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

Diff of /code/trunk/pcre_jit_compile.c

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

revision 941 by zherczeg, Tue Feb 28 11:33:34 2012 UTC revision 1474 by zherczeg, Thu Apr 24 06:43:50 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) 1997-2012 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) 2010-2012                        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
# Line 65  system files. */ Line 65  system files. */
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 82  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 108  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 {
# Line 154  typedef struct jit_arguments { Line 166  typedef struct jit_arguments {
166    int *offsets;    int *offsets;
167    pcre_uchar *uchar_ptr;    pcre_uchar *uchar_ptr;
168    pcre_uchar *mark_ptr;    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      int offset_count;
174    pcre_uint8 notbol;    pcre_uint8 notbol;
175    pcre_uint8 noteol;    pcre_uint8 noteol;
176    pcre_uint8 notempty;    pcre_uint8 notempty;
# Line 165  typedef struct jit_arguments { Line 179  typedef struct jit_arguments {
179    
180  typedef struct executable_functions {  typedef struct executable_functions {
181    void *executable_funcs[JIT_NUMBER_OF_COMPILE_MODES];    void *executable_funcs[JIT_NUMBER_OF_COMPILE_MODES];
182      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;    PUBL(jit_callback) callback;
185    void *userdata;    void *userdata;
186    sljit_uw executable_sizes[JIT_NUMBER_OF_COMPILE_MODES];    pcre_uint32 top_bracket;
187      pcre_uint32 limit_match;
188  } executable_functions;  } executable_functions;
189    
190  typedef struct jump_list {  typedef struct jump_list {
# Line 175  typedef struct jump_list { Line 192  typedef struct jump_list {
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    pcre_uchar *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 261  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      /* First byte code. */
322    pcre_uchar *start;    pcre_uchar *start;
323      /* Maps private data offset to each opcode. */
324    /* Local stack area size and variable pointers. */    sljit_si *private_data_ptrs;
325    int localsize;    /* This read-only data is available during runtime. */
326    int *localptrs;    sljit_uw *read_only_data;
327    int cbraptr;    /* The total size of the read-only data. */
328    /* OVector starting point. Must be divisible by 2. */    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;    int ovector_start;
341    /* Last known position of the requested byte. */    /* Last known position of the requested byte. */
342    int req_char_ptr;    int req_char_ptr;
343    /* Head of the last recursion. */    /* Head of the last recursion. */
344    int recursive_head;    int recursive_head_ptr;
345    /* First inspected character for partial matching. */    /* First inspected character for partial matching. */
346    int start_used_ptr;    int start_used_ptr;
347    /* Starting pointer for partial soft matches. */    /* Starting pointer for partial soft matches. */
# Line 290  typedef struct compiler_common { Line 350  typedef struct compiler_common {
350    int first_line_end;    int first_line_end;
351    /* Points to the marked string. */    /* Points to the marked string. */
352    int mark_ptr;    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    /* Other  */    /* Flipped and lower case tables. */
361    const pcre_uint8 *fcc;    const pcre_uint8 *fcc;
362    sljit_w lcc;    sljit_sw lcc;
363      /* Mode can be PCRE_STUDY_JIT_COMPILE and others. */
364    int mode;    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    BOOL has_set_som;    /* Tables. */
390    sljit_w ctypes;    sljit_sw ctypes;
391    sljit_uw name_table;    /* Named capturing brackets. */
392    sljit_w name_count;    pcre_uchar *name_table;
393    sljit_w name_entry_size;    sljit_sw name_count;
394      sljit_sw name_entry_size;
395    
396    /* Labels and jump lists. */    /* Labels and jump lists. */
397    struct sljit_label *partialmatchlabel;    struct sljit_label *partialmatchlabel;
398    struct sljit_label *leavelabel;    struct sljit_label *quit_label;
399    struct sljit_label *acceptlabel;    struct sljit_label *forced_quit_label;
400      struct sljit_label *accept_label;
401      struct sljit_label *ff_newline_shortcut;
402    stub_list *stubs;    stub_list *stubs;
403      label_addr_list *label_addrs;
404    recurse_entry *entries;    recurse_entry *entries;
405    recurse_entry *currententry;    recurse_entry *currententry;
406    jump_list *partialmatch;    jump_list *partialmatch;
407    jump_list *leave;    jump_list *quit;
408      jump_list *positive_assert_quit;
409      jump_list *forced_quit;
410    jump_list *accept;    jump_list *accept;
411    jump_list *calllimit;    jump_list *calllimit;
412    jump_list *stackalloc;    jump_list *stackalloc;
# Line 324  typedef struct compiler_common { Line 417  typedef struct compiler_common {
417    jump_list *vspace;    jump_list *vspace;
418    jump_list *casefulcmp;    jump_list *casefulcmp;
419    jump_list *caselesscmp;    jump_list *caselesscmp;
420      jump_list *reset_match;
421    BOOL jscript_compat;    BOOL jscript_compat;
422  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
423    BOOL utf;    BOOL utf;
424  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
425    BOOL use_ucp;    BOOL use_ucp;
426  #endif  #endif
   jump_list *utfreadchar;  
427  #ifdef COMPILE_PCRE8  #ifdef COMPILE_PCRE8
428      jump_list *utfreadchar;
429      jump_list *utfreadchar16;
430    jump_list *utfreadtype8;    jump_list *utfreadtype8;
431  #endif  #endif
432  #endif /* SUPPORT_UTF */  #endif /* SUPPORT_UTF */
# Line 348  typedef struct compare_context { Line 443  typedef struct compare_context {
443  #if defined SLJIT_UNALIGNED && SLJIT_UNALIGNED  #if defined SLJIT_UNALIGNED && SLJIT_UNALIGNED
444    int ucharptr;    int ucharptr;
445    union {    union {
446      sljit_i asint;      sljit_si asint;
447      sljit_uh asushort;      sljit_uh asushort;
448  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
449      sljit_ub asbyte;      sljit_ub asbyte;
450      sljit_ub asuchars[4];      sljit_ub asuchars[4];
451  #else  #elif defined COMPILE_PCRE16
 #ifdef COMPILE_PCRE16  
452      sljit_uh asuchars[2];      sljit_uh asuchars[2];
453  #endif  #elif defined COMPILE_PCRE32
454        sljit_ui asuchars[1];
455  #endif  #endif
456    } c;    } c;
457    union {    union {
458      sljit_i asint;      sljit_si asint;
459      sljit_uh asushort;      sljit_uh asushort;
460  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
461      sljit_ub asbyte;      sljit_ub asbyte;
462      sljit_ub asuchars[4];      sljit_ub asuchars[4];
463  #else  #elif defined COMPILE_PCRE16
 #ifdef COMPILE_PCRE16  
464      sljit_uh asuchars[2];      sljit_uh asuchars[2];
465  #endif  #elif defined COMPILE_PCRE32
466        sljit_ui asuchars[1];
467  #endif  #endif
468    } oc;    } oc;
469  #endif  #endif
470  } compare_context;  } compare_context;
471    
 enum {  
   frame_end = 0,  
   frame_setstrbegin = -1,  
   frame_setmark = -2  
 };  
   
472  /* Undefine sljit macros. */  /* Undefine sljit macros. */
473  #undef CMP  #undef CMP
474    
475  /* Used for accessing the elements of the stack. */  /* Used for accessing the elements of the stack. */
476  #define STACK(i)      ((-(i) - 1) * (int)sizeof(sljit_w))  #define STACK(i)      ((-(i) - 1) * (int)sizeof(sljit_sw))
