;
886      case OP_ASSERTBACK:      common->control_head_ptr = 1;
887      case OP_ASSERTBACK_NOT:      /* Fall through. */
888      case OP_ONCE:  
889      if (needs_frame || length > 0)      case OP_PRUNE:
890        {      case OP_SKIP:
891        cc = bracketend(cc);      common->needs_start_ptr = TRUE;
892        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);  
       }  
893      break;      break;
894    
895      case OP_CBRA:      case OP_SKIP_ARG:
896      case OP_CBRAPOS:      common->control_head_ptr = 1;
897      case OP_SCBRA:      common->has_skip_arg = TRUE;
898      case OP_SCBRAPOS:      cc += 1 + 2 + cc[1];
     length += 3;  
     cc += 1 + LINK_SIZE + 2;  
899      break;      break;
900    
901      default:      default:
902      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
903      SLJIT_ASSERT(cc != NULL);      if (cc == NULL)
904          return FALSE;
905      break;      break;
906      }      }
907      }
908  /* 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;  
909  }  }
910    
911  static void init_frame(compiler_common *common, uschar *cc, int stackpos, int stacktop, BOOL recursive)  static int get_class_iterator_size(pcre_uchar *cc)
912  {  {
913  /* TMP2 must contain STACK_TOP - (-STACK(stackpos)) */  switch(*cc)
914  DEFINE_COMPILER;    {
915  uschar *ccend = bracketend(cc);    case OP_CRSTAR:
916  BOOL setsom_found = FALSE;    case OP_CRPLUS:
917  int offset;    return 2;
918    
919      case OP_CRMINSTAR:
920      case OP_CRMINPLUS:
921      case OP_CRQUERY:
922      case OP_CRMINQUERY:
923      return 1;
924    
925      case OP_CRRANGE:
926      case OP_CRMINRANGE:
927      if (GET2(cc, 1) == GET2(cc, 1 + IMM2_SIZE))
928        return 0;
929      return 2;
930    
931      default:
932      return 0;
933      }
934    }
935    
936  if (stackpos < stacktop)  static BOOL detect_repeat(compiler_common *common, pcre_uchar *begin)
937    {
938    pcre_uchar *end = bracketend(begin);
939    pcre_uchar *next;
940    pcre_uchar *next_end;
941    pcre_uchar *max_end;
942    pcre_uchar type;
943    sljit_sw length = end - begin;
944    int min, max, i;
945    
946    /* Detect fixed iterations first. */
947    if (end[-(1 + LINK_SIZE)] != OP_KET)
948      return FALSE;
949    
950    /* Already detected repeat. */
951    if (common->private_data_ptrs[end - common->start - LINK_SIZE] != 0)
952      return TRUE;
953    
954    next = end;
955    min = 1;
956    while (1)
957    {    {
958    SLJIT_ASSERT(stackpos + 1 == stacktop);    if (*next != *begin)
959    return;      break;
960      next_end = bracketend(next);
961      if (next_end - next != length || memcmp(begin, next, IN_UCHARS(length)) != 0)
962        break;
963      next = next_end;
964      min++;
965    }    }
966    
967  stackpos = STACK(stackpos);  if (min == 2)
968  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);  
969    
970  if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))  max = 0;
971    cc = next_opcode(common, cc);  max_end = next;
972  SLJIT_ASSERT(cc != NULL);  if (*next == OP_BRAZERO || *next == OP_BRAMINZERO)
973  while (cc < ccend)    {
974    switch(*cc)    type = *next;
975      while (1)
976      {      {
977      case OP_SET_SOM:      if (next[0] != type || next[1] != OP_BRA || next[2 + LINK_SIZE] != *begin)
978      case OP_RECURSE:        break;
979      if (!setsom_found)      next_end = bracketend(next + 2 + LINK_SIZE);
980        if (next_end - next != (length + 2 + LINK_SIZE) || memcmp(begin, next + 2 + LINK_SIZE, IN_UCHARS(length)) != 0)
981          break;
982        next = next_end;
983        max++;
984        }
985    
986      if (next[0] == type && next[1] == *begin && max >= 1)
987        {
988        next_end = bracketend(next + 1);
989        if (next_end - next == (length + 1) && memcmp(begin, next + 1, IN_UCHARS(length)) == 0)
990        {        {
991        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));        for (i = 0; i < max; i++, next_end += 1 + LINK_SIZE)
992        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setstrbegin);          if (*next_end != OP_KET)
993        stackpos += (int)sizeof(sljit_w);            break;
994        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);  
995        stackpos += (int)sizeof(sljit_w);        if (i == max)
996        setsom_found = TRUE;          {
997            common->private_data_ptrs[max_end - common->start - LINK_SIZE] = next_end - max_end;
998            common->private_data_ptrs[max_end - common->start - LINK_SIZE + 1] = (type == OP_BRAZERO) ? OP_UPTO : OP_MINUPTO;
999            /* +2 the original and the last. */
1000            common->private_data_ptrs[max_end - common->start - LINK_SIZE + 2] = max + 2;
1001            if (min == 1)
1002              return TRUE;
1003            min--;
1004            max_end -= (1 + LINK_SIZE) + GET(max_end, -LINK_SIZE);
1005            }
1006        }        }
1007      cc += (*cc == OP_SET_SOM) ? 1 : 1 + LINK_SIZE;      }
1008      break;    }
1009    
1010      case OP_ASSERT:  if (min >= 3)
1011      case OP_ASSERT_NOT:    {
1012      case OP_ASSERTBACK:    common->private_data_ptrs[end - common->start - LINK_SIZE] = max_end - end;
1013      case OP_ASSERTBACK_NOT:    common->private_data_ptrs[end - common->start - LINK_SIZE + 1] = OP_EXACT;
1014      case OP_ONCE:    common->private_data_ptrs[end - common->start - LINK_SIZE + 2] = min;
1015      cc = bracketend(cc);    return TRUE;
1016      break;    }
1017    
1018    return FALSE;
1019    }
1020    
1021    #define CASE_ITERATOR_PRIVATE_DATA_1 \
1022        case OP_MINSTAR: \
1023        case OP_MINPLUS: \
1024        case OP_QUERY: \
1025        case OP_MINQUERY: \
1026        case OP_MINSTARI: \
1027        case OP_MINPLUSI: \
1028        case OP_QUERYI: \
1029        case OP_MINQUERYI: \
1030        case OP_NOTMINSTAR: \
1031        case OP_NOTMINPLUS: \
1032        case OP_NOTQUERY: \
1033        case OP_NOTMINQUERY: \
1034        case OP_NOTMINSTARI: \
1035        case OP_NOTMINPLUSI: \
1036        case OP_NOTQUERYI: \
1037        case OP_NOTMINQUERYI:
1038    
1039      case OP_CBRA:  #define CASE_ITERATOR_PRIVATE_DATA_2A \
1040      case OP_CBRAPOS:      case OP_STAR: \
1041      case OP_SCBRA:      case OP_PLUS: \
1042      case OP_SCBRAPOS:      case OP_STARI: \
1043      offset = (GET2(cc, 1 + LINK_SIZE)) << 1;      case OP_PLUSI: \
1044      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));      case OP_NOTSTAR: \
1045      stackpos += (int)sizeof(sljit_w);      case OP_NOTPLUS: \
1046      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset));      case OP_NOTSTARI: \
1047      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset + 1));      case OP_NOTPLUSI:
     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);  
1048    
1049      cc += 1 + LINK_SIZE + 2;  #define CASE_ITERATOR_PRIVATE_DATA_2B \
1050      break;      case OP_UPTO: \
1051        case OP_MINUPTO: \
1052        case OP_UPTOI: \
1053        case OP_MINUPTOI: \
1054        case OP_NOTUPTO: \
1055        case OP_NOTMINUPTO: \
1056        case OP_NOTUPTOI: \
1057        case OP_NOTMINUPTOI:
1058    
1059      default:  #define CASE_ITERATOR_TYPE_PRIVATE_DATA_1 \
1060      cc = next_opcode(common, cc);      case OP_TYPEMINSTAR: \
1061      SLJIT_ASSERT(cc != NULL);      case OP_TYPEMINPLUS: \
1062      break;      case OP_TYPEQUERY: \
1063      }      case OP_TYPEMINQUERY:
1064    
1065  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_end);  #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2A \
1066  SLJIT_ASSERT(stackpos == STACK(stacktop + 1));      case OP_TYPESTAR: \
1067  }      case OP_TYPEPLUS:
1068    
1069    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2B \
1070        case OP_TYPEUPTO: \
1071        case OP_TYPEMINUPTO:
1072    
1073  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)
1074  {  {
1075  int localsize = 2;  pcre_uchar *cc = common->start;
1076  uschar *alternative;  pcre_uchar *alternative;
1077  /* Calculate the sum of the local variables. */  pcre_uchar *end = NULL;
1078    int private_data_ptr = *private_data_start;
1079    int space, size, bracketlen;
1080    
1081  while (cc < ccend)  while (cc < ccend)
1082    {    {
1083      space = 0;
1084      size = 0;
1085      bracketlen = 0;
1086      if (private_data_ptr > SLJIT_MAX_LOCAL_SIZE)
1087        return;
1088    
1089      if (*cc == OP_ONCE || *cc == OP_ONCE_NC || *cc == OP_BRA || *cc == OP_CBRA || *cc == OP_COND)
1090        if (detect_repeat(common, cc))
1091          {
1092          /* These brackets are converted to repeats, so no global
1093          based single character repeat is allowed. */
1094          if (cc >= end)
1095            end = bracketend(cc);
1096          }
1097    
1098    switch(*cc)    switch(*cc)
1099      {      {
1100        case OP_KET:
1101        if (common->private_data_ptrs[cc + 1 - common->start] != 0)
1102          {
1103          common->private_data_ptrs[cc - common->start] = private_data_ptr;
1104          private_data_ptr += sizeof(sljit_sw);
1105          cc += common->private_data_ptrs[cc + 1 - common->start];
1106          }
1107        cc += 1 + LINK_SIZE;
1108        break;
1109    
1110      case OP_ASSERT:      case OP_ASSERT:
1111      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1112      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1113      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1114      case OP_ONCE:      case OP_ONCE:
1115        case OP_ONCE_NC:
1116      case OP_BRAPOS:      case OP_BRAPOS:
1117      case OP_SBRA:      case OP_SBRA:
1118      case OP_SBRAPOS:      case OP_SBRAPOS:
1119      case OP_SCOND:      case OP_SCOND:
1120      localsize++;      common->private_data_ptrs[cc - common->start] = private_data_ptr;
1121      cc += 1 + LINK_SIZE;      private_data_ptr += sizeof(sljit_sw);
1122      break;      bracketlen = 1 + LINK_SIZE;
   
     case OP_CBRA:  
     case OP_SCBRA:  
     localsize++;  
     cc += 1 + LINK_SIZE + 2;  
1123      break;      break;
1124    
1125      case OP_CBRAPOS:      case OP_CBRAPOS:
1126      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1127      localsize += 2;      common->private_data_ptrs[cc - common->start] = private_data_ptr;
1128      cc += 1 + LINK_SIZE + 2;      private_data_ptr += sizeof(sljit_sw);
1129        bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
1130      break;      break;
1131    
1132      case OP_COND:      case OP_COND:
1133      /* Might be a hidden SCOND. */      /* Might be a hidden SCOND. */
1134      alternative = cc + GET(cc, 1);      alternative = cc + GET(cc, 1);
1135      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1136        localsize++;        {
1137      cc += 1 + LINK_SIZE;        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1138          private_data_ptr += sizeof(sljit_sw);
1139          }
1140        bracketlen = 1 + LINK_SIZE;
1141        break;
1142    
1143        case OP_BRA:
1144        bracketlen = 1 + LINK_SIZE;
1145        break;
1146    
1147        case OP_CBRA:
1148        case OP_SCBRA:
1149        bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
1150        break;
1151    
1152        CASE_ITERATOR_PRIVATE_DATA_1
1153        space = 1;
1154        size = -2;
1155        break;
1156    
1157        CASE_ITERATOR_PRIVATE_DATA_2A
1158        space = 2;
1159        size = -2;
1160        break;
1161    
1162        CASE_ITERATOR_PRIVATE_DATA_2B
1163        space = 2;
1164        size = -(2 + IMM2_SIZE);
1165      break;      break;
1166    
1167        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1168        space = 1;
1169        size = 1;
1170        break;
1171    
1172        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1173        if (cc[1] != OP_ANYNL && cc[1] != OP_EXTUNI)
1174          space = 2;
1175        size = 1;
1176        break;
1177    
1178        CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1179        if (cc[1 + IMM2_SIZE] != OP_ANYNL && cc[1 + IMM2_SIZE] != OP_EXTUNI)
1180          space = 2;
1181        size = 1 + IMM2_SIZE;
1182        break;
1183    
1184        case OP_CLASS:
1185        case OP_NCLASS:
1186        size += 1 + 32 / sizeof(pcre_uchar);
1187        space = get_class_iterator_size(cc + size);
1188        break;
1189    
1190    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1191        case OP_XCLASS:
1192        size = GET(cc, 1);
1193        space = get_class_iterator_size(cc + size);
1194        break;
1195    #endif
1196    
1197      default:      default:
1198      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
1199      SLJIT_ASSERT(cc != NULL);      SLJIT_ASSERT(cc != NULL);
1200      break;      break;
1201      }      }
1202    
1203      /* Character iterators, which are not inside a repeated bracket,
1204         gets a private slot instead of allocating it on the stack. */
1205      if (space > 0 && cc >= end)
1206        {
1207        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1208        private_data_ptr += sizeof(sljit_sw) * space;
1209        }
1210    
1211      if (size != 0)
1212        {
1213        if (size < 0)
1214          {
1215          cc += -size;
1216    #ifdef SUPPORT_UTF
1217          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1218    #endif
1219          }
1220        else
1221          cc += size;
1222        }