477    
478  #define TMP1          SLJIT_TEMPORARY_REG1  #define TMP1          SLJIT_SCRATCH_REG1
479  #define TMP2          SLJIT_TEMPORARY_REG3  #define TMP2          SLJIT_SCRATCH_REG3
480  #define TMP3          SLJIT_TEMPORARY_EREG2  #define TMP3          SLJIT_TEMPORARY_EREG2
481  #define STR_PTR       SLJIT_SAVED_REG1  #define STR_PTR       SLJIT_SAVED_REG1
482  #define STR_END       SLJIT_SAVED_REG2  #define STR_END       SLJIT_SAVED_REG2
483  #define STACK_TOP     SLJIT_TEMPORARY_REG2  #define STACK_TOP     SLJIT_SCRATCH_REG2
484  #define STACK_LIMIT   SLJIT_SAVED_REG3  #define STACK_LIMIT   SLJIT_SAVED_REG3
485  #define ARGUMENTS     SLJIT_SAVED_EREG1  #define ARGUMENTS     SLJIT_SAVED_EREG1
486  #define CALL_COUNT    SLJIT_SAVED_EREG2  #define COUNT_MATCH   SLJIT_SAVED_EREG2
487  #define RETURN_ADDR   SLJIT_TEMPORARY_EREG1  #define RETURN_ADDR   SLJIT_TEMPORARY_EREG1
488    
489  /* Locals layout. */  /* Local space layout. */
490  /* These two locals can be used by the current opcode. */  /* These two locals can be used by the current opcode. */
491  #define LOCALS0          (0 * sizeof(sljit_w))  #define LOCALS0          (0 * sizeof(sljit_sw))
492  #define LOCALS1          (1 * sizeof(sljit_w))  #define LOCALS1          (1 * sizeof(sljit_sw))
493  /* Two local variables for possessive quantifiers (char1 cannot use them). */  /* Two local variables for possessive quantifiers (char1 cannot use them). */
494  #define POSSESSIVE0      (2 * sizeof(sljit_w))  #define POSSESSIVE0      (2 * sizeof(sljit_sw))
495  #define POSSESSIVE1      (3 * sizeof(sljit_w))  #define POSSESSIVE1      (3 * sizeof(sljit_sw))
496  /* Max limit of recursions. */  /* Max limit of recursions. */
497  #define CALL_LIMIT       (4 * sizeof(sljit_w))  #define LIMIT_MATCH      (4 * sizeof(sljit_sw))
498  /* The output vector is stored on the stack, and contains pointers  /* The output vector is stored on the stack, and contains pointers
499  to characters. The vector data is divided into two groups: the first  to characters. The vector data is divided into two groups: the first
500  group contains the start / end character pointers, and the second is  group contains the start / end character pointers, and the second is
501  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. */
502  #define OVECTOR_START    (common->ovector_start)  #define OVECTOR_START    (common->ovector_start)
503  #define OVECTOR(i)       (OVECTOR_START + (i) * sizeof(sljit_w))  #define OVECTOR(i)       (OVECTOR_START + (i) * (sljit_sw)sizeof(sljit_sw))
504  #define OVECTOR_PRIV(i)  (common->cbraptr + (i) * sizeof(sljit_w))  #define OVECTOR_PRIV(i)  (common->cbra_ptr + (i) * (sljit_sw)sizeof(sljit_sw))
505  #define PRIV_DATA(cc)    (common->localptrs[(cc) - common->start])  #define PRIVATE_DATA(cc) (common->private_data_ptrs[(cc) - common->start])
506    
507  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
508  #define MOV_UCHAR  SLJIT_MOV_UB  #define MOV_UCHAR  SLJIT_MOV_UB
509  #define MOVU_UCHAR SLJIT_MOVU_UB  #define MOVU_UCHAR SLJIT_MOVU_UB
510  #else  #elif defined COMPILE_PCRE16
 #ifdef COMPILE_PCRE16  
511  #define MOV_UCHAR  SLJIT_MOV_UH  #define MOV_UCHAR  SLJIT_MOV_UH
512  #define MOVU_UCHAR SLJIT_MOVU_UH  #define MOVU_UCHAR SLJIT_MOVU_UH
513    #elif defined COMPILE_PCRE32
514    #define MOV_UCHAR  SLJIT_MOV_UI
515    #define MOVU_UCHAR SLJIT_MOVU_UI
516  #else  #else
517  #error Unsupported compiling mode  #error Unsupported compiling mode
518  #endif  #endif
 #endif  
519    
520  /* Shortcuts. */  /* Shortcuts. */
521  #define DEFINE_COMPILER \  #define DEFINE_COMPILER \
# Line 442  the start pointers when the end of the c Line 532  the start pointers when the end of the c
532    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))
533  #define JUMPHERE(jump) \  #define JUMPHERE(jump) \
534    sljit_set_label((jump), sljit_emit_label(compiler))    sljit_set_label((jump), sljit_emit_label(compiler))
535    #define SET_LABEL(jump, label) \
536      sljit_set_label((jump), (label))
537  #define CMP(type, src1, src1w, src2, src2w) \  #define CMP(type, src1, src1w, src2, src2w) \
538    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))
539  #define CMPTO(type, src1, src1w, src2, src2w, label) \  #define CMPTO(type, src1, src1w, src2, src2w, label) \
540    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))
541  #define COND_VALUE(op, dst, dstw, type) \  #define OP_FLAGS(op, dst, dstw, src, srcw, type) \
542    sljit_emit_cond_value(compiler, (op), (dst), (dstw), (type))    sljit_emit_op_flags(compiler, (op), (dst), (dstw), (src), (srcw), (type))
543    #define GET_LOCAL_BASE(dst, dstw, offset) \
544      sljit_get_local_base(compiler, (dst), (dstw), (offset))
545    
546    #define READ_CHAR_MAX 0x7fffffff
547    
548  static pcre_uchar* bracketend(pcre_uchar* cc)  static pcre_uchar* bracketend(pcre_uchar* cc)
549  {  {
# Line 458  cc += 1 + LINK_SIZE; Line 554  cc += 1 + LINK_SIZE;
554  return cc;  return cc;
555  }  }
556    
557    static int no_alternatives(pcre_uchar* cc)
558    {
559    int count = 0;
560    SLJIT_ASSERT((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND));
561    do
562      {
563      cc += GET(cc, 1);
564      count++;
565      }
566    while (*cc == OP_ALT);
567    SLJIT_ASSERT(*cc >= OP_KET && *cc <= OP_KETRPOS);
568    return count;
569    }
570    
571    static int ones_in_half_byte[16] = {
572      /* 0 */ 0, 1, 1, 2, /* 4 */ 1, 2, 2, 3,
573      /* 8 */ 1, 2, 2, 3, /* 12 */ 2, 3, 3, 4
574    };
575    
576  /* Functions whose might need modification for all new supported opcodes:  /* Functions whose might need modification for all new supported opcodes:
577   next_opcode   next_opcode
578   get_localspace   check_opcode_types
579   set_localptrs   set_private_data_ptrs
580   get_framesize   get_framesize
581   init_frame   init_frame
582   get_localsize   get_private_data_copy_length
583   copy_locals   copy_private_data
584   compile_hotpath   compile_matchingpath
585   compile_fallbackpath   compile_backtrackingpath
586  */  */
587    
588  static pcre_uchar *next_opcode(compiler_common *common, pcre_uchar *cc)  static pcre_uchar *next_opcode(compiler_common *common, pcre_uchar *cc)
# Line 488  switch(*cc) Line 603  switch(*cc)
603    case OP_WORDCHAR:    case OP_WORDCHAR:
604    case OP_ANY:    case OP_ANY:
605    case OP_ALLANY:    case OP_ALLANY:
606      case OP_NOTPROP:
607      case OP_PROP:
608    case OP_ANYNL:    case OP_ANYNL:
609    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
610    case OP_HSPACE:    case OP_HSPACE:
# Line 500  switch(*cc) Line 617  switch(*cc)
617    case OP_CIRCM:    case OP_CIRCM:
618    case OP_DOLL:    case OP_DOLL:
619    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:  
620    case OP_CRSTAR:    case OP_CRSTAR:
621    case OP_CRMINSTAR:    case OP_CRMINSTAR:
622    case OP_CRPLUS:    case OP_CRPLUS:
623    case OP_CRMINPLUS:    case OP_CRMINPLUS:
624    case OP_CRQUERY:    case OP_CRQUERY:
625    case OP_CRMINQUERY:    case OP_CRMINQUERY:
626      case OP_CRRANGE:
627      case OP_CRMINRANGE:
628      case OP_CRPOSSTAR:
629      case OP_CRPOSPLUS:
630      case OP_CRPOSQUERY:
631      case OP_CRPOSRANGE:
632      case OP_CLASS:
633      case OP_NCLASS:
634      case OP_REF:
635      case OP_REFI:
636      case OP_DNREF:
637      case OP_DNREFI:
638      case OP_RECURSE:
639      case OP_CALLOUT:
640      case OP_ALT:
641      case OP_KET:
642      case OP_KETRMAX:
643      case OP_KETRMIN:
644      case OP_KETRPOS:
645      case OP_REVERSE:
646      case OP_ASSERT:
647      case OP_ASSERT_NOT:
648      case OP_ASSERTBACK:
649      case OP_ASSERTBACK_NOT:
650      case OP_ONCE:
651      case OP_ONCE_NC:
652      case OP_BRA:
653      case OP_BRAPOS:
654      case OP_CBRA:
655      case OP_CBRAPOS:
656      case OP_COND:
657      case OP_SBRA:
658      case OP_SBRAPOS:
659      case OP_SCBRA:
660      case OP_SCBRAPOS:
661      case OP_SCOND:
662      case OP_CREF:
663      case OP_DNCREF:
664      case OP_RREF:
665      case OP_DNRREF:
666    case OP_DEF:    case OP_DEF:
667    case OP_BRAZERO:    case OP_BRAZERO:
668    case OP_BRAMINZERO:    case OP_BRAMINZERO:
669    case OP_BRAPOSZERO:    case OP_BRAPOSZERO:
670      case OP_PRUNE:
671      case OP_SKIP:
672      case OP_THEN:
673    case OP_COMMIT:    case OP_COMMIT:
674    case OP_FAIL:    case OP_FAIL:
675    case OP_ACCEPT:    case OP_ACCEPT:
676    case OP_ASSERT_ACCEPT:    case OP_ASSERT_ACCEPT:
677      case OP_CLOSE:
678    case OP_SKIPZERO:    case OP_SKIPZERO:
679    return cc + 1;    return cc + PRIV(OP_lengths)[*cc];
   
   case OP_ANYBYTE:  
 #ifdef SUPPORT_UTF  
   if (common->utf) return NULL;  
 #endif  
   return cc + 1;  
680    
681    case OP_CHAR:    case OP_CHAR:
682    case OP_CHARI:    case OP_CHARI:
# Line 542  switch(*cc) Line 688  switch(*cc)
688    case OP_MINPLUS:    case OP_MINPLUS:
689    case OP_QUERY:    case OP_QUERY:
690    case OP_MINQUERY:    case OP_MINQUERY:
691      case OP_UPTO:
692      case OP_MINUPTO:
693      case OP_EXACT:
694    case OP_POSSTAR:    case OP_POSSTAR:
695    case OP_POSPLUS:    case OP_POSPLUS:
696    case OP_POSQUERY:    case OP_POSQUERY:
697      case OP_POSUPTO:
698    case OP_STARI:    case OP_STARI:
699    case OP_MINSTARI:    case OP_MINSTARI:
700    case OP_PLUSI:    case OP_PLUSI:
701    case OP_MINPLUSI:    case OP_MINPLUSI:
702    case OP_QUERYI:    case OP_QUERYI:
703    case OP_MINQUERYI:    case OP_MINQUERYI:
704      case OP_UPTOI:
705      case OP_MINUPTOI:
706      case OP_EXACTI:
707    case OP_POSSTARI:    case OP_POSSTARI:
708    case OP_POSPLUSI:    case OP_POSPLUSI:
709    case OP_POSQUERYI:    case OP_POSQUERYI:
710      case OP_POSUPTOI:
711    case OP_NOTSTAR:    case OP_NOTSTAR:
712    case OP_NOTMINSTAR:    case OP_NOTMINSTAR:
713    case OP_NOTPLUS:    case OP_NOTPLUS:
714    case OP_NOTMINPLUS:    case OP_NOTMINPLUS:
715    case OP_NOTQUERY:    case OP_NOTQUERY:
716    case OP_NOTMINQUERY:    case OP_NOTMINQUERY:
717      case OP_NOTUPTO:
718      case OP_NOTMINUPTO:
719      case OP_NOTEXACT:
720    case OP_NOTPOSSTAR:    case OP_NOTPOSSTAR:
721    case OP_NOTPOSPLUS:    case OP_NOTPOSPLUS:
722    case OP_NOTPOSQUERY:    case OP_NOTPOSQUERY:
723      case OP_NOTPOSUPTO:
724    case OP_NOTSTARI:    case OP_NOTSTARI:
725    case OP_NOTMINSTARI:    case OP_NOTMINSTARI:
726    case OP_NOTPLUSI:    case OP_NOTPLUSI:
727    case OP_NOTMINPLUSI:    case OP_NOTMINPLUSI:
728    case OP_NOTQUERYI:    case OP_NOTQUERYI:
729    case OP_NOTMINQUERYI:    case OP_NOTMINQUERYI:
   case OP_NOTPOSSTARI:  
   case OP_NOTPOSPLUSI:  
   case OP_NOTPOSQUERYI:  
   cc += 2;  
 #ifdef SUPPORT_UTF  
   if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);  
 #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:  
730    case OP_NOTUPTOI:    case OP_NOTUPTOI:
731    case OP_NOTMINUPTOI:    case OP_NOTMINUPTOI:
732    case OP_NOTEXACTI:    case OP_NOTEXACTI:
733      case OP_NOTPOSSTARI:
734      case OP_NOTPOSPLUSI:
735      case OP_NOTPOSQUERYI:
736    case OP_NOTPOSUPTOI:    case OP_NOTPOSUPTOI:
737    cc += 2 + IMM2_SIZE;    cc += PRIV(OP_lengths)[*cc];
738  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
739    if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);    if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
740  #endif  #endif
741    return cc;    return cc;
742    
743    case OP_NOTPROP:    /* Special cases. */
744    case OP_PROP:    case OP_TYPESTAR:
745    return cc + 1 + 2;    case OP_TYPEMINSTAR:
746      case OP_TYPEPLUS:
747      case OP_TYPEMINPLUS:
748      case OP_TYPEQUERY:
749      case OP_TYPEMINQUERY:
750    case OP_TYPEUPTO:    case OP_TYPEUPTO:
751    case OP_TYPEMINUPTO:    case OP_TYPEMINUPTO:
752    case OP_TYPEEXACT:    case OP_TYPEEXACT:
753      case OP_TYPEPOSSTAR:
754      case OP_TYPEPOSPLUS:
755      case OP_TYPEPOSQUERY:
756    case OP_TYPEPOSUPTO:    case OP_TYPEPOSUPTO:
757    case OP_REF:    return cc + PRIV(OP_lengths)[*cc] - 1;
   case OP_REFI:  
   case OP_CREF:  
   case OP_NCREF:  
   case OP_RREF:  
   case OP_NRREF:  
   case OP_CLOSE:  
   cc += 1 + IMM2_SIZE;  
   return cc;  
   
   case OP_CRRANGE:  
   case OP_CRMINRANGE:  
   return cc + 1 + 2 * IMM2_SIZE;  
758    
759    case OP_CLASS:    case OP_ANYBYTE:
760    case OP_NCLASS:  #ifdef SUPPORT_UTF
761    return cc + 1 + 32 / sizeof(pcre_uchar);    if (common->utf) return NULL;
762    #endif
763      return cc + 1;
764    
765  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
766    case OP_XCLASS:    case OP_XCLASS:
767    return cc + GET(cc, 1);    return cc + GET(cc, 1);
768  #endif  #endif
769    
   case OP_RECURSE:  
   case OP_ASSERT:  
   case OP_ASSERT_NOT:  
   case OP_ASSERTBACK:  
   case OP_ASSERTBACK_NOT:  
   case OP_REVERSE:  
   case OP_ONCE:  
   case OP_ONCE_NC:  
   case OP_BRA:  
   case OP_BRAPOS:  
   case OP_COND:  
   case OP_SBRA:  
   case OP_SBRAPOS:  
   case OP_SCOND:  
   case OP_ALT:  
   case OP_KET:  
   case OP_KETRMAX:  
   case OP_KETRMIN:  
   case OP_KETRPOS:  
   return cc + 1 + LINK_SIZE;  
   
   case OP_CBRA:  
   case OP_CBRAPOS:  
   case OP_SCBRA:  
   case OP_SCBRAPOS:  
   return cc + 1 + LINK_SIZE + IMM2_SIZE;  
   
770    case OP_MARK:    case OP_MARK:
771      case OP_PRUNE_ARG:
772      case OP_SKIP_ARG:
773      case OP_THEN_ARG:
774    return cc + 1 + 2 + cc[1];    return cc + 1 + 2 + cc[1];
775    
776    default:    default:
777      /* All opcodes are supported now! */
778      SLJIT_ASSERT_STOP();
779    return NULL;    return NULL;
780    }    }
781  }  }
782    
783  static int get_localspace(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend)  static BOOL check_opcode_types(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend)
784  {  {
785  int localspace = 0;  int count;
786  pcre_uchar *alternative;  pcre_uchar *slot;
787    
788  /* 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. */
789  while (cc < ccend)  while (cc < ccend)
790    {    {
# Line 677  while (cc < ccend) Line 792  while (cc < ccend)
792      {      {
793      case OP_SET_SOM:      case OP_SET_SOM:
794      common->has_set_som = TRUE;      common->has_set_som = TRUE;
795        common->might_be_empty = TRUE;
796      cc += 1;      cc += 1;
797      break;      break;
798    
799      case OP_ASSERT:      case OP_REF:
800      case OP_ASSERT_NOT:      case OP_REFI:
801      case OP_ASSERTBACK:      common->optimized_cbracket[GET2(cc, 1)] = 0;
802      case OP_ASSERTBACK_NOT:      cc += 1 + IMM2_SIZE;
803      case OP_ONCE:      break;
804      case OP_ONCE_NC:  
805      case OP_BRAPOS:      case OP_BRA:
806        case OP_CBRA:
807      case OP_SBRA:      case OP_SBRA:
808      case OP_SBRAPOS:      case OP_SCBRA:
809      case OP_SCOND:      count = no_alternatives(cc);
810      localspace += sizeof(sljit_w);      if (count > 4)
811      cc += 1 + LINK_SIZE;        common->read_only_data_size += count * sizeof(sljit_uw);
812        cc += 1 + LINK_SIZE + (*cc == OP_CBRA || *cc == OP_SCBRA ? IMM2_SIZE : 0);
813      break;      break;
814    
815      case OP_CBRAPOS:      case OP_CBRAPOS:
816      case OP_SCBRAPOS:      case OP_SCBRAPOS:
817      localspace += sizeof(sljit_w);      common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] = 0;
818      cc += 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
819      break;      break;
820    
821      case OP_COND:      case OP_COND:
822      /* Might be a hidden SCOND. */      case OP_SCOND:
823      alternative = cc + GET(cc, 1);      /* Only AUTO_CALLOUT can insert this opcode. We do
824      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)         not intend to support this case. */
825        localspace += sizeof(sljit_w);      if (cc[1 + LINK_SIZE] == OP_CALLOUT)
826          return FALSE;
827      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
828      break;      break;
829    
830        case OP_CREF:
831        common->optimized_cbracket[GET2(cc, 1)] = 0;
832        cc += 1 + IMM2_SIZE;
833        break;
834    
835        case OP_DNREF:
836        case OP_DNREFI:
837        case OP_DNCREF:
838        count = GET2(cc, 1 + IMM2_SIZE);
839        slot = common->name_table + GET2(cc, 1) * common->name_entry_size;
840        while (count-- > 0)
841          {
842          common->optimized_cbracket[GET2(slot, 0)] = 0;
843          slot += common->name_entry_size;
844          }
845        cc += 1 + 2 * IMM2_SIZE;
846        break;
847    
848      case OP_RECURSE:      case OP_RECURSE:
849      /* Set its value only once. */      /* Set its value only once. */
850      if (common->recursive_head == 0)      if (common->recursive_head_ptr == 0)
851        {        {
852        common->recursive_head = common->ovector_start;        common->recursive_head_ptr = common->ovector_start;
853        common->ovector_start += sizeof(sljit_w);        common->ovector_start += sizeof(sljit_sw);
854        }        }
855      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
856      break;      break;
857    
858        case OP_CALLOUT:
859        if (common->capture_last_ptr == 0)
860          {
861          common->capture_last_ptr = common->ovector_start;
862          common->ovector_start += sizeof(sljit_sw);
863          }
864        cc += 2 + 2 * LINK_SIZE;
865        break;
866    
867        case OP_THEN_ARG:
868        common->has_then = TRUE;
869        common->control_head_ptr = 1;
870        /* Fall through. */
871    
872        case OP_PRUNE_ARG:
873        common->needs_start_ptr = TRUE;
874        /* Fall through. */
875    
876      case OP_MARK:      case OP_MARK:
877      if (common->mark_ptr == 0)      if (common->mark_ptr == 0)
878        {        {
879        common->mark_ptr = common->ovector_start;        common->mark_ptr = common->ovector_start;
880        common->ovector_start += sizeof(sljit_w);        common->ovector_start += sizeof(sljit_sw);
881        }        }
882      cc += 1 + 2 + cc[1];      cc += 1 + 2 + cc[1];
883      break;      break;
884    
885        case OP_THEN:
886        common->has_then = TRUE;
887        common->control_head_ptr = 1;
888        /* Fall through. */
889    
890        case OP_PRUNE:
891        case OP_SKIP:
892        common->needs_start_ptr = TRUE;
893        cc += 1;
894        break;
895    
896        case OP_SKIP_ARG:
897        common->control_head_ptr = 1;
898        common->has_skip_arg = TRUE;
899        cc += 1 + 2 + cc[1];
900        break;
901    
902      default:      default:
903      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
904      if (cc == NULL)      if (cc == NULL)
905        return -1;        return FALSE;
906      break;      break;
907      }      }
908    }    }
909  return localspace;  return TRUE;
910  }  }
911    
912  static void set_localptrs(compiler_common *common, int localptr, pcre_uchar *ccend)  static int get_class_iterator_size(pcre_uchar *cc)
913  {  {
914  pcre_uchar *cc = common->start;  switch(*cc)
 pcre_uchar *alternative;  
 while (cc < ccend)  
915    {    {
916    switch(*cc)    case OP_CRSTAR:
917      {    case OP_CRPLUS:
918      case OP_ASSERT:    return 2;
     case OP_ASSERT_NOT:  
     case OP_ASSERTBACK:  
     case OP_ASSERTBACK_NOT:  
     case OP_ONCE:  
     case OP_ONCE_NC:  
     case OP_BRAPOS:  
     case OP_SBRA:  
     case OP_SBRAPOS:  
     case OP_SCOND:  
     common->localptrs[cc - common->start] = localptr;  
     localptr += sizeof(sljit_w);  
     cc += 1 + LINK_SIZE;  
     break;  
919    
920      case OP_CBRAPOS:    case OP_CRMINSTAR:
921      case OP_SCBRAPOS:    case OP_CRMINPLUS:
922      common->localptrs[cc - common->start] = localptr;    case OP_CRQUERY:
923      localptr += sizeof(sljit_w);    case OP_CRMINQUERY:
924      cc += 1 + LINK_SIZE + IMM2_SIZE;    return 1;
     break;  
925    
926      case OP_COND:    case OP_CRRANGE:
927      /* Might be a hidden SCOND. */    case OP_CRMINRANGE:
928      alternative = cc + GET(cc, 1);    if (GET2(cc, 1) == GET2(cc, 1 + IMM2_SIZE))
929      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)      return 0;
930        {    return 2;
931        common->localptrs[cc - common->start] = localptr;  
932        localptr += sizeof(sljit_w);    default:
933        }    return 0;
934      }
935    }
936    
937    static BOOL detect_repeat(compiler_common *common, pcre_uchar *begin)
938    {
939    pcre_uchar *end = bracketend(begin);
940    pcre_uchar *next;
941    pcre_uchar *next_end;
942    pcre_uchar *max_end;
943    pcre_uchar type;
944    sljit_sw length = end - begin;
945    int min, max, i;
946    
947    /* Detect fixed iterations first. */
948    if (end[-(1 + LINK_SIZE)] != OP_KET)
949      return FALSE;
950    
951    /* Already detected repeat. */
952    if (common->private_data_ptrs[end - common->start - LINK_SIZE] != 0)
953      return TRUE;
954    
955    next = end;
956    min = 1;
957    while (1)
958      {
959      if (*next != *begin)
960        break;
961      next_end = bracketend(next);
962      if (next_end - next != length || memcmp(begin, next, IN_UCHARS(length)) != 0)
963        break;
964      next = next_end;
965      min++;
966      }
967    
968    if (min == 2)
969      return FALSE;
970    
971    max = 0;
972    max_end = next;
973    if (*next == OP_BRAZERO || *next == OP_BRAMINZERO)
974      {
975      type = *next;
976      while (1)
977        {
978        if (next[0] != type || next[1] != OP_BRA || next[2 + LINK_SIZE] != *begin)
979          break;
980        next_end = bracketend(next + 2 + LINK_SIZE);
981        if (next_end - next != (length + 2 + LINK_SIZE) || memcmp(begin, next + 2 + LINK_SIZE, IN_UCHARS(length)) != 0)
982          break;
983        next = next_end;
984        max++;
985        }
986    
987      if (next[0] == type && next[1] == *begin && max >= 1)
988        {
989        next_end = bracketend(next + 1);
990        if (next_end - next == (length + 1) && memcmp(begin, next + 1, IN_UCHARS(length)) == 0)
991          {
992          for (i = 0; i < max; i++, next_end += 1 + LINK_SIZE)
993            if (*next_end != OP_KET)
994              break;
995    
996          if (i == max)
997            {
998            common->private_data_ptrs[max_end - common->start - LINK_SIZE] = next_end - max_end;
999            common->private_data_ptrs[max_end - common->start - LINK_SIZE + 1] = (type == OP_BRAZERO) ? OP_UPTO : OP_MINUPTO;
1000            /* +2 the original and the last. */
1001            common->private_data_ptrs[max_end - common->start - LINK_SIZE + 2] = max + 2;
1002            if (min == 1)
1003              return TRUE;
1004            min--;
1005            max_end -= (1 + LINK_SIZE) + GET(max_end, -LINK_SIZE);
1006            }
1007          }
1008        }
1009      }
1010    
1011    if (min >= 3)
1012      {
1013      common->private_data_ptrs[end - common->start - LINK_SIZE] = max_end - end;
1014      common->private_data_ptrs[end - common->start - LINK_SIZE + 1] = OP_EXACT;
1015      common->private_data_ptrs[end - common->start - LINK_SIZE + 2] = min;
1016      return TRUE;
1017      }
1018    
1019    return FALSE;
1020    }
1021    
1022    #define CASE_ITERATOR_PRIVATE_DATA_1 \
1023        case OP_MINSTAR: \
1024        case OP_MINPLUS: \
1025        case OP_QUERY: \
1026        case OP_MINQUERY: \
1027        case OP_MINSTARI: \
1028        case OP_MINPLUSI: \
1029        case OP_QUERYI: \
1030        case OP_MINQUERYI: \
1031        case OP_NOTMINSTAR: \
1032        case OP_NOTMINPLUS: \
1033        case OP_NOTQUERY: \
1034        case OP_NOTMINQUERY: \
1035        case OP_NOTMINSTARI: \
1036        case OP_NOTMINPLUSI: \
1037        case OP_NOTQUERYI: \
1038        case OP_NOTMINQUERYI:
1039    
1040    #define CASE_ITERATOR_PRIVATE_DATA_2A \
1041        case OP_STAR: \
1042        case OP_PLUS: \
1043        case OP_STARI: \
1044        case OP_PLUSI: \
1045        case OP_NOTSTAR: \
1046        case OP_NOTPLUS: \
1047        case OP_NOTSTARI: \
1048        case OP_NOTPLUSI:
1049    
1050    #define CASE_ITERATOR_PRIVATE_DATA_2B \
1051        case OP_UPTO: \
1052        case OP_MINUPTO: \
1053        case OP_UPTOI: \
1054        case OP_MINUPTOI: \
1055        case OP_NOTUPTO: \
1056        case OP_NOTMINUPTO: \
1057        case OP_NOTUPTOI: \
1058        case OP_NOTMINUPTOI:
1059    
1060    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_1 \
1061        case OP_TYPEMINSTAR: \
1062        case OP_TYPEMINPLUS: \
1063        case OP_TYPEQUERY: \
1064        case OP_TYPEMINQUERY:
1065    
1066    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2A \
1067        case OP_TYPESTAR: \
1068        case OP_TYPEPLUS:
1069    
1070    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2B \
1071        case OP_TYPEUPTO: \
1072        case OP_TYPEMINUPTO:
1073    
1074    static void set_private_data_ptrs(compiler_common *common, int *private_data_start, pcre_uchar *ccend)
1075    {
1076    pcre_uchar *cc = common->start;
1077    pcre_uchar *alternative;
1078    pcre_uchar *end = NULL;
1079    int private_data_ptr = *private_data_start;
1080    int space, size, bracketlen;
1081    
1082    while (cc < ccend)
1083      {
1084      space = 0;
1085      size = 0;
1086      bracketlen = 0;
1087      if (private_data_ptr > SLJIT_MAX_LOCAL_SIZE)
1088        return;
1089    
1090      if (*cc == OP_ONCE || *cc == OP_ONCE_NC || *cc == OP_BRA || *cc == OP_CBRA || *cc == OP_COND)
1091        if (detect_repeat(common, cc))
1092          {
1093          /* These brackets are converted to repeats, so no global
1094          based single character repeat is allowed. */
1095          if (cc >= end)
1096            end = bracketend(cc);
1097          }
1098    
1099      switch(*cc)
1100        {
1101        case OP_KET:
1102        if (common->private_data_ptrs[cc + 1 - common->start] != 0)
1103          {
1104          common->private_data_ptrs[cc - common->start] = private_data_ptr;
1105          private_data_ptr += sizeof(sljit_sw);
1106          cc += common->private_data_ptrs[cc + 1 - common->start];
1107          }
1108      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1109      break;      break;
1110    
1111        case OP_ASSERT:
1112        case OP_ASSERT_NOT:
1113        case OP_ASSERTBACK:
1114        case OP_ASSERTBACK_NOT:
1115        case OP_ONCE:
1116        case OP_ONCE_NC:
1117        case OP_BRAPOS:
1118        case OP_SBRA:
1119        case OP_SBRAPOS:
1120        case OP_SCOND:
1121        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1122        private_data_ptr += sizeof(sljit_sw);
1123        bracketlen = 1 + LINK_SIZE;
1124        break;
1125    
1126        case OP_CBRAPOS:
1127        case OP_SCBRAPOS:
1128        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1129        private_data_ptr += sizeof(sljit_sw);
1130        bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
1131        break;
1132    
1133        case OP_COND:
1134        /* Might be a hidden SCOND. */
1135        alternative = cc + GET(cc, 1);
1136        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1137          {
1138          common->private_data_ptrs[cc - common->start] = private_data_ptr;
1139          private_data_ptr += sizeof(sljit_sw);
1140          }
1141        bracketlen = 1 + LINK_SIZE;
1142        break;
1143    
1144        case OP_BRA:
1145        bracketlen = 1 + LINK_SIZE;
1146        break;
1147    
1148        case OP_CBRA:
1149        case OP_SCBRA:
1150        bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
1151        break;
1152    
1153        CASE_ITERATOR_PRIVATE_DATA_1
1154        space = 1;
1155        size = -2;
1156        break;
1157    
1158        CASE_ITERATOR_PRIVATE_DATA_2A
1159        space = 2;
1160        size = -2;
1161        break;
1162    
1163        CASE_ITERATOR_PRIVATE_DATA_2B
1164        space = 2;
1165        size = -(2 + IMM2_SIZE);
1166        break;
1167    
1168        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1169        space = 1;
1170        size = 1;
1171        break;
1172    
1173        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1174        if (cc[1] != OP_ANYNL && cc[1] != OP_EXTUNI)
1175          space = 2;
1176        size = 1;
1177        break;
1178    
1179        CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1180        if (cc[1 + IMM2_SIZE] != OP_ANYNL && cc[1 + IMM2_SIZE] != OP_EXTUNI)
1181          space = 2;
1182        size = 1 + IMM2_SIZE;
1183        break;
1184    
1185        case OP_CLASS:
1186        case OP_NCLASS:
1187        size += 1 + 32 / sizeof(pcre_uchar);
1188        space = get_class_iterator_size(cc + size);
1189        break;
1190    
1191    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1192        case OP_XCLASS:
1193        size = GET(cc, 1);
1194        space = get_class_iterator_size(cc + size);
1195        break;
1196    #endif
1197    
1198      default:      default:
1199      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
1200      SLJIT_ASSERT(cc != NULL);      SLJIT_ASSERT(cc != NULL);
1201      break;      break;
1202      }      }
1203    
1204      /* Character iterators, which are not inside a repeated bracket,
1205         gets a private slot instead of allocating it on the stack. */
1206      if (space > 0 && cc >= end)
1207        {
1208        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1209        private_data_ptr += sizeof(sljit_sw) * space;
1210        }
1211    
1212      if (size != 0)
1213        {
1214        if (size < 0)
1215          {
1216          cc += -size;
1217    #ifdef SUPPORT_UTF
1218          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1219    #endif
1220          }
1221        else
1222          cc += size;
1223        }
1224    
1225      if (bracketlen > 0)
1226        {
1227        if (cc >= end)
1228          {
1229          end = bracketend(cc);
1230          if (end[-1 - LINK_SIZE] == OP_KET)
1231            end = NULL;
1232          }
1233        cc += bracketlen;
1234        }
1235    }    }
1236    *private_data_start = private_data_ptr;
1237  }  }
1238    
1239  /* Returns with -1 if no need for frame. */  /* Returns with a frame_types (always < 0) if no need for frame. */