1223    
1224      if (bracketlen > 0)
1225        {
1226        if (cc >= end)
1227          {
1228          end = bracketend(cc);
1229          if (end[-1 - LINK_SIZE] == OP_KET)
1230            end = NULL;
1231          }
1232        cc += bracketlen;
1233        }
1234    }    }
1235  SLJIT_ASSERT(cc == ccend);  *private_data_start = private_data_ptr;
 return localsize;  
1236  }  }
1237    
1238  static void copy_locals(compiler_common *common, uschar *cc, uschar *ccend,  /* Returns with a frame_types (always < 0) if no need for frame. */
1239    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)
1240  {  {
1241  DEFINE_COMPILER;  int length = 0;
1242  int srcw[2];  int possessive = 0;
1243  int count;  BOOL stack_restore = FALSE;
1244  BOOL tmp1next = TRUE;  BOOL setsom_found = recursive;
1245  BOOL tmp1empty = TRUE;  BOOL setmark_found = recursive;
1246  BOOL tmp2empty = TRUE;  /* The last capture is a local variable even for recursions. */
1247  uschar *alternative;  BOOL capture_last_found = FALSE;
1248  enum {  
1249    start,  #if defined DEBUG_FORCE_CONTROL_HEAD && DEBUG_FORCE_CONTROL_HEAD
1250    loop,  SLJIT_ASSERT(common->control_head_ptr != 0);
1251    end  *needs_control_head = TRUE;
1252  } status;  #else
1253    *needs_control_head = FALSE;
1254  status = save ? start : loop;  #endif
 stackptr = STACK(stackptr - 2);  
 stacktop = STACK(stacktop - 1);  
1255    
1256  if (!save)  if (ccend == NULL)
1257    {    {
1258    stackptr += sizeof(sljit_w);    ccend = bracketend(cc) - (1 + LINK_SIZE);
1259    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)  
1260      {      {
1261      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);      possessive = length = (common->capture_last_ptr != 0) ? 5 : 3;
1262      stackptr += sizeof(sljit_w);      /* This is correct regardless of common->capture_last_ptr. */
1263      tmp2empty = FALSE;      capture_last_found = TRUE;
1264      }      }
1265    /* The tmp1next must be TRUE in either way. */    cc = next_opcode(common, cc);
1266    }    }
1267    
1268  while (status != end)  SLJIT_ASSERT(cc != NULL);
1269    {  while (cc < ccend)
1270    count = 0;    switch(*cc)
   switch(status)  
1271      {      {
1272      case start:      case OP_SET_SOM:
1273      SLJIT_ASSERT(save);      SLJIT_ASSERT(common->has_set_som);
1274      count = 1;      stack_restore = TRUE;
1275      srcw[0] = RECURSIVE_HEAD;      if (!setsom_found)
1276      status = loop;        {
1277          length += 2;
1278          setsom_found = TRUE;
1279          }
1280        cc += 1;
1281      break;      break;
1282    
1283      case loop:      case OP_MARK:
1284      if (cc >= ccend)      case OP_PRUNE_ARG:
1285        case OP_THEN_ARG:
1286        SLJIT_ASSERT(common->mark_ptr != 0);
1287        stack_restore = TRUE;
1288        if (!setmark_found)
1289        {        {
1290        status = end;        length += 2;
1291        break;        setmark_found = TRUE;
1292        }        }
1293        if (common->control_head_ptr != 0)
1294          *needs_control_head = TRUE;
1295        cc += 1 + 2 + cc[1];
1296        break;
1297    
1298      switch(*cc)      case OP_RECURSE:
1299        stack_restore = TRUE;
1300        if (common->has_set_som && !setsom_found)
1301        {        {
1302        case OP_ASSERT:        length += 2;
1303        case OP_ASSERT_NOT:        setsom_found = TRUE;
1304        case OP_ASSERTBACK:        }
1305        case OP_ASSERTBACK_NOT:      if (common->mark_ptr != 0 && !setmark_found)
1306        case OP_ONCE:        {
1307        case OP_BRAPOS:        length += 2;
1308        case OP_SBRA:        setmark_found = TRUE;
1309        case OP_SBRAPOS:        }
1310        case OP_SCOND:      if (common->capture_last_ptr != 0 && !capture_last_found)
1311        count = 1;        {
1312        srcw[0] = PRIV(cc);        length += 2;
1313        SLJIT_ASSERT(srcw[0] != 0);        capture_last_found = TRUE;
       cc += 1 + LINK_SIZE;  
       break;  
   
       case OP_CBRA:  
       case OP_SCBRA:  
       count = 1;  
       srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));  
       cc += 1 + LINK_SIZE + 2;  
       break;  
   
       case OP_CBRAPOS:  
       case OP_SCBRAPOS:  
       count = 2;  
       srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));  
       srcw[0] = PRIV(cc);  
       SLJIT_ASSERT(srcw[0] != 0);  
       cc += 1 + LINK_SIZE + 2;  
       break;  
   
       case OP_COND:  
       /* Might be a hidden SCOND. */  
       alternative = cc + GET(cc, 1);  
       if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)  
         {  
         count = 1;  
         srcw[0] = PRIV(cc);  
         SLJIT_ASSERT(srcw[0] != 0);  
         }  
       cc += 1 + LINK_SIZE;  
       break;  
   
       default:  
       cc = next_opcode(common, cc);  
       SLJIT_ASSERT(cc != NULL);  
       break;  
1314        }        }
1315        cc += 1 + LINK_SIZE;
1316      break;      break;
1317    
1318      case end:      case OP_CBRA:
1319      SLJIT_ASSERT_STOP();      case OP_CBRAPOS:
1320        case OP_SCBRA:
1321        case OP_SCBRAPOS:
1322        stack_restore = TRUE;
1323        if (common->capture_last_ptr != 0 && !capture_last_found)
1324          {
1325          length += 2;
1326          capture_last_found = TRUE;
1327          }
1328        length += 3;
1329        cc += 1 + LINK_SIZE + IMM2_SIZE;
1330        break;
1331    
1332        default:
1333        stack_restore = TRUE;
1334        /* Fall through. */
1335    
1336        case OP_NOT_WORD_BOUNDARY:
1337        case OP_WORD_BOUNDARY:
1338        case OP_NOT_DIGIT:
1339        case OP_DIGIT:
1340        case OP_NOT_WHITESPACE:
1341        case OP_WHITESPACE:
1342        case OP_NOT_WORDCHAR:
1343        case OP_WORDCHAR:
1344        case OP_ANY:
1345        case OP_ALLANY:
1346        case OP_ANYBYTE:
1347        case OP_NOTPROP:
1348        case OP_PROP:
1349        case OP_ANYNL:
1350        case OP_NOT_HSPACE:
1351        case OP_HSPACE:
1352        case OP_NOT_VSPACE:
1353        case OP_VSPACE:
1354        case OP_EXTUNI:
1355        case OP_EODN:
1356        case OP_EOD:
1357        case OP_CIRC:
1358        case OP_CIRCM:
1359        case OP_DOLL:
1360        case OP_DOLLM:
1361        case OP_CHAR:
1362        case OP_CHARI:
1363        case OP_NOT:
1364        case OP_NOTI:
1365    
1366        case OP_EXACT:
1367        case OP_POSSTAR:
1368        case OP_POSPLUS:
1369        case OP_POSQUERY:
1370        case OP_POSUPTO:
1371    
1372        case OP_EXACTI:
1373        case OP_POSSTARI:
1374        case OP_POSPLUSI:
1375        case OP_POSQUERYI:
1376        case OP_POSUPTOI:
1377    
1378        case OP_NOTEXACT:
1379        case OP_NOTPOSSTAR:
1380        case OP_NOTPOSPLUS:
1381        case OP_NOTPOSQUERY:
1382        case OP_NOTPOSUPTO:
1383    
1384        case OP_NOTEXACTI:
1385        case OP_NOTPOSSTARI:
1386        case OP_NOTPOSPLUSI:
1387        case OP_NOTPOSQUERYI:
1388        case OP_NOTPOSUPTOI:
1389    
1390        case OP_TYPEEXACT:
1391        case OP_TYPEPOSSTAR:
1392        case OP_TYPEPOSPLUS:
1393        case OP_TYPEPOSQUERY:
1394        case OP_TYPEPOSUPTO:
1395    
1396        case OP_CLASS:
1397        case OP_NCLASS:
1398        case OP_XCLASS:
1399    
1400        cc = next_opcode(common, cc);
1401        SLJIT_ASSERT(cc != NULL);
1402        break;
1403        }
1404    
1405    /* Possessive quantifiers can use a special case. */
1406    if (SLJIT_UNLIKELY(possessive == length))
1407      return stack_restore ? no_frame : no_stack;
1408    
1409    if (length > 0)
1410      return length + 1;
1411    return stack_restore ? no_frame : no_stack;
1412    }
1413    
1414    static void init_frame(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, int stackpos, int stacktop, BOOL recursive)
1415    {
1416    DEFINE_COMPILER;
1417    BOOL setsom_found = recursive;
1418    BOOL setmark_found = recursive;
1419    /* The last capture is a local variable even for recursions. */
1420    BOOL capture_last_found = FALSE;
1421    int offset;
1422    
1423    /* >= 1 + shortest item size (2) */
1424    SLJIT_UNUSED_ARG(stacktop);
1425    SLJIT_ASSERT(stackpos >= stacktop + 2);
1426    
1427    stackpos = STACK(stackpos);
1428    if (ccend == NULL)
1429      {
1430      ccend = bracketend(cc) - (1 + LINK_SIZE);
1431      if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))
1432        cc = next_opcode(common, cc);
1433      }
1434    
1435    SLJIT_ASSERT(cc != NULL);
1436    while (cc < ccend)
1437      switch(*cc)
1438        {
1439        case OP_SET_SOM:
1440        SLJIT_ASSERT(common->has_set_som);
1441        if (!setsom_found)
1442          {
1443          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));
1444          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1445          stackpos += (int)sizeof(sljit_sw);
1446          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1447          stackpos += (int)sizeof(sljit_sw);
1448          setsom_found = TRUE;
1449          }
1450        cc += 1;
1451        break;
1452    
1453        case OP_MARK:
1454        case OP_PRUNE_ARG:
1455        case OP_THEN_ARG:
1456        SLJIT_ASSERT(common->mark_ptr != 0);
1457        if (!setmark_found)
1458          {
1459          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);
1460          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
1461          stackpos += (int)sizeof(sljit_sw);
1462          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1463          stackpos += (int)sizeof(sljit_sw);
1464          setmark_found = TRUE;
1465          }
1466        cc += 1 + 2 + cc[1];
1467        break;
1468    
1469        case OP_RECURSE:
1470        if (common->has_set_som && !setsom_found)
1471          {
1472          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));
1473          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1474          stackpos += (int)sizeof(sljit_sw);
1475          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1476          stackpos += (int)sizeof(sljit_sw);
1477          setsom_found = TRUE;
1478          }
1479        if (common->mark_ptr != 0 && !setmark_found)
1480          {
1481          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);
1482          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
1483          stackpos += (int)sizeof(sljit_sw);
1484          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1485          stackpos += (int)sizeof(sljit_sw);
1486          setmark_found = TRUE;
1487          }
1488        if (common->capture_last_ptr != 0 && !capture_last_found)
1489          {
1490          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->capture_last_ptr);
1491          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1492          stackpos += (int)sizeof(sljit_sw);
1493          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1494          stackpos += (int)sizeof(sljit_sw);
1495          capture_last_found = TRUE;
1496          }
1497        cc += 1 + LINK_SIZE;
1498        break;
1499    
1500        case OP_CBRA:
1501        case OP_CBRAPOS:
1502        case OP_SCBRA:
1503        case OP_SCBRAPOS:
1504        if (common->capture_last_ptr != 0 && !capture_last_found)
1505          {
1506          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->capture_last_ptr);
1507          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1508          stackpos += (int)sizeof(sljit_sw);
1509          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1510          stackpos += (int)sizeof(sljit_sw);
1511          capture_last_found = TRUE;
1512          }
1513        offset = (GET2(cc, 1 + LINK_SIZE)) << 1;
1514        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));
1515        stackpos += (int)sizeof(sljit_sw);
1516        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset));
1517        OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset + 1));
1518        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1519        stackpos += (int)sizeof(sljit_sw);
1520        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);
1521        stackpos += (int)sizeof(sljit_sw);
1522    
1523        cc += 1 + LINK_SIZE + IMM2_SIZE;
1524        break;
1525    
1526        default:
1527        cc = next_opcode(common, cc);
1528        SLJIT_ASSERT(cc != NULL);
1529        break;
1530        }
1531    
1532    OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, 0);
1533    SLJIT_ASSERT(stackpos == STACK(stacktop));
1534    }
1535    
1536    static SLJIT_INLINE int get_private_data_copy_length(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, BOOL needs_control_head)
1537    {
1538    int private_data_length = needs_control_head ? 3 : 2;
1539    int size;
1540    pcre_uchar *alternative;
1541    /* Calculate the sum of the private machine words. */
1542    while (cc < ccend)
1543      {
1544      size = 0;
1545      switch(*cc)
1546        {
1547        case OP_KET:
1548        if (PRIVATE_DATA(cc) != 0)
1549          private_data_length++;
1550        cc += 1 + LINK_SIZE;
1551        break;
1552    
1553        case OP_ASSERT:
1554        case OP_ASSERT_NOT:
1555        case OP_ASSERTBACK:
1556        case OP_ASSERTBACK_NOT:
1557        case OP_ONCE:
1558        case OP_ONCE_NC:
1559        case OP_BRAPOS:
1560        case OP_SBRA:
1561        case OP_SBRAPOS:
1562        case OP_SCOND:
1563        private_data_length++;
1564        cc += 1 + LINK_SIZE;
1565        break;
1566    
1567        case OP_CBRA:
1568        case OP_SCBRA:
1569        if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1570          private_data_length++;
1571        cc += 1 + LINK_SIZE + IMM2_SIZE;
1572        break;
1573    
1574        case OP_CBRAPOS:
1575        case OP_SCBRAPOS:
1576        private_data_length += 2;
1577        cc += 1 + LINK_SIZE + IMM2_SIZE;
1578        break;
1579    
1580        case OP_COND:
1581        /* Might be a hidden SCOND. */
1582        alternative = cc + GET(cc, 1);
1583        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1584          private_data_length++;
1585        cc += 1 + LINK_SIZE;
1586        break;
1587    
1588        CASE_ITERATOR_PRIVATE_DATA_1
1589        if (PRIVATE_DATA(cc))
1590          private_data_length++;
1591        cc += 2;
1592    #ifdef SUPPORT_UTF
1593        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1594    #endif
1595        break;
1596    
1597        CASE_ITERATOR_PRIVATE_DATA_2A
1598        if (PRIVATE_DATA(cc))
1599          private_data_length += 2;
1600        cc += 2;
1601    #ifdef SUPPORT_UTF
1602        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1603    #endif
1604        break;
1605    
1606        CASE_ITERATOR_PRIVATE_DATA_2B
1607        if (PRIVATE_DATA(cc))
1608          private_data_length += 2;
1609        cc += 2 + IMM2_SIZE;
1610    #ifdef SUPPORT_UTF
1611        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1612    #endif
1613        break;
1614    
1615        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1616        if (PRIVATE_DATA(cc))
1617          private_data_length++;
1618        cc += 1;
1619        break;
1620    
1621        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1622        if (PRIVATE_DATA(cc))
1623          private_data_length += 2;
1624        cc += 1;
1625        break;
1626    
1627        CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1628        if (PRIVATE_DATA(cc))
1629          private_data_length += 2;
1630        cc += 1 + IMM2_SIZE;
1631        break;
1632    
1633        case OP_CLASS:
1634        case OP_NCLASS:
1635    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1636        case OP_XCLASS:
1637        size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
1638    #else
1639        size = 1 + 32 / (int)sizeof(pcre_uchar);
1640    #endif
1641        if (PRIVATE_DATA(cc))
1642          private_data_length += get_class_iterator_size(cc + size);
1643        cc += size;
1644        break;
1645    
1646        default:
1647        cc = next_opcode(common, cc);
1648        SLJIT_ASSERT(cc != NULL);
1649        break;
1650        }
1651      }
1652    SLJIT_ASSERT(cc == ccend);
1653    return private_data_length;
1654    }
1655    
1656    static void copy_private_data(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend,
1657      BOOL save, int stackptr, int stacktop, BOOL needs_control_head)
1658    {
1659    DEFINE_COMPILER;
1660    int srcw[2];
1661    int count, size;
1662    BOOL tmp1next = TRUE;
1663    BOOL tmp1empty = TRUE;
1664    BOOL tmp2empty = TRUE;
1665    pcre_uchar *alternative;
1666    enum {
1667      start,
1668      loop,
1669      end
1670    } status;
1671    
1672    status = save ? start : loop;
1673    stackptr = STACK(stackptr - 2);
1674    stacktop = STACK(stacktop - 1);
1675    
1676    if (!save)
1677      {
1678      stackptr += (needs_control_head ? 2 : 1) * sizeof(sljit_sw);
1679      if (stackptr < stacktop)
1680        {
1681        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1682        stackptr += sizeof(sljit_sw);
1683        tmp1empty = FALSE;
1684        }
1685      if (stackptr < stacktop)
1686        {
1687        OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1688        stackptr += sizeof(sljit_sw);
1689        tmp2empty = FALSE;
1690        }
1691      /* The tmp1next must be TRUE in either way. */
1692      }
1693    
1694    do
1695      {
1696      count = 0;
1697      switch(status)
1698        {
1699        case start:
1700        SLJIT_ASSERT(save && common->recursive_head_ptr != 0);
1701        count = 1;
1702        srcw[0] = common->recursive_head_ptr;
1703        if (needs_control_head)
1704          {
1705          SLJIT_ASSERT(common->control_head_ptr != 0);
1706          count = 2;
1707          srcw[1] = common->control_head_ptr;
1708          }
1709        status = loop;
1710        break;
1711    
1712        case loop:
1713        if (cc >= ccend)
1714          {
1715          status = end;
1716          break;
1717          }
1718    
1719        switch(*cc)
1720          {
1721          case OP_KET:
1722          if (PRIVATE_DATA(cc) != 0)
1723            {
1724            count = 1;
1725            srcw[0] = PRIVATE_DATA(cc);
1726            }
1727          cc += 1 + LINK_SIZE;
1728          break;
1729    
1730          case OP_ASSERT:
1731          case OP_ASSERT_NOT:
1732          case OP_ASSERTBACK:
1733          case OP_ASSERTBACK_NOT:
1734          case OP_ONCE:
1735          case OP_ONCE_NC:
1736          case OP_BRAPOS:
1737          case OP_SBRA:
1738          case OP_SBRAPOS:
1739          case OP_SCOND:
1740          count = 1;
1741          srcw[0] = PRIVATE_DATA(cc);
1742          SLJIT_ASSERT(srcw[0] != 0);
1743          cc += 1 + LINK_SIZE;
1744          break;
1745    
1746          case OP_CBRA:
1747          case OP_SCBRA:
1748          if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1749            {
1750            count = 1;
1751            srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1752            }
1753          cc += 1 + LINK_SIZE + IMM2_SIZE;
1754          break;
1755    
1756          case OP_CBRAPOS:
1757          case OP_SCBRAPOS:
1758          count = 2;
1759          srcw[0] = PRIVATE_DATA(cc);
1760          srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1761          SLJIT_ASSERT(srcw[0] != 0 && srcw[1] != 0);
1762          cc += 1 + LINK_SIZE + IMM2_SIZE;
1763          break;
1764    
1765          case OP_COND:
1766          /* Might be a hidden SCOND. */
1767          alternative = cc + GET(cc, 1);
1768          if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1769            {
1770            count = 1;
1771            srcw[0] = PRIVATE_DATA(cc);
1772            SLJIT_ASSERT(srcw[0] != 0);
1773            }
1774          cc += 1 + LINK_SIZE;
1775          break;
1776    
1777          CASE_ITERATOR_PRIVATE_DATA_1
1778          if (PRIVATE_DATA(cc))
1779            {
1780            count = 1;
1781            srcw[0] = PRIVATE_DATA(cc);
1782            }
1783          cc += 2;
1784    #ifdef SUPPORT_UTF
1785          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1786    #endif
1787          break;
1788    
1789          CASE_ITERATOR_PRIVATE_DATA_2A
1790          if (PRIVATE_DATA(cc))
1791            {
1792            count = 2;
1793            srcw[0] = PRIVATE_DATA(cc);
1794            srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1795            }
1796          cc += 2;
1797    #ifdef SUPPORT_UTF
1798          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1799    #endif
1800          break;
1801    
1802          CASE_ITERATOR_PRIVATE_DATA_2B
1803          if (PRIVATE_DATA(cc))
1804            {
1805            count = 2;
1806            srcw[0] = PRIVATE_DATA(cc);
1807            srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1808            }
1809          cc += 2 + IMM2_SIZE;
1810    #ifdef SUPPORT_UTF
1811          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1812    #endif
1813          break;
1814    
1815          CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1816          if (PRIVATE_DATA(cc))
1817            {
1818            count = 1;
1819            srcw[0] = PRIVATE_DATA(cc);
1820            }
1821          cc += 1;
1822          break;
1823    
1824          CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1825          if (PRIVATE_DATA(cc))
1826            {
1827            count = 2;
1828            srcw[0] = PRIVATE_DATA(cc);
1829            srcw[1] = srcw[0] + sizeof(sljit_sw);
1830            }
1831          cc += 1;
1832          break;
1833    
1834          CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1835          if (PRIVATE_DATA(cc))
1836            {
1837            count = 2;
1838            srcw[0] = PRIVATE_DATA(cc);
1839            srcw[1] = srcw[0] + sizeof(sljit_sw);
1840            }
1841          cc += 1 + IMM2_SIZE;
1842          break;
1843    
1844          case OP_CLASS:
1845          case OP_NCLASS:
1846    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1847          case OP_XCLASS:
1848          size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
1849    #else
1850          size = 1 + 32 / (int)sizeof(pcre_uchar);
1851    #endif
1852          if (PRIVATE_DATA(cc))
1853            switch(get_class_iterator_size(cc + size))
1854              {
1855              case 1:
1856              count = 1;
1857              srcw[0] = PRIVATE_DATA(cc);
1858              break;
1859    
1860              case 2:
1861              count = 2;
1862              srcw[0] = PRIVATE_DATA(cc);
1863              srcw[1] = srcw[0] + sizeof(sljit_sw);
1864              break;
1865    
1866              default:
1867              SLJIT_ASSERT_STOP();
1868              break;
1869              }
1870          cc += size;
1871          break;
1872    
1873          default:
1874          cc = next_opcode(common, cc);
1875          SLJIT_ASSERT(cc != NULL);
1876          break;
1877          }
1878        break;
1879    
1880        case end:
1881        SLJIT_ASSERT_STOP();
1882      break;      break;
1883      }      }
1884    
# Line 1015  while (status != end) Line 1892  while (status != end)
1892          if (!tmp1empty)          if (!tmp1empty)
1893            {            {
1894            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1895            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1896            }            }
1897          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);
1898          tmp1empty = FALSE;          tmp1empty = FALSE;
# Line 1026  while (status != end) Line 1903  while (status != end)
1903          if (!tmp2empty)          if (!tmp2empty)
1904            {            {
1905            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1906            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1907            }            }
1908          OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);          OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);
1909          tmp2empty = FALSE;          tmp2empty = FALSE;
# Line 1043  while (status != end) Line 1920  while (status != end)
1920          if (!tmp1empty)          if (!tmp1empty)
1921            {            {
1922            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1923            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1924            }            }
1925          tmp1next = FALSE;          tmp1next = FALSE;
1926          }          }
# Line 1055  while (status != end) Line 1932  while (status != end)
1932          if (!tmp2empty)          if (!tmp2empty)
1933            {            {
1934            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1935            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1936            }            }
1937          tmp1next = TRUE;          tmp1next = TRUE;
1938          }          }
1939        }        }
1940      }      }
1941    }    }
1942    while (status != end);
1943    
1944  if (save)  if (save)
1945    {    {
# Line 1070  if (save) Line 1948  if (save)
1948      if (!tmp1empty)      if (!tmp1empty)
1949        {        {
1950        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1951        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1952        }        }
1953      if (!tmp2empty)      if (!tmp2empty)
1954        {        {
1955        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1956        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1957        }        }
1958      }      }
1959    else    else
# Line 1083  if (save) Line 1961  if (save)
1961      if (!tmp2empty)      if (!tmp2empty)
1962        {        {
1963        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1964        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1965        }        }
1966      if (!tmp1empty)      if (!tmp1empty)
1967        {        {
1968        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1969        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1970        }        }
1971      }      }
1972    }    }
1973  SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));  SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));
1974  }  }
1975    
1976  static SLJIT_INLINE BOOL ispowerof2(unsigned int value)  static SLJIT_INLINE pcre_uchar *set_then_offsets(compiler_common *common, pcre_uchar *cc, pcre_uint8 *current_offset)
1977    {
1978    pcre_uchar *end = bracketend(cc);
1979    BOOL has_alternatives = cc[GET(cc, 1)] == OP_ALT;
1980    
1981    /* Assert captures then. */
1982    if (*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT)
1983      current_offset = NULL;
1984    /* Conditional block does not. */