1240  static int get_framesize(compiler_common *common, pcre_uchar *cc, BOOL recursive)  static int get_framesize(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, BOOL recursive, BOOL* needs_control_head)
1241  {  {
 pcre_uchar *ccend = bracketend(cc);  
1242  int length = 0;  int length = 0;
1243  BOOL possessive = FALSE;  int possessive = 0;
1244    BOOL stack_restore = FALSE;
1245  BOOL setsom_found = recursive;  BOOL setsom_found = recursive;
1246  BOOL setmark_found = recursive;  BOOL setmark_found = recursive;
1247    /* The last capture is a local variable even for recursions. */
1248    BOOL capture_last_found = FALSE;
1249    
1250    #if defined DEBUG_FORCE_CONTROL_HEAD && DEBUG_FORCE_CONTROL_HEAD
1251    SLJIT_ASSERT(common->control_head_ptr != 0);
1252    *needs_control_head = TRUE;
1253    #else
1254    *needs_control_head = FALSE;
1255    #endif
1256    
1257  if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))  if (ccend == NULL)
1258    {    {
1259    length = 3;    ccend = bracketend(cc) - (1 + LINK_SIZE);
1260    possessive = TRUE;    if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))
1261        {
1262        possessive = length = (common->capture_last_ptr != 0) ? 5 : 3;
1263        /* This is correct regardless of common->capture_last_ptr. */
1264        capture_last_found = TRUE;
1265        }
1266      cc = next_opcode(common, cc);
1267    }    }
1268    
 cc = next_opcode(common, cc);  
1269  SLJIT_ASSERT(cc != NULL);  SLJIT_ASSERT(cc != NULL);
1270  while (cc < ccend)  while (cc < ccend)
1271    switch(*cc)    switch(*cc)
1272      {      {
1273      case OP_SET_SOM:      case OP_SET_SOM:
1274      SLJIT_ASSERT(common->has_set_som);      SLJIT_ASSERT(common->has_set_som);
1275        stack_restore = TRUE;
1276      if (!setsom_found)      if (!setsom_found)
1277        {        {
1278        length += 2;        length += 2;
# Line 817  while (cc < ccend) Line 1282  while (cc < ccend)
1282      break;      break;
1283    
1284      case OP_MARK:      case OP_MARK:
1285        case OP_PRUNE_ARG:
1286        case OP_THEN_ARG:
1287      SLJIT_ASSERT(common->mark_ptr != 0);      SLJIT_ASSERT(common->mark_ptr != 0);
1288        stack_restore = TRUE;
1289      if (!setmark_found)      if (!setmark_found)
1290        {        {
1291        length += 2;        length += 2;
1292        setmark_found = TRUE;        setmark_found = TRUE;
1293        }        }
1294        if (common->control_head_ptr != 0)
1295          *needs_control_head = TRUE;
1296      cc += 1 + 2 + cc[1];      cc += 1 + 2 + cc[1];
1297      break;      break;
1298    
1299      case OP_RECURSE:      case OP_RECURSE:
1300        stack_restore = TRUE;
1301      if (common->has_set_som && !setsom_found)      if (common->has_set_som && !setsom_found)
1302        {        {
1303        length += 2;        length += 2;
# Line 837  while (cc < ccend) Line 1308  while (cc < ccend)
1308        length += 2;        length += 2;
1309        setmark_found = TRUE;        setmark_found = TRUE;
1310        }        }
1311        if (common->capture_last_ptr != 0 && !capture_last_found)
1312          {
1313          length += 2;
1314          capture_last_found = TRUE;
1315          }
1316      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1317      break;      break;
1318    
# Line 844  while (cc < ccend) Line 1320  while (cc < ccend)
1320      case OP_CBRAPOS:      case OP_CBRAPOS:
1321      case OP_SCBRA:      case OP_SCBRA:
1322      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1323        stack_restore = TRUE;
1324        if (common->capture_last_ptr != 0 && !capture_last_found)
1325          {
1326          length += 2;
1327          capture_last_found = TRUE;
1328          }
1329      length += 3;      length += 3;
1330      cc += 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1331      break;      break;
1332    
1333      default:      default:
1334        stack_restore = TRUE;
1335        /* Fall through. */
1336    
1337        case OP_NOT_WORD_BOUNDARY:
1338        case OP_WORD_BOUNDARY:
1339        case OP_NOT_DIGIT:
1340        case OP_DIGIT:
1341        case OP_NOT_WHITESPACE:
1342        case OP_WHITESPACE:
1343        case OP_NOT_WORDCHAR:
1344        case OP_WORDCHAR:
1345        case OP_ANY:
1346        case OP_ALLANY:
1347        case OP_ANYBYTE:
1348        case OP_NOTPROP:
1349        case OP_PROP:
1350        case OP_ANYNL:
1351        case OP_NOT_HSPACE:
1352        case OP_HSPACE:
1353        case OP_NOT_VSPACE:
1354        case OP_VSPACE:
1355        case OP_EXTUNI:
1356        case OP_EODN:
1357        case OP_EOD:
1358        case OP_CIRC:
1359        case OP_CIRCM:
1360        case OP_DOLL:
1361        case OP_DOLLM:
1362        case OP_CHAR:
1363        case OP_CHARI:
1364        case OP_NOT:
1365        case OP_NOTI:
1366    
1367        case OP_EXACT:
1368        case OP_POSSTAR:
1369        case OP_POSPLUS:
1370        case OP_POSQUERY:
1371        case OP_POSUPTO:
1372    
1373        case OP_EXACTI:
1374        case OP_POSSTARI:
1375        case OP_POSPLUSI:
1376        case OP_POSQUERYI:
1377        case OP_POSUPTOI:
1378    
1379        case OP_NOTEXACT:
1380        case OP_NOTPOSSTAR:
1381        case OP_NOTPOSPLUS:
1382        case OP_NOTPOSQUERY:
1383        case OP_NOTPOSUPTO:
1384    
1385        case OP_NOTEXACTI:
1386        case OP_NOTPOSSTARI:
1387        case OP_NOTPOSPLUSI:
1388        case OP_NOTPOSQUERYI:
1389        case OP_NOTPOSUPTOI:
1390    
1391        case OP_TYPEEXACT:
1392        case OP_TYPEPOSSTAR:
1393        case OP_TYPEPOSPLUS:
1394        case OP_TYPEPOSQUERY:
1395        case OP_TYPEPOSUPTO:
1396    
1397        case OP_CLASS:
1398        case OP_NCLASS:
1399        case OP_XCLASS:
1400    
1401      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
1402      SLJIT_ASSERT(cc != NULL);      SLJIT_ASSERT(cc != NULL);
1403      break;      break;
1404      }      }
1405    
1406  /* Possessive quantifiers can use a special case. */  /* Possessive quantifiers can use a special case. */
1407  if (SLJIT_UNLIKELY(possessive) && length == 3)  if (SLJIT_UNLIKELY(possessive == length))
1408    return -1;    return stack_restore ? no_frame : no_stack;
1409    
1410  if (length > 0)  if (length > 0)
1411    return length + 1;    return length + 1;
1412  return -1;  return stack_restore ? no_frame : no_stack;
1413  }  }
1414    
1415  static void init_frame(compiler_common *common, pcre_uchar *cc, int stackpos, int stacktop, BOOL recursive)  static void init_frame(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, int stackpos, int stacktop, BOOL recursive)
1416  {  {
1417  DEFINE_COMPILER;  DEFINE_COMPILER;
 pcre_uchar *ccend = bracketend(cc);  
1418  BOOL setsom_found = recursive;  BOOL setsom_found = recursive;
1419  BOOL setmark_found = recursive;  BOOL setmark_found = recursive;
1420    /* The last capture is a local variable even for recursions. */
1421    BOOL capture_last_found = FALSE;
1422  int offset;  int offset;
1423    
1424  /* >= 1 + shortest item size (2) */  /* >= 1 + shortest item size (2) */
# Line 876  SLJIT_UNUSED_ARG(stacktop); Line 1426  SLJIT_UNUSED_ARG(stacktop);
1426  SLJIT_ASSERT(stackpos >= stacktop + 2);  SLJIT_ASSERT(stackpos >= stacktop + 2);
1427    
1428  stackpos = STACK(stackpos);  stackpos = STACK(stackpos);
1429  if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))  if (ccend == NULL)
1430    cc = next_opcode(common, cc);    {
1431      ccend = bracketend(cc) - (1 + LINK_SIZE);
1432      if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))
1433        cc = next_opcode(common, cc);
1434      }
1435    