1985    if (*cc == OP_COND || *cc == OP_SCOND)
1986      has_alternatives = FALSE;
1987    
1988    cc = next_opcode(common, cc);
1989    if (has_alternatives)
1990      current_offset = common->then_offsets + (cc - common->start);
1991    
1992    while (cc < end)
1993      {
1994      if ((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND))
1995        cc = set_then_offsets(common, cc, current_offset);
1996      else
1997        {
1998        if (*cc == OP_ALT && has_alternatives)
1999          current_offset = common->then_offsets + (cc + 1 + LINK_SIZE - common->start);
2000        if (*cc >= OP_THEN && *cc <= OP_THEN_ARG && current_offset != NULL)
2001          *current_offset = 1;
2002        cc = next_opcode(common, cc);
2003        }
2004      }
2005    
2006    return end;
2007    }
2008    
2009    #undef CASE_ITERATOR_PRIVATE_DATA_1
2010    #undef CASE_ITERATOR_PRIVATE_DATA_2A
2011    #undef CASE_ITERATOR_PRIVATE_DATA_2B
2012    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_1
2013    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
2014    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
2015    
2016    static SLJIT_INLINE BOOL is_powerof2(unsigned int value)
2017  {  {
2018  return (value & (value - 1)) == 0;  return (value & (value - 1)) == 0;
2019  }  }
# Line 1105  static SLJIT_INLINE void set_jumps(jump_ Line 2023  static SLJIT_INLINE void set_jumps(jump_
2023  while (list)  while (list)
2024    {    {
2025    /* sljit_set_label is clever enough to do nothing    /* sljit_set_label is clever enough to do nothing
2026    if either the jump or the label is NULL */    if either the jump or the label is NULL. */
2027    sljit_set_label(list->jump, label);    SET_LABEL(list->jump, label);
2028    list = list->next;    list = list->next;
2029    }    }
2030  }  }
# Line 1122  if (list_item) Line 2040  if (list_item)
2040    }    }
2041  }  }
2042    
2043  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)
2044  {  {
2045  DEFINE_COMPILER;  DEFINE_COMPILER;
2046  stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));  stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));
2047    
2048  if (list_item)  if (list_item)
2049    {    {
   list_item->type = type;  
   list_item->data = data;  
2050    list_item->start = start;    list_item->start = start;
2051    list_item->leave = LABEL();    list_item->quit = LABEL();
2052    list_item->next = common->stubs;    list_item->next = common->stubs;
2053    common->stubs = list_item;    common->stubs = list_item;
2054    }    }
# Line 1146  stub_list* list_item = common->stubs; Line 2062  stub_list* list_item = common->stubs;
2062  while (list_item)  while (list_item)
2063    {    {
2064    JUMPHERE(list_item->start);    JUMPHERE(list_item->start);
2065    switch(list_item->type)    add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));
2066      {    JUMPTO(SLJIT_JUMP, list_item->quit);
     case stack_alloc:  
     add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));  
     break;  
     }  
   JUMPTO(SLJIT_JUMP, list_item->leave);  
2067    list_item = list_item->next;    list_item = list_item->next;
2068    }    }
2069  common->stubs = NULL;  common->stubs = NULL;
2070  }  }
2071    
2072  static SLJIT_INLINE void decrease_call_count(compiler_common *common)  static void add_label_addr(compiler_common *common)
2073    {
2074    DEFINE_COMPILER;
2075    label_addr_list *label_addr;
2076    
2077    label_addr = sljit_alloc_memory(compiler, sizeof(label_addr_list));
2078    if (label_addr == NULL)
2079      return;
2080    label_addr->label = LABEL();
2081    label_addr->addr = common->read_only_data_ptr;
2082    label_addr->next = common->label_addrs;
2083    common->label_addrs = label_addr;
2084    common->read_only_data_ptr++;
2085    }
2086    
2087    static SLJIT_INLINE void count_match(compiler_common *common)
2088  {  {
2089  DEFINE_COMPILER;  DEFINE_COMPILER;
2090    
2091  OP2(SLJIT_SUB | SLJIT_SET_E, CALL_COUNT, 0, CALL_COUNT, 0, SLJIT_IMM, 1);  OP2(SLJIT_SUB | SLJIT_SET_E, COUNT_MATCH, 0, COUNT_MATCH, 0, SLJIT_IMM, 1);
2092  add_jump(compiler, &common->calllimit, JUMP(SLJIT_C_ZERO));  add_jump(compiler, &common->calllimit, JUMP(SLJIT_C_ZERO));
2093  }  }
2094    
# Line 1171  static SLJIT_INLINE void allocate_stack( Line 2097  static SLJIT_INLINE void allocate_stack(
2097  /* May destroy all locals and registers except TMP2. */  /* May destroy all locals and registers except TMP2. */
2098  DEFINE_COMPILER;  DEFINE_COMPILER;
2099    
2100  OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_w));  OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_sw));
2101  #ifdef DESTROY_REGISTERS  #ifdef DESTROY_REGISTERS
2102  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);
2103  OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);  OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);
# Line 1179  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0); Line 2105  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);
2105  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);
2106  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);
2107  #endif  #endif
2108  add_stub(common, stack_alloc, 0, CMP(SLJIT_C_GREATER, STACK_TOP, 0, STACK_LIMIT, 0));  add_stub(common, CMP(SLJIT_C_GREATER, STACK_TOP, 0, STACK_LIMIT, 0));
2109  }  }
2110    
2111  static SLJIT_INLINE void free_stack(compiler_common *common, int size)  static SLJIT_INLINE void free_stack(compiler_common *common, int size)
2112  {  {
2113  DEFINE_COMPILER;  DEFINE_COMPILER;
2114  OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_w));  OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_sw));
2115  }  }
2116    
2117  static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)  static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)
# Line 1193  static SLJIT_INLINE void reset_ovector(c Line 2119  static SLJIT_INLINE void reset_ovector(c
2119  DEFINE_COMPILER;  DEFINE_COMPILER;
2120  struct sljit_label *loop;  struct sljit_label *loop;
2121  int i;  int i;
2122    
2123  /* At this point we can freely use all temporary registers. */  /* At this point we can freely use all temporary registers. */
2124    SLJIT_ASSERT(length > 1);
2125  /* TMP1 returns with begin - 1. */  /* TMP1 returns with begin - 1. */
2126  OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG1, 0, SLJIT_MEM1(SLJIT_GENERAL_REG1), SLJIT_OFFSETOF(jit_arguments, begin), SLJIT_IMM, 1);  OP2(SLJIT_SUB, SLJIT_SCRATCH_REG1, 0, SLJIT_MEM1(SLJIT_SAVED_REG1), SLJIT_OFFSETOF(jit_arguments, begin), SLJIT_IMM, IN_UCHARS(1));
2127    if (length < 8)
2128      {
2129      for (i = 1; i < length; i++)
2130        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), SLJIT_SCRATCH_REG1, 0);
2131      }
2132    else
2133      {
2134      GET_LOCAL_BASE(SLJIT_SCRATCH_REG2, 0, OVECTOR_START);
2135      OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, length - 1);
2136      loop = LABEL();
2137      OP1(SLJIT_MOVU, SLJIT_MEM1(SLJIT_SCRATCH_REG2), sizeof(sljit_sw), SLJIT_SCRATCH_REG1, 0);
2138      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 1);
2139      JUMPTO(SLJIT_C_NOT_ZERO, loop);
2140      }
2141    }
2142    
2143    static SLJIT_INLINE void do_reset_match(compiler_common *common, int length)
2144    {
2145    DEFINE_COMPILER;
2146    struct sljit_label *loop;
2147    int i;
2148    
2149    SLJIT_ASSERT(length > 1);
2150    /* OVECTOR(1) contains the "string begin - 1" constant. */
2151    if (length > 2)
2152      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));
2153  if (length < 8)  if (length < 8)
2154    {    {
2155    for (i = 0; i < length; i++)    for (i = 2; i < length; i++)
2156      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), SLJIT_TEMPORARY_REG1, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), TMP1, 0);
2157    }    }
2158  else  else
2159    {    {
2160    OP2(SLJIT_ADD, SLJIT_TEMPORARY_REG2, 0, SLJIT_LOCALS_REG, 0, SLJIT_IMM, OVECTOR_START - sizeof(sljit_w));    GET_LOCAL_BASE(TMP2, 0, OVECTOR_START + sizeof(sljit_sw));
2161    OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, length);    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_IMM, length - 2);
2162    loop = LABEL();    loop = LABEL();
2163    OP1(SLJIT_MOVU, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), sizeof(sljit_w), SLJIT_TEMPORARY_REG1, 0);    OP1(SLJIT_MOVU, SLJIT_MEM1(TMP2), sizeof(sljit_sw), TMP1, 0);
2164    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_TEMPORARY_REG3, 0, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, 1);    OP2(SLJIT_SUB | SLJIT_SET_E, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, 1);
2165    JUMPTO(SLJIT_C_NOT_ZERO, loop);    JUMPTO(SLJIT_C_NOT_ZERO, loop);
2166    }    }
2167    
2168    OP1(SLJIT_MOV, STACK_TOP, 0, ARGUMENTS, 0);
2169    if (common->mark_ptr != 0)
2170      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr, SLJIT_IMM, 0);
2171    if (common->control_head_ptr != 0)
2172      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->control_head_ptr, SLJIT_IMM, 0);
2173    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(jit_arguments, stack));
2174    OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_ptr);
2175    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(struct sljit_stack, base));
2176    }
2177    
2178    static sljit_sw SLJIT_CALL do_search_mark(sljit_sw *current, const pcre_uchar *skip_arg)
2179    {
2180    while (current != NULL)
2181      {
2182      switch (current[-2])
2183        {
2184        case type_then_trap:
2185        break;
2186    
2187        case type_mark:
2188        if (STRCMP_UC_UC(skip_arg, (pcre_uchar *)current[-3]) == 0)
2189          return current[-4];
2190        break;
2191    
2192        default:
2193        SLJIT_ASSERT_STOP();
2194        break;
2195        }
2196      current = (sljit_sw*)current[-1];
2197      }
2198    return -1;
2199  }  }
2200    
2201  static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)  static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)
2202  {  {
2203  DEFINE_COMPILER;  DEFINE_COMPILER;
2204  struct sljit_label *loop;  struct sljit_label *loop;
2205  struct sljit_jump *earlyexit;  struct sljit_jump *early_quit;
2206    
2207  /* At this point we can freely use all registers. */  /* At this point we can freely use all registers. */
2208  OP1(SLJIT_MOV, SLJIT_GENERAL_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));  OP1(SLJIT_MOV, SLJIT_SAVED_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));
2209  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1), STR_PTR, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1), STR_PTR, 0);
2210    
2211  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, ARGUMENTS, 0);  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG1, 0, ARGUMENTS, 0);
2212  OP1(SLJIT_MOV_SI, SLJIT_TEMPORARY_REG2, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, offsetcount));  if (common->mark_ptr != 0)
2213  OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);
2214  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, begin));  OP1(SLJIT_MOV_SI, SLJIT_SCRATCH_REG2, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, offset_count));
2215  OP2(SLJIT_ADD, SLJIT_GENERAL_REG1, 0, SLJIT_LOCALS_REG, 0, SLJIT_IMM, OVECTOR_START);  if (common->mark_ptr != 0)
2216      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, mark_ptr), SLJIT_SCRATCH_REG3, 0);
2217    OP2(SLJIT_SUB, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));
2218    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG1, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, begin));
2219    GET_LOCAL_BASE(SLJIT_SAVED_REG1, 0, OVECTOR_START);
2220  /* Unlikely, but possible */  /* Unlikely, but possible */
2221  earlyexit = CMP(SLJIT_C_EQUAL, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, 0);  early_quit = CMP(SLJIT_C_EQUAL, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, 0);
2222  loop = LABEL();  loop = LABEL();
2223  OP2(SLJIT_SUB, SLJIT_GENERAL_REG2, 0, SLJIT_MEM1(SLJIT_GENERAL_REG1), 0, SLJIT_TEMPORARY_REG1, 0);  OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, SLJIT_MEM1(SLJIT_SAVED_REG1), 0, SLJIT_SCRATCH_REG1, 0);
2224  OP2(SLJIT_ADD, SLJIT_GENERAL_REG1, 0, SLJIT_GENERAL_REG1, 0, SLJIT_IMM, sizeof(sljit_w));  OP2(SLJIT_ADD, SLJIT_SAVED_REG1, 0, SLJIT_SAVED_REG1, 0, SLJIT_IMM, sizeof(sljit_sw));
2225  /* Copy the integer value to the output buffer */  /* Copy the integer value to the output buffer */
2226  OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_TEMPORARY_REG3), sizeof(int), SLJIT_GENERAL_REG2, 0);  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2227  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_TEMPORARY_REG2, 0, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, 1);  OP2(SLJIT_ASHR, SLJIT_SAVED_REG2, 0, SLJIT_SAVED_REG2, 0, SLJIT_IMM, UCHAR_SHIFT);
2228    #endif
2229    OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG3), sizeof(int), SLJIT_SAVED_REG2, 0);
2230    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_SCRATCH_REG2, 0, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, 1);
2231  JUMPTO(SLJIT_C_NOT_ZERO, loop);  JUMPTO(SLJIT_C_NOT_ZERO, loop);
2232  JUMPHERE(earlyexit);  JUMPHERE(early_quit);
2233    
2234  /* Calculate the return value, which is the maximum ovector value. */  /* Calculate the return value, which is the maximum ovector value. */
2235  if (topbracket > 1)  if (topbracket > 1)
2236    {    {
2237    OP2(SLJIT_ADD, SLJIT_TEMPORARY_REG1, 0, SLJIT_LOCALS_REG, 0, SLJIT_IMM, OVECTOR_START + topbracket * 2 * sizeof(sljit_w));    GET_LOCAL_BASE(SLJIT_SCRATCH_REG1, 0, OVECTOR_START + topbracket * 2 * sizeof(sljit_sw));
2238    OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, topbracket + 1);    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, topbracket + 1);
2239    
2240    /* OVECTOR(0) is never equal to SLJIT_GENERAL_REG3. */    /* OVECTOR(0) is never equal to SLJIT_SAVED_REG3. */
2241    loop = LABEL();    loop = LABEL();
2242    OP1(SLJIT_MOVU, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), -(2 * (sljit_w)sizeof(sljit_w)));    OP1(SLJIT_MOVU, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), -(2 * (sljit_sw)sizeof(sljit_sw)));
2243    OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG2, 0, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, 1);    OP2(SLJIT_SUB, SLJIT_SCRATCH_REG2, 0, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, 1);
2244    CMPTO(SLJIT_C_EQUAL, SLJIT_TEMPORARY_REG3, 0, SLJIT_GENERAL_REG3, 0, loop);    CMPTO(SLJIT_C_EQUAL, SLJIT_SCRATCH_REG3, 0, SLJIT_SAVED_REG3, 0, loop);
2245    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_TEMPORARY_REG2, 0);    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_SCRATCH_REG2, 0);
2246    }    }
2247  else  else
2248    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, 1);    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, 1);
2249  }  }
2250    
2251  static SLJIT_INLINE BOOL char_has_othercase(compiler_common *common, uschar* cc)  static SLJIT_INLINE void return_with_partial_match(compiler_common *common, struct sljit_label *quit)
2252    {
2253    DEFINE_COMPILER;
2254    struct sljit_jump *jump;
2255    
2256    SLJIT_COMPILE_ASSERT(STR_END == SLJIT_SAVED_REG2, str_end_must_be_saved_reg2);
2257    SLJIT_ASSERT(common->start_used_ptr != 0 && common->start_ptr != 0
2258      && (common->mode == JIT_PARTIAL_SOFT_COMPILE ? common->hit_start != 0 : common->hit_start == 0));
2259    
2260    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, ARGUMENTS, 0);
2261    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, PCRE_ERROR_PARTIAL);
2262    OP1(SLJIT_MOV_SI, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, real_offset_count));
2263    CMPTO(SLJIT_C_SIG_LESS, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 2, quit);
2264    
2265    /* Store match begin and end. */
2266    OP1(SLJIT_MOV, SLJIT_SAVED_REG1, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, begin));
2267    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, offsets));
2268    
2269    jump = CMP(SLJIT_C_SIG_LESS, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 3);
2270    OP2(SLJIT_SUB, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mode == JIT_PARTIAL_HARD_COMPILE ? common->start_ptr : (common->hit_start + (int)sizeof(sljit_sw)), SLJIT_SAVED_REG1, 0);
2271    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2272    OP2(SLJIT_ASHR, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, UCHAR_SHIFT);
2273    #endif
2274    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG2), 2 * sizeof(int), SLJIT_SCRATCH_REG3, 0);
2275    JUMPHERE(jump);
2276    
2277    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mode == JIT_PARTIAL_HARD_COMPILE ? common->start_used_ptr : common->hit_start);
2278    OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, STR_END, 0, SLJIT_SAVED_REG1, 0);
2279    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2280    OP2(SLJIT_ASHR, SLJIT_SAVED_REG2, 0, SLJIT_SAVED_REG2, 0, SLJIT_IMM, UCHAR_SHIFT);
2281    #endif
2282    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG2), sizeof(int), SLJIT_SAVED_REG2, 0);
2283    
2284    OP2(SLJIT_SUB, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_SAVED_REG1, 0);
2285    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2286    OP2(SLJIT_ASHR, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, UCHAR_SHIFT);
2287    #endif
2288    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG2), 0, SLJIT_SCRATCH_REG3, 0);
2289    
2290    JUMPTO(SLJIT_JUMP, quit);
2291    }
2292    
2293    static SLJIT_INLINE void check_start_used_ptr(compiler_common *common)
2294    {
2295    /* May destroy TMP1. */
2296    DEFINE_COMPILER;
2297    struct sljit_jump *jump;
2298    
2299    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2300      {
2301      /* The value of -1 must be kept for start_used_ptr! */
2302      OP2(SLJIT_ADD, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, SLJIT_IMM, 1);
2303      /* Jumps if start_used_ptr < STR_PTR, or start_used_ptr == -1. Although overwriting
2304      is not necessary if start_used_ptr == STR_PTR, it does not hurt as well. */
2305      jump = CMP(SLJIT_C_LESS_EQUAL, TMP1, 0, STR_PTR, 0);
2306      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);
2307      JUMPHERE(jump);
2308      }
2309    else if (common->mode == JIT_PARTIAL_HARD_COMPILE)
2310      {
2311      jump = CMP(SLJIT_C_LESS_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);
2312      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);
2313      JUMPHERE(jump);
2314      }
2315    }
2316    
2317    static SLJIT_INLINE BOOL char_has_othercase(compiler_common *common, pcre_uchar* cc)
2318  {  {
2319  /* Detects if the character has an othercase. */  /* Detects if the character has an othercase. */
2320  unsigned int c;  unsigned int c;
2321    
2322  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2323  if (common->utf8)  if (common->utf)
2324    {    {
2325    GETCHAR(c, cc);    GETCHAR(c, cc);
2326    if (c > 127)    if (c > 127)
# Line 1272  if (common->utf8) Line 2331  if (common->utf8)
2331      return FALSE;      return FALSE;
2332  #endif  #endif
2333      }      }
2334    #ifndef COMPILE_PCRE8
2335      return common->fcc[c] != c;
2336    #endif
2337    }    }
2338  else  else
2339  #endif  #endif
2340    c = *cc;    c = *cc;
2341  return common->fcc[c] != c;  return MAX_255(c) ? common->fcc[c] != c : FALSE;
2342  }  }
2343    
2344  static SLJIT_INLINE unsigned int char_othercase(compiler_common *common, unsigned int c)  static SLJIT_INLINE unsigned int char_othercase(compiler_common *common, unsigned int c)
2345  {  {
2346  /* Returns with the othercase. */  /* Returns with the othercase. */
2347  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2348  if (common->utf8 && c > 127)  if (common->utf && c > 127)
2349    {    {
2350  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2351    return UCD_OTHERCASE(c);    return UCD_OTHERCASE(c);
# Line 1292  if (common->utf8 && c > 127) Line 2354  if (common->utf8 && c > 127)
2354  #endif  #endif
2355    }    }
2356  #endif  #endif
2357  return common->fcc[c];  return TABLE_GET(c, common->fcc, c);
2358  }  }
2359    
2360  static unsigned int char_get_othercase_bit(compiler_common *common, uschar* cc)  static unsigned int char_get_othercase_bit(compiler_common *common, pcre_uchar* cc)
2361  {  {
2362  /* Detects if the character and its othercase has only 1 bit difference. */  /* Detects if the character and its othercase has only 1 bit difference. */
2363  unsigned int c, oc, bit;  unsigned int c, oc, bit;
2364  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2365  int n;  int n;
2366  #endif  #endif
2367    
2368  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2369  if (common->utf8)  if (common->utf)
2370    {    {
2371    GETCHAR(c, cc);    GETCHAR(c, cc);
2372    if (c <= 127)    if (c <= 127)
# Line 1321  if (common->utf8) Line 2383  if (common->utf8)
2383  else  else
2384    {    {
2385    c = *cc;    c = *cc;
2386    oc = common->fcc[c];    oc = TABLE_GET(c, common->fcc, c);
2387    }    }
2388  #else  #else
2389  c = *cc;  c = *cc;
2390  oc = common->fcc[c];  oc = TABLE_GET(c, common->fcc, c);
2391  #endif  #endif
2392    
2393  SLJIT_ASSERT(c != oc);  SLJIT_ASSERT(c != oc);
# Line 1336  if (c <= 127 && bit == 0x20) Line 2398  if (c <= 127 && bit == 0x20)
2398    return (0 << 8) | 0x20;    return (0 << 8) | 0x20;
2399    
2400  /* Since c != oc, they must have at least 1 bit difference. */  /* Since c != oc, they must have at least 1 bit difference. */
2401  if (!ispowerof2(bit))  if (!is_powerof2(bit))
2402    return 0;    return 0;
2403    
2404  #ifdef SUPPORT_UTF8  #if defined COMPILE_PCRE8
2405  if (common->utf8 && c > 127)  
2406    #ifdef SUPPORT_UTF
2407    if (common->utf && c > 127)
2408    {    {
2409    n = _pcre_utf8_table4[*cc & 0x3f];    n = GET_EXTRALEN(*cc);
2410    while ((bit & 0x3f) == 0)    while ((bit & 0x3f) == 0)
2411      {      {
2412      n--;      n--;
# Line 1350  if (common->utf8 && c > 127) Line 2414  if (common->utf8 && c > 127)
2414      }      }
2415    return (n << 8) | bit;    return (n << 8) | bit;
2416    }    }
2417  #endif  #endif /* SUPPORT_UTF */
2418  return (0 << 8) | bit;  return (0 << 8) | bit;
2419    
2420    #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2421    
2422    #ifdef SUPPORT_UTF
2423    if (common->utf && c > 65535)
2424      {
2425      if (bit >= (1 << 10))
2426        bit >>= 10;
2427      else
2428        return (bit < 256) ? ((2 << 8) | bit) : ((3 << 8) | (bit >> 8));
2429      }
2430    #endif /* SUPPORT_UTF */
2431    return (bit < 256) ? ((0 << 8) | bit) : ((1 << 8) | (bit >> 8));
2432    
2433    #endif /* COMPILE_PCRE[8|16|32] */
2434    }
2435    
2436    static void check_partial(compiler_common *common, BOOL force)
2437    {
2438    /* Checks whether a partial matching is occurred. Does not modify registers. */
2439    DEFINE_COMPILER;
2440    struct sljit_jump *jump = NULL;
2441    
2442    SLJIT_ASSERT(!force || common->mode != JIT_COMPILE);
2443    
2444    if (common->mode == JIT_COMPILE)
2445      return;
2446    
2447    if (!force)
2448      jump = CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);
2449    else if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2450      jump = CMP(SLJIT_C_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, SLJIT_IMM, -1);
2451    
2452    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2453      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2454    else
2455      {
2456      if (common->partialmatchlabel != NULL)
2457        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2458      else
2459        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2460      }
2461    
2462    if (jump != NULL)
2463      JUMPHERE(jump);
2464    }
2465    
2466    static void check_str_end(compiler_common *common, jump_list **end_reached)
2467    {
2468    /* Does not affect registers. Usually used in a tight spot. */
2469    DEFINE_COMPILER;
2470    struct sljit_jump *jump;
2471    
2472    if (common->mode == JIT_COMPILE)
2473      {
2474      add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2475      return;
2476      }
2477    
2478    jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);
2479    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2480      {
2481      add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));
2482      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2483      add_jump(compiler, end_reached, JUMP(SLJIT_JUMP));
2484      }
2485    else
2486      {
2487      add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));
2488      if (common->partialmatchlabel != NULL)
2489        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2490      else
2491        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2492      }
2493    JUMPHERE(jump);
2494  }  }
2495    
2496  static SLJIT_INLINE void check_input_end(compiler_common *common, jump_list **fallbacks)  static void detect_partial_match(compiler_common *common, jump_list **backtracks)
2497  {  {
2498  DEFINE_COMPILER;  DEFINE_COMPILER;
2499  add_jump(compiler, fallbacks, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));  struct sljit_jump *jump;
2500    
2501    if (common->mode == JIT_COMPILE)
2502      {
2503      add_jump(compiler, backtracks, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2504      return;
2505      }
2506    
2507    /* Partial matching mode. */
2508    jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);
2509    add_jump(compiler, backtracks, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));
2510    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2511      {
2512      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2513      add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));
2514      }
2515    else
2516      {
2517      if (common->partialmatchlabel != NULL)
2518        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2519      else
2520        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2521      }
2522    JUMPHERE(jump);
2523  }  }
2524    
2525  static void read_char(compiler_common *common)  static void peek_char(compiler_common *common, pcre_uint32 max)
2526  {  {
2527  /* Reads the character into TMP1, updates STR_PTR.  /* Reads the character into TMP1, keeps STR_PTR.
2528  Does not check STR_END. TMP2 Destroyed. */  Does not check STR_END. TMP2 Destroyed. */
2529  DEFINE_COMPILER;  DEFINE_COMPILER;
2530  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2531  struct sljit_jump *jump;  struct sljit_jump *jump;
2532  #endif  #endif
2533    
2534  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  SLJIT_UNUSED_ARG(max);