1436  SLJIT_ASSERT(cc != NULL);  SLJIT_ASSERT(cc != NULL);
1437  while (cc < ccend)  while (cc < ccend)
1438    switch(*cc)    switch(*cc)
# Line 887  while (cc < ccend) Line 1442  while (cc < ccend)
1442      if (!setsom_found)      if (!setsom_found)
1443        {        {
1444        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));
1445        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setstrbegin);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1446        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1447        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1448        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1449        setsom_found = TRUE;        setsom_found = TRUE;
1450        }        }
1451      cc += 1;      cc += 1;
1452      break;      break;
1453    
1454      case OP_MARK:      case OP_MARK:
1455        case OP_PRUNE_ARG:
1456        case OP_THEN_ARG:
1457      SLJIT_ASSERT(common->mark_ptr != 0);      SLJIT_ASSERT(common->mark_ptr != 0);
1458      if (!setmark_found)      if (!setmark_found)
1459        {        {
1460        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);
1461        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setmark);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
1462        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1463        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1464        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1465        setmark_found = TRUE;        setmark_found = TRUE;
1466        }        }
1467      cc += 1 + 2 + cc[1];      cc += 1 + 2 + cc[1];
# Line 914  while (cc < ccend) Line 1471  while (cc < ccend)
1471      if (common->has_set_som && !setsom_found)      if (common->has_set_som && !setsom_found)
1472        {        {
1473        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));
1474        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setstrbegin);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1475        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1476        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1477        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1478        setsom_found = TRUE;        setsom_found = TRUE;
1479        }        }
1480      if (common->mark_ptr != 0 && !setmark_found)      if (common->mark_ptr != 0 && !setmark_found)
1481        {        {
1482        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);
1483        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setmark);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
1484        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1485        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1486        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1487        setmark_found = TRUE;        setmark_found = TRUE;
1488        }        }
1489        if (common->capture_last_ptr != 0 && !capture_last_found)
1490          {
1491          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->capture_last_ptr);
1492          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1493          stackpos += (int)sizeof(sljit_sw);
1494          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1495          stackpos += (int)sizeof(sljit_sw);
1496          capture_last_found = TRUE;
1497          }
1498      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1499      break;      break;
1500    
# Line 936  while (cc < ccend) Line 1502  while (cc < ccend)
1502      case OP_CBRAPOS:      case OP_CBRAPOS:
1503      case OP_SCBRA:      case OP_SCBRA:
1504      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1505        if (common->capture_last_ptr != 0 && !capture_last_found)
1506          {
1507          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->capture_last_ptr);
1508          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1509          stackpos += (int)sizeof(sljit_sw);
1510          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1511          stackpos += (int)sizeof(sljit_sw);
1512          capture_last_found = TRUE;
1513          }
1514      offset = (GET2(cc, 1 + LINK_SIZE)) << 1;      offset = (GET2(cc, 1 + LINK_SIZE)) << 1;
1515      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));
1516      stackpos += (int)sizeof(sljit_w);      stackpos += (int)sizeof(sljit_sw);
1517      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset));      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset));
1518      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset + 1));      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset + 1));
1519      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1520      stackpos += (int)sizeof(sljit_w);      stackpos += (int)sizeof(sljit_sw);
1521      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);
1522      stackpos += (int)sizeof(sljit_w);      stackpos += (int)sizeof(sljit_sw);
1523    
1524      cc += 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1525      break;      break;
# Line 955  while (cc < ccend) Line 1530  while (cc < ccend)
1530      break;      break;
1531      }      }
1532    
1533  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_end);  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, 0);
1534  SLJIT_ASSERT(stackpos == STACK(stacktop));  SLJIT_ASSERT(stackpos == STACK(stacktop));
1535  }  }
1536    
1537  static SLJIT_INLINE int get_localsize(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend)  static SLJIT_INLINE int get_private_data_copy_length(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, BOOL needs_control_head)
1538  {  {
1539  int localsize = 2;  int private_data_length = needs_control_head ? 3 : 2;
1540    int size;
1541  pcre_uchar *alternative;  pcre_uchar *alternative;
1542  /* Calculate the sum of the local variables. */  /* Calculate the sum of the private machine words. */
1543  while (cc < ccend)  while (cc < ccend)
1544    {    {
1545      size = 0;
1546    switch(*cc)    switch(*cc)
1547      {      {
1548        case OP_KET:
1549        if (PRIVATE_DATA(cc) != 0)
1550          private_data_length++;
1551        cc += 1 + LINK_SIZE;
1552        break;
1553    
1554      case OP_ASSERT:      case OP_ASSERT:
1555      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1556      case OP_ASSERTBACK:      case OP_ASSERTBACK:
# Line 978  while (cc < ccend) Line 1561  while (cc < ccend)
1561      case OP_SBRA:      case OP_SBRA:
1562      case OP_SBRAPOS:      case OP_SBRAPOS:
1563      case OP_SCOND:      case OP_SCOND:
1564      localsize++;      private_data_length++;
1565      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1566      break;      break;
1567    
1568      case OP_CBRA:      case OP_CBRA:
1569      case OP_SCBRA:      case OP_SCBRA:
1570      localsize++;      if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1571          private_data_length++;
1572      cc += 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1573      break;      break;
1574    
1575      case OP_CBRAPOS:      case OP_CBRAPOS:
1576      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1577      localsize += 2;      private_data_length += 2;
1578      cc += 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1579      break;      break;
1580    
# Line 998  while (cc < ccend) Line 1582  while (cc < ccend)
1582      /* Might be a hidden SCOND. */      /* Might be a hidden SCOND. */
1583      alternative = cc + GET(cc, 1);      alternative = cc + GET(cc, 1);