2535  #ifdef SUPPORT_UTF8  
2536  if (common->utf8)  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2537    {  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2538    /* Should not found a value between 128 and 192 here. */  if (common->utf)
2539    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 192);    {
2540    add_jump(compiler, &common->utf8readchar, JUMP(SLJIT_FAST_CALL));    if (max < 128) return;
2541    
2542      jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2543      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2544      add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2545      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2546      JUMPHERE(jump);
2547      }
2548    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2549    
2550    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2551    if (common->utf)
2552      {
2553      if (max < 0xd800) return;
2554    
2555      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2556      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2557      /* TMP2 contains the high surrogate. */
2558      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2559      OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2560      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2561      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2562      OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2563    JUMPHERE(jump);    JUMPHERE(jump);
2564    }    }
2565  #endif  #endif
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);  
2566  }  }
2567    
2568  static void peek_char(compiler_common *common)  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2569    
2570    static BOOL is_char7_bitset(const pcre_uint8* bitset, BOOL nclass)
2571  {  {
2572  /* Reads the character into TMP1, keeps STR_PTR.  /* Tells whether the character codes below 128 are enough
2573  Does not check STR_END. TMP2 Destroyed. */  to determine a match. */
2574    const pcre_uint8 value = nclass ? 0xff : 0;
2575    const pcre_uint8* end = bitset + 32;
2576    
2577    bitset += 16;
2578    do
2579      {
2580      if (*bitset++ != value)
2581        return FALSE;
2582      }
2583    while (bitset < end);
2584    return TRUE;
2585    }
2586    
2587    static void read_char7_type(compiler_common *common, BOOL full_read)
2588    {
2589    /* Reads the precise character type of a character into TMP1, if the character
2590    is less than 128. Otherwise it returns with zero. Does not check STR_END. The
2591    full_read argument tells whether characters above max are accepted or not. */
2592    DEFINE_COMPILER;
2593    struct sljit_jump *jump;
2594    
2595    SLJIT_ASSERT(common->utf);
2596    
2597    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2598    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2599    
2600    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2601    
2602    if (full_read)
2603      {
2604      jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2605      OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2606      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2607      JUMPHERE(jump);
2608      }
2609    }
2610    
2611    #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2612    
2613    static void read_char_range(compiler_common *common, pcre_uint32 min, pcre_uint32 max, BOOL update_str_ptr)
2614    {
2615    /* Reads the precise value of a character into TMP1, if the character is
2616    between min and max (c >= min && c <= max). Otherwise it returns with a value
2617    outside the range. Does not check STR_END. */
2618  DEFINE_COMPILER;  DEFINE_COMPILER;
2619  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2620  struct sljit_jump *jump;  struct sljit_jump *jump;
2621  #endif  #endif
2622    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2623    struct sljit_jump *jump2;
2624    #endif
2625    
2626  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  SLJIT_UNUSED_ARG(update_str_ptr);
2627  #ifdef SUPPORT_UTF8  SLJIT_UNUSED_ARG(min);
2628  if (common->utf8)  SLJIT_UNUSED_ARG(max);
2629    {  SLJIT_ASSERT(min <= max);
2630    /* Should not found a value between 128 and 192 here. */  
2631    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 192);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2632    add_jump(compiler, &common->utf8readchar, JUMP(SLJIT_FAST_CALL));  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2633    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);  
2634    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2635    if (common->utf)
2636      {
2637      if (max < 128 && !update_str_ptr) return;
2638    
2639      jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2640      if (min >= 0x10000)
2641        {
2642        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xf0);
2643        if (update_str_ptr)
2644          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2645        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2646        jump2 = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0x7);
2647        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2648        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2649        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2650        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2651        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2652        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2653        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2654        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2655        if (!update_str_ptr)
2656          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2657        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2658        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2659        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2660        JUMPHERE(jump2);
2661        if (update_str_ptr)
2662          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2663        }
2664      else if (min >= 0x800 && max <= 0xffff)
2665        {
2666        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xe0);
2667        if (update_str_ptr)
2668          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2669        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2670        jump2 = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xf);
2671        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2672        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2673        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2674        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2675        if (!update_str_ptr)
2676          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2677        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2678        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2679        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2680        JUMPHERE(jump2);
2681        if (update_str_ptr)
2682          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2683        }
2684      else if (max >= 0x800)
2685        add_jump(compiler, (max < 0x10000) ? &common->utfreadchar16 : &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2686      else if (max < 128)
2687        {
2688        OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2689        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2690        }
2691      else
2692        {
2693        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2694        if (!update_str_ptr)
2695          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2696        else
2697          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2698        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2699        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2700        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2701        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2702        if (update_str_ptr)
2703          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2704        }
2705    JUMPHERE(jump);    JUMPHERE(jump);
2706    }    }
2707  #endif  #endif
2708    
2709    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2710    if (common->utf)
2711      {
2712      if (max >= 0x10000)
2713        {
2714        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2715        jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2716        /* TMP2 contains the high surrogate. */
2717        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2718        OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2719        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2720        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2721        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2722        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2723        JUMPHERE(jump);
2724        return;
2725        }
2726    
2727      if (max < 0xd800 && !update_str_ptr) return;
2728    
2729      /* Skip low surrogate if necessary. */
2730      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2731      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2732      if (update_str_ptr)
2733        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2734      if (max >= 0xd800)
2735        OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0x10000);
2736      JUMPHERE(jump);
2737      }
2738    #endif
2739    }
2740    
2741    static SLJIT_INLINE void read_char(compiler_common *common)
2742    {
2743    read_char_range(common, 0, READ_CHAR_MAX, TRUE);
2744  }  }
2745    
2746  static void read_char8_type(compiler_common *common)  static void read_char8_type(compiler_common *common, BOOL update_str_ptr)
2747  {  {
2748  /* Reads the character type into TMP1, updates STR_PTR. Does not check STR_END. */  /* Reads the character type into TMP1, updates STR_PTR. Does not check STR_END. */
2749  DEFINE_COMPILER;  DEFINE_COMPILER;
2750  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2751  struct sljit_jump *jump;  struct sljit_jump *jump;
2752  #endif  #endif
2753    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2754    struct sljit_jump *jump2;
2755    #endif
2756    
2757    SLJIT_UNUSED_ARG(update_str_ptr);
2758    
2759  #ifdef SUPPORT_UTF8  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2760  if (common->utf8)  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2761    
2762    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2763    if (common->utf)
2764    {    {
   OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);  
   OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);  
2765    /* This can be an extra read in some situations, but hopefully    /* This can be an extra read in some situations, but hopefully
2766    it is a clever early read in most cases. */    it is needed in most cases. */
2767    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2768    /* Should not found a value between 128 and 192 here. */    jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2769    jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 192);    if (!update_str_ptr)
2770    add_jump(compiler, &common->utf8readtype8, JUMP(SLJIT_FAST_CALL));      {
2771        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2772        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2773        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2774        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2775        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2776        OP2(SLJIT_OR, TMP2, 0, TMP2, 0, TMP1, 0);
2777        OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2778        jump2 = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);
2779        OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2780        JUMPHERE(jump2);
2781        }
2782      else
2783        add_jump(compiler, &common->utfreadtype8, JUMP(SLJIT_FAST_CALL));
2784    JUMPHERE(jump);    JUMPHERE(jump);
2785    return;    return;
2786    }    }
2787    #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2788    
2789    #if !defined COMPILE_PCRE8
2790    /* The ctypes array contains only 256 values. */
2791    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2792    jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);
2793    #endif
2794    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2795    #if !defined COMPILE_PCRE8
2796    JUMPHERE(jump);
2797  #endif  #endif
2798  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  
2799  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);  #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2800  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), common->ctypes);  if (common->utf && update_str_ptr)
2801      {
2802      /* Skip low surrogate if necessary. */
2803      OP2(SLJIT_SUB, TMP2, 0, TMP2, 0, SLJIT_IMM, 0xd800);
2804      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2805      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2806      JUMPHERE(jump);
2807      }
2808    #endif /* SUPPORT_UTF && COMPILE_PCRE16 */
2809  }  }
2810    
2811  static void skip_char_back(compiler_common *common)  static void skip_char_back(compiler_common *common)
2812  {  {
2813  /* Goes one character back. Only affects STR_PTR. Does not check begin. */  /* Goes one character back. Affects STR_PTR and TMP1. Does not check begin. */
2814  DEFINE_COMPILER;  DEFINE_COMPILER;
2815  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2816    #if defined COMPILE_PCRE8
2817  struct sljit_label *label;  struct sljit_label *label;
2818    
2819  if (common->utf8)  if (common->utf)
2820    {    {
2821    label = LABEL();    label = LABEL();
2822    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
2823    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2824    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xc0);    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xc0);
2825    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, 0x80, label);    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, 0x80, label);
2826    return;    return;
2827    }    }
2828  #endif  #elif defined COMPILE_PCRE16
2829  OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);  if (common->utf)
2830      {
2831      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
2832      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2833      /* Skip low surrogate if necessary. */
2834      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
2835      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xdc00);
2836      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
2837      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
2838      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2839      return;
2840      }
2841    #endif /* COMPILE_PCRE[8|16] */
2842    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2843    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2844  }  }
2845    
2846  static void check_newlinechar(compiler_common *common, int nltype, jump_list **fallbacks, BOOL jumpiftrue)  static void check_newlinechar(compiler_common *common, int nltype, jump_list **backtracks, BOOL jumpifmatch)
2847  {  {
2848  /* Character comes in TMP1. Checks if it is a newline. TMP2 may be destroyed. */  /* Character comes in TMP1. Checks if it is a newline. TMP2 may be destroyed. */
2849  DEFINE_COMPILER;  DEFINE_COMPILER;
2850    struct sljit_jump *jump;
2851    
2852  if (nltype == NLTYPE_ANY)  if (nltype == NLTYPE_ANY)
2853    {    {
2854    add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));
2855    add_jump(compiler, fallbacks, JUMP(jumpiftrue ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));    add_jump(compiler, backtracks, JUMP(jumpifmatch ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));
2856    }    }
2857  else if (nltype == NLTYPE_ANYCRLF)  else if (nltype == NLTYPE_ANYCRLF)
2858    {    {
2859    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_CR);    if (jumpifmatch)
2860    COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_EQUAL);      {
2861    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_NL);      add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR));
2862    COND_VALUE(SLJIT_OR | SLJIT_SET_E, TMP2, 0, SLJIT_C_EQUAL);      add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL));
2863    add_jump(compiler, fallbacks, JUMP(jumpiftrue ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));      }
2864      else
2865        {
2866        jump = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR);
2867        add_jump(compiler, backtracks, CMP(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL));
2868        JUMPHERE(jump);
2869        }
2870    }    }
2871  else  else
2872    {    {
2873    SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline <= 255);    SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline < 256);
2874    add_jump(compiler, fallbacks, CMP(jumpiftrue ? SLJIT_C_EQUAL : SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, common->newline));    add_jump(compiler, backtracks, CMP(jumpifmatch ? SLJIT_C_EQUAL : SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, common->newline));
2875    }    }
2876  }  }
2877    
2878  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2879  static void do_utf8readchar(compiler_common *common)  
2880    #if defined COMPILE_PCRE8
2881    static void do_utfreadchar(compiler_common *common)
2882  {  {
2883  /* Fast decoding an utf8 character. TMP1 contains the first byte  /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2884  of the character (>= 192). Return char value in TMP1, length - 1 in TMP2. */  of the character (>= 0xc0). Return char value in TMP1, length in TMP2. */
2885  DEFINE_COMPILER;  DEFINE_COMPILER;
2886  struct sljit_jump *jump;  struct sljit_jump *jump;
2887    
2888  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2889  /* Searching for the first zero. */  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2890  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x20);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
 jump = JUMP(SLJIT_C_NOT_ZERO);  
 /* 2 byte sequence */  
 OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 1);  
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);  
 OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x1f);  
2891  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2892  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2893  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
 OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 1);  
 sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  
 JUMPHERE(jump);  
2894    
2895  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x10);  /* Searching for the first zero. */
2896    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
2897  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_C_NOT_ZERO);
2898  /* 3 byte sequence */  /* Two byte sequence. */
2899  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 1);  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2900  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x0f);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(2));
 OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 12);  
 OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  
 OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
 OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 2);  
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 2);  
 OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
 OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 2);  
2901  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
 JUMPHERE(jump);  
2902    
2903  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x08);  JUMPHERE(jump);
2904  jump = JUMP(SLJIT_C_NOT_ZERO);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2905  /* 4 byte sequence */  OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
2906  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 1);  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
 OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x07);  
 OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 18);  
 OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  
 OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 12);  
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
 OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 2);  
 OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  
 OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
 OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 3);  
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 3);  
2907  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2908  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2909  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 3);  
2910    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2911    jump = JUMP(SLJIT_C_NOT_ZERO);
2912    /* Three byte sequence. */
2913    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2914    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(3));
2915  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
 JUMPHERE(jump);  
2916    
2917  /* 5 byte sequence */  /* Four byte sequence. */
2918  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 1);  JUMPHERE(jump);
2919  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x03);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2920  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 24);  OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2921  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2922  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 18);  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
 OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 2);  
2923  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
 OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 12);  
2924  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2925  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 3);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(4));
2926    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2927    }
2928    
2929    static void do_utfreadchar16(compiler_common *common)
2930    {
2931    /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2932    of the character (>= 0xc0). Return value in TMP1. */
2933    DEFINE_COMPILER;
2934    struct sljit_jump *jump;
2935    
2936    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2937    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2938    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2939    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2940  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
 OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  
2941  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2942  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 4);  
2943  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 4);  /* Searching for the first zero. */
2944    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
2945    jump = JUMP(SLJIT_C_NOT_ZERO);
2946    /* Two byte sequence. */
2947    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2948    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2949    
2950    JUMPHERE(jump);
2951    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x400);
2952    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_NOT_ZERO);
2953    /* This code runs only in 8 bit mode. No need to shift the value. */
2954    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2955    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2956    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
2957    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2958  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2959  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2960  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 4);  /* Three byte sequence. */
2961    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2962  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2963  }  }
2964    
2965  static void do_utf8readtype8(compiler_common *common)  static void do_utfreadtype8(compiler_common *common)
2966  {  {
2967  /* Fast decoding an utf8 character type. TMP2 contains the first byte  /* Fast decoding a UTF-8 character type. TMP2 contains the first byte
2968  of the character (>= 192) and TMP1 is destroyed. Return value in TMP1. */  of the character (>= 0xc0). Return value in TMP1. */
2969  DEFINE_COMPILER;  DEFINE_COMPILER;
2970  struct sljit_jump *jump;  struct sljit_jump *jump;
2971  struct sljit_jump *compare;  struct sljit_jump *compare;
2972    
2973  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2974    
2975  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0x20);  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0x20);
2976  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_C_NOT_ZERO);
2977  /* 2 byte sequence */  /* Two byte sequence. */
2978  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2979  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2980  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);
2981    /* The upper 5 bits are known at this point. */
2982    compare = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0x3);
2983  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2984  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2985  OP2(SLJIT_OR, TMP2, 0, TMP2, 0, TMP1, 0);  OP2(SLJIT_OR, TMP2, 0, TMP2, 0, TMP1, 0);
 compare = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);  
2986  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2987  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2988    
2989  JUMPHERE(compare);  JUMPHERE(compare);
2990  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2991  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2992    
2993    /* We only have types for characters less than 256. */
2994  JUMPHERE(jump);  JUMPHERE(jump);
2995    OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2996    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2997    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2998    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2999    }
3000    
3001    #endif /* COMPILE_PCRE8 */
3002    
3003    #endif /* SUPPORT_UTF */
3004    
3005    #ifdef SUPPORT_UCP
3006    
3007    /* UCD_BLOCK_SIZE must be 128 (see the assert below). */
3008    #define UCD_BLOCK_MASK 127
3009    #define UCD_BLOCK_SHIFT 7
3010    
3011    static void do_getucd(compiler_common *common)
3012    {
3013    /* Search the UCD record for the character comes in TMP1.
3014    Returns chartype in TMP1 and UCD offset in TMP2. */
3015    DEFINE_COMPILER;
3016    
3017    SLJIT_ASSERT(UCD_BLOCK_SIZE == 128 && sizeof(ucd_record) == 8);
3018    
3019    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
3020    OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
3021    OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(ucd_stage1));
3022    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);
3023    OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
3024    OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, TMP2, 0);
3025    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_stage2));
3026    OP1(SLJIT_MOV_UH, TMP2, 0, SLJIT_MEM2(TMP2, TMP1), 1);
3027    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_records) + SLJIT_OFFSETOF(ucd_record, chartype));
3028    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM2(TMP1, TMP2), 3);
3029    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3030    }
3031    #endif
3032    
3033    static SLJIT_INLINE struct sljit_label *mainloop_entry(compiler_common *common, BOOL hascrorlf, BOOL firstline)
3034    {
3035    DEFINE_COMPILER;
3036    struct sljit_label *mainloop;
3037    struct sljit_label *newlinelabel = NULL;
3038    struct sljit_jump *start;
3039    struct sljit_jump *end = NULL;
3040    struct sljit_jump *nl = NULL;
3041    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3042    struct sljit_jump *singlechar;
3043    #endif
3044    jump_list *newline = NULL;
3045    BOOL newlinecheck = FALSE;
3046    BOOL readuchar = FALSE;
3047    
3048    if (!(hascrorlf || firstline) && (common->nltype == NLTYPE_ANY ||
3049        common->nltype == NLTYPE_ANYCRLF || common->newline > 255))
3050      newlinecheck = TRUE;
3051    
3052    if (firstline)
3053      {
3054      /* Search for the end of the first line. */
3055      SLJIT_ASSERT(common->first_line_end != 0);
3056      OP1(SLJIT_MOV, TMP3, 0, STR_PTR, 0);
3057    
3058      if (common->nltype == NLTYPE_FIXED && common->newline > 255)
3059        {
3060        mainloop = LABEL();
3061        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3062        end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3063        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(-1));
3064        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
3065        CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, mainloop);
3066        CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, mainloop);
3067        JUMPHERE(end);
3068        OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3069        }
3070      else
3071        {
3072        end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3073        mainloop = LABEL();
3074        /* Continual stores does not cause data dependency. */
3075        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);
3076        read_char_range(common, common->nlmin, common->nlmax, TRUE);
3077        check_newlinechar(common, common->nltype, &newline, TRUE);
3078        CMPTO(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0, mainloop);
3079        JUMPHERE(end);
3080        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);
3081        set_jumps(newline, LABEL());
3082        }
3083    
3084      OP1(SLJIT_MOV, STR_PTR, 0, TMP3, 0);
3085      }
3086    
3087    start = JUMP(SLJIT_JUMP);
3088    
3089    if (newlinecheck)
3090      {
3091      newlinelabel = LABEL();
3092      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3093      end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3094      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3095      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, common->newline & 0xff);
3096      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
3097    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
3098      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, UCHAR_SHIFT);
3099    #endif
3100      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3101      nl = JUMP(SLJIT_JUMP);
3102      }
3103    
3104    mainloop = LABEL();
3105    
3106    /* Increasing the STR_PTR here requires one less jump in the most common case. */
3107    #ifdef SUPPORT_UTF
3108    if (common->utf) readuchar = TRUE;
3109    #endif
3110    if (newlinecheck) readuchar = TRUE;
3111    
3112    if (readuchar)
3113      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3114    
3115    if (newlinecheck)
3116      CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, newlinelabel);
3117    
3118    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3119    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3120    #if defined COMPILE_PCRE8
3121    if (common->utf)
3122      {
3123      singlechar = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
3124      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
3125      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3126      JUMPHERE(singlechar);
3127      }
3128    #elif defined COMPILE_PCRE16
3129    if (common->utf)
3130      {
3131      singlechar = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);
3132      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
3133      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);
3134      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
3135      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
3136      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3137      JUMPHERE(singlechar);
3138      }
3139    #endif /* COMPILE_PCRE[8|16] */
3140    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
3141    JUMPHERE(start);
3142    
3143    if (newlinecheck)
3144      {
3145      JUMPHERE(end);
3146      JUMPHERE(nl);
3147      }
3148    
3149    return mainloop;
3150    }
3151    
3152    static int scan_prefix(compiler_common *common, pcre_uchar *cc, pcre_uint32 *chars, int max_chars)
3153    {
3154    /* Recursive function, which scans prefix literals. */
3155    int len, repeat, len_save, consumed = 0;
3156    pcre_uint32 caseless, chr, mask;
3157    pcre_uchar *alternative, *cc_save;
3158    BOOL last, any;
3159    
3160    repeat = 1;
3161    while (TRUE)
3162      {
3163      last = TRUE;
3164      any = FALSE;
3165      caseless = 0;
3166      switch (*cc)
3167        {
3168        case OP_CHARI:
3169        caseless = 1;
3170        case OP_CHAR:
3171        last = FALSE;
3172        cc++;
3173        break;
3174    
3175        case OP_SOD:
3176        case OP_SOM:
3177        case OP_SET_SOM:
3178        case OP_NOT_WORD_BOUNDARY:
3179        case OP_WORD_BOUNDARY:
3180        case OP_EODN:
3181        case OP_EOD:
3182        case OP_CIRC:
3183        case OP_CIRCM:
3184        case OP_DOLL:
3185        case OP_DOLLM:
3186        /* Zero width assertions. */
3187        cc++;
3188        continue;
3189    
3190        case OP_PLUS:
3191        case OP_MINPLUS:
3192        case OP_POSPLUS:
3193        cc++;
3194        break;
3195    
3196        case OP_EXACTI:
3197        caseless = 1;
3198        case OP_EXACT:
3199        repeat = GET2(cc, 1);
3200        last = FALSE;
3201        cc += 1 + IMM2_SIZE;
3202        break;
3203    
3204        case OP_PLUSI:
3205        case OP_MINPLUSI:
3206        case OP_POSPLUSI:
3207        caseless = 1;
3208        cc++;
3209        break;
3210    
3211        case OP_KET:
3212        cc += 1 + LINK_SIZE;
3213        continue;
3214    
3215        case OP_ALT:
3216        cc += GET(cc, 1);
3217        continue;
3218    
3219        case OP_ONCE:
3220        case OP_ONCE_NC:
3221        case OP_BRA:
3222        case OP_BRAPOS:
3223        case OP_CBRA:
3224        case OP_CBRAPOS:
3225        alternative = cc + GET(cc, 1);
3226        while (*alternative == OP_ALT)
3227          {
3228          max_chars = scan_prefix(common, alternative + 1 + LINK_SIZE, chars, max_chars);
3229          if (max_chars == 0)
3230            return consumed;
3231          alternative += GET(alternative, 1);
3232          }
3233    
3234        if (*cc == OP_CBRA || *cc == OP_CBRAPOS)
3235          cc += IMM2_SIZE;
3236        cc += 1 + LINK_SIZE;
3237        continue;
3238    
3239        case OP_CLASS:
3240        case OP_NCLASS:
3241        any = TRUE;
3242        cc += 1 + 32 / sizeof(pcre_uchar);
3243        break;
3244    
3245    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3246        case OP_XCLASS:
3247        any = TRUE;
3248        cc += GET(cc, 1);
3249        break;
3250    #endif
3251    
3252        case OP_NOT_DIGIT:
3253        case OP_DIGIT:
3254        case OP_NOT_WHITESPACE:
3255        case OP_WHITESPACE:
3256        case OP_NOT_WORDCHAR:
3257        case OP_WORDCHAR:
3258        case OP_ANY:
3259        case OP_ALLANY:
3260        any = TRUE;
3261        cc++;
3262        break;
3263    
3264    #ifdef SUPPORT_UCP
3265        case OP_NOTPROP:
3266        case OP_PROP:
3267        any = TRUE;
3268        cc += 1 + 2;
3269        break;
3270    #endif
3271    
3272        case OP_TYPEEXACT:
3273        repeat = GET2(cc, 1);
3274        cc += 1 + IMM2_SIZE;
3275        continue;
3276    
3277        default:
3278        return consumed;
3279        }
3280    
3281      if (any)
3282        {
3283    #ifdef SUPPORT_UTF
3284        if (common->utf) return consumed;
3285    #endif
3286    #if defined COMPILE_PCRE8
3287        mask = 0xff;
3288    #elif defined COMPILE_PCRE16
3289        mask = 0xffff;
3290    #elif defined COMPILE_PCRE32
3291        mask = 0xffffffff;
3292    #else
3293        SLJIT_ASSERT_STOP();
3294    #endif
3295    
3296        do
3297          {
3298          chars[0] = mask;
3299          chars[1] = mask;
3300    
3301          if (--max_chars == 0)
3302            return consumed;
3303          consumed++;
3304          chars += 2;
3305          }
3306        while (--repeat > 0);
3307    
3308  /* We only have types for characters less than 256. */      repeat = 1;
3309  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), (sljit_w)_pcre_utf8_char_sizes);      continue;
3310  OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);      }
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);  
 OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);  
 sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  
 }  
3311    
3312      len = 1;
3313    #ifdef SUPPORT_UTF
3314      if (common->utf && HAS_EXTRALEN(*cc)) len += GET_EXTRALEN(*cc);
3315  #endif  #endif
3316    
3317  #ifdef SUPPORT_UCP    if (caseless != 0 && char_has_othercase(common, cc))
3318        {
3319        caseless = char_get_othercase_bit(common, cc);
3320        if (caseless == 0)
3321          return consumed;
3322    #ifdef COMPILE_PCRE8
3323        caseless = ((caseless & 0xff) << 8) | (len - (caseless >> 8));
3324    #else
3325        if ((caseless & 0x100) != 0)
3326          caseless = ((caseless & 0xff) << 16) | (len - (caseless >> 9));
3327        else
3328          caseless = ((caseless & 0xff) << 8) | (len - (caseless >> 9));
3329    #endif
3330        }
3331      else
3332        caseless = 0;
3333    
3334  /* UCD_BLOCK_SIZE must be 128 (see the assert below). */    len_save = len;
3335  #define UCD_BLOCK_MASK 127    cc_save = cc;
3336  #define UCD_BLOCK_SHIFT 7    while (TRUE)
3337        {
3338        do
3339          {
3340          chr = *cc;
3341    #ifdef COMPILE_PCRE32
3342          if (SLJIT_UNLIKELY(chr == NOTACHAR))
3343            return consumed;
3344    #endif
3345          mask = 0;
3346          if ((pcre_uint32)len == (caseless & 0xff))
3347            {
3348            mask = caseless >> 8;
3349            chr |= mask;
3350            }
3351    
3352  static void do_getucd(compiler_common *common)        if (chars[0] == NOTACHAR)
3353  {          {
3354  /* Search the UCD record for the character comes in TMP1.          chars[0] = chr;
3355  Returns chartype in TMP1 and UCD offset in TMP2. */          chars[1] = mask;
3356  DEFINE_COMPILER;          }
3357          else
3358            {
3359            mask |= chars[0] ^ chr;
3360            chr |= mask;
3361            chars[0] = chr;
3362            chars[1] |= mask;
3363            }
3364    
3365  SLJIT_ASSERT(UCD_BLOCK_SIZE == 128 && sizeof(ucd_record) == 8);        len--;
3366          if (--max_chars == 0)
3367            return consumed;
3368          consumed++;
3369          chars += 2;
3370          cc++;
3371          }
3372        while (len > 0);
3373    
3374  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);      if (--repeat == 0)
3375  OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);        break;
3376  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_w)_pcre_ucd_stage1);  
3377  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);      len = len_save;
3378  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);      cc = cc_save;
3379  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, TMP2, 0);      }
3380  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, (sljit_w)_pcre_ucd_stage2);  
3381  OP1(SLJIT_MOV_UH, TMP2, 0, SLJIT_MEM2(TMP2, TMP1), 1);    repeat = 1;
3382  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, (sljit_w)_pcre_ucd_records + SLJIT_OFFSETOF(ucd_record, chartype));    if (last)
3383  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM2(TMP1, TMP2), 3);      return consumed;
3384  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);    }
3385  }  }
 #endif  
3386    
3387  static SLJIT_INLINE struct sljit_label *mainloop_entry(compiler_common *common, BOOL hascrorlf, BOOL firstline)  #define MAX_N_CHARS 16
3388    
3389    static SLJIT_INLINE BOOL fast_forward_first_n_chars(compiler_common *common, BOOL firstline)
3390  {  {
3391  DEFINE_COMPILER;  DEFINE_COMPILER;
3392  struct sljit_label *mainloop;  struct sljit_label *start;
3393  struct sljit_label *newlinelabel = NULL;  struct sljit_jump *quit;
3394  struct sljit_jump *start;  pcre_uint32 chars[MAX_N_CHARS * 2];
3395  struct sljit_jump *end = NULL;  pcre_uint8 ones[MAX_N_CHARS];
3396  struct sljit_jump *nl = NULL;  pcre_uint32 mask;
3397  jump_list *newline = NULL;  int i, max;
3398  BOOL newlinecheck = FALSE;  int offsets[3];
 BOOL readbyte = FALSE;  
   
 if (!(hascrorlf || firstline) && (common->nltype == NLTYPE_ANY ||  
     common->nltype == NLTYPE_ANYCRLF || common->newline > 255))  
   newlinecheck = TRUE;  
3399    
3400  if (firstline)  for (i = 0; i < MAX_N_CHARS; i++)
3401    {    {
3402    /* Search for the end of the first line. */    chars[i << 1] = NOTACHAR;
3403    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, STR_PTR, 0);    chars[(i << 1) + 1] = 0;
3404    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), FIRSTLINE_END, STR_END, 0);    }
3405    
3406    if (common->nltype == NLTYPE_FIXED && common->newline > 255)  max = scan_prefix(common, common->start, chars, MAX_N_CHARS);
3407    
3408    if (max <= 1)
3409      return FALSE;
3410    
3411    for (i = 0; i < max; i++)
3412      {
3413      mask = chars[(i << 1) + 1];
3414      ones[i] = ones_in_half_byte[mask & 0xf];
3415      mask >>= 4;
3416      while (mask != 0)
3417      {      {
3418      mainloop = LABEL();      ones[i] += ones_in_half_byte[mask & 0xf];
3419      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);      mask >>= 4;
     end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);  
     OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), -1);  
     OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);  
     CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, mainloop);  
     CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, mainloop);  
     OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_LOCALS_REG), FIRSTLINE_END, STR_PTR, 0, SLJIT_IMM, 1);  
3420      }      }
3421    else    }
3422    
3423    offsets[0] = -1;
3424    /* Scan forward. */
3425    for (i = 0; i < max; i++)
3426      if (ones[i] <= 2) {
3427        offsets[0] = i;
3428        break;
3429      }
3430    
3431    if (offsets[0] == -1)
3432      return FALSE;
3433    
3434    /* Scan backward. */
3435    offsets[1] = -1;
3436    for (i = max - 1; i > offsets[0]; i--)
3437      if (ones[i] <= 2) {