1584      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1585        localsize++;        private_data_length++;
1586      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1587      break;      break;
1588    
1589      default:      CASE_ITERATOR_PRIVATE_DATA_1
1590      cc = next_opcode(common, cc);      if (PRIVATE_DATA(cc))
1591      SLJIT_ASSERT(cc != NULL);        private_data_length++;
1592        cc += 2;
1593    #ifdef SUPPORT_UTF
1594        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1595    #endif
1596      break;      break;
     }  
   }  
 SLJIT_ASSERT(cc == ccend);  
 return localsize;  
 }  
   
 static void copy_locals(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend,  
   BOOL save, int stackptr, int stacktop)  
 {  
 DEFINE_COMPILER;  
 int srcw[2];  
 int count;  
 BOOL tmp1next = TRUE;  
 BOOL tmp1empty = TRUE;  
 BOOL tmp2empty = TRUE;  
 pcre_uchar *alternative;  
 enum {  
   start,  
   loop,  
   end  
 } status;  
1597    
1598  status = save ? start : loop;      CASE_ITERATOR_PRIVATE_DATA_2A
1599  stackptr = STACK(stackptr - 2);      if (PRIVATE_DATA(cc))
1600  stacktop = STACK(stacktop - 1);        private_data_length += 2;
1601        cc += 2;
1602    #ifdef SUPPORT_UTF
1603        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1604    #endif
1605        break;
1606    
1607        CASE_ITERATOR_PRIVATE_DATA_2B
1608        if (PRIVATE_DATA(cc))
1609          private_data_length += 2;
1610        cc += 2 + IMM2_SIZE;
1611    #ifdef SUPPORT_UTF
1612        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1613    #endif
1614        break;
1615    
1616        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1617        if (PRIVATE_DATA(cc))
1618          private_data_length++;
1619        cc += 1;
1620        break;
1621    
1622        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1623        if (PRIVATE_DATA(cc))
1624          private_data_length += 2;
1625        cc += 1;
1626        break;
1627    
1628        CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1629        if (PRIVATE_DATA(cc))
1630          private_data_length += 2;
1631        cc += 1 + IMM2_SIZE;
1632        break;
1633    
1634        case OP_CLASS:
1635        case OP_NCLASS:
1636    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1637        case OP_XCLASS:
1638        size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
1639    #else
1640        size = 1 + 32 / (int)sizeof(pcre_uchar);
1641    #endif
1642        if (PRIVATE_DATA(cc))
1643          private_data_length += get_class_iterator_size(cc + size);
1644        cc += size;
1645        break;
1646    
1647        default:
1648        cc = next_opcode(common, cc);
1649        SLJIT_ASSERT(cc != NULL);
1650        break;
1651        }
1652      }
1653    SLJIT_ASSERT(cc == ccend);
1654    return private_data_length;
1655    }
1656    
1657    static void copy_private_data(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend,
1658      BOOL save, int stackptr, int stacktop, BOOL needs_control_head)
1659    {
1660    DEFINE_COMPILER;
1661    int srcw[2];
1662    int count, size;
1663    BOOL tmp1next = TRUE;
1664    BOOL tmp1empty = TRUE;
1665    BOOL tmp2empty = TRUE;
1666    pcre_uchar *alternative;
1667    enum {
1668      start,
1669      loop,
1670      end
1671    } status;
1672    
1673    status = save ? start : loop;
1674    stackptr = STACK(stackptr - 2);
1675    stacktop = STACK(stacktop - 1);
1676    
1677  if (!save)  if (!save)
1678    {    {
1679    stackptr += sizeof(sljit_w);    stackptr += (needs_control_head ? 2 : 1) * sizeof(sljit_sw);
1680    if (stackptr < stacktop)    if (stackptr < stacktop)
1681      {      {
1682      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1683      stackptr += sizeof(sljit_w);      stackptr += sizeof(sljit_sw);
1684      tmp1empty = FALSE;      tmp1empty = FALSE;
1685      }      }
1686    if (stackptr < stacktop)    if (stackptr < stacktop)
1687      {      {
1688      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1689      stackptr += sizeof(sljit_w);      stackptr += sizeof(sljit_sw);
1690      tmp2empty = FALSE;      tmp2empty = FALSE;
1691      }      }
1692    /* The tmp1next must be TRUE in either way. */    /* The tmp1next must be TRUE in either way. */
1693    }    }
1694    
1695  while (status != end)  do
1696    {    {
1697    count = 0;    count = 0;
1698    switch(status)    switch(status)
1699      {      {
1700      case start:      case start:
1701      SLJIT_ASSERT(save && common->recursive_head != 0);      SLJIT_ASSERT(save && common->recursive_head_ptr != 0);
1702      count = 1;      count = 1;
1703      srcw[0] = common->recursive_head;      srcw[0] = common->recursive_head_ptr;
1704        if (needs_control_head)
1705          {
1706          SLJIT_ASSERT(common->control_head_ptr != 0);
1707          count = 2;
1708          srcw[1] = common->control_head_ptr;
1709          }
1710      status = loop;      status = loop;
1711      break;      break;
1712    
# Line 1071  while (status != end) Line 1719  while (status != end)
1719    
1720      switch(*cc)      switch(*cc)
1721        {        {
1722          case OP_KET:
1723          if (PRIVATE_DATA(cc) != 0)
1724            {
1725            count = 1;
1726            srcw[0] = PRIVATE_DATA(cc);
1727            }
1728          cc += 1 + LINK_SIZE;
1729          break;
1730    
1731        case OP_ASSERT:        case OP_ASSERT:
1732        case OP_ASSERT_NOT:        case OP_ASSERT_NOT:
1733        case OP_ASSERTBACK:        case OP_ASSERTBACK:
# Line 1082  while (status != end) Line 1739  while (status != end)
1739        case OP_SBRAPOS:        case OP_SBRAPOS:
1740        case OP_SCOND:        case OP_SCOND:
1741        count = 1;        count = 1;
1742        srcw[0] = PRIV_DATA(cc);        srcw[0] = PRIVATE_DATA(cc);
1743        SLJIT_ASSERT(srcw[0] != 0);        SLJIT_ASSERT(srcw[0] != 0);
1744        cc += 1 + LINK_SIZE;        cc += 1 + LINK_SIZE;
1745        break;        break;
1746    
1747        case OP_CBRA:        case OP_CBRA:
1748        case OP_SCBRA:        case OP_SCBRA:
1749        count = 1;        if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1750        srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));          {
1751            count = 1;
1752            srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1753            }
1754        cc += 1 + LINK_SIZE + IMM2_SIZE;        cc += 1 + LINK_SIZE + IMM2_SIZE;
1755        break;        break;
1756    
1757        case OP_CBRAPOS:        case OP_CBRAPOS:
1758        case OP_SCBRAPOS:        case OP_SCBRAPOS:
1759        count = 2;        count = 2;
1760          srcw[0] = PRIVATE_DATA(cc);
1761        srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));        srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1762        srcw[0] = PRIV_DATA(cc);        SLJIT_ASSERT(srcw[0] != 0 && srcw[1] != 0);
       SLJIT_ASSERT(srcw[0] != 0);  
1763        cc += 1 + LINK_SIZE + IMM2_SIZE;        cc += 1 + LINK_SIZE + IMM2_SIZE;
1764        break;        break;
1765    
# Line 1109  while (status != end) Line 1769  while (status != end)
1769        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1770          {          {
1771          count = 1;          count = 1;
1772          srcw[0] = PRIV_DATA(cc);          srcw[0] = PRIVATE_DATA(cc);
1773          SLJIT_ASSERT(srcw[0] != 0);          SLJIT_ASSERT(srcw[0] != 0);
1774          }          }
1775        cc += 1 + LINK_SIZE;        cc += 1 + LINK_SIZE;
1776        break;        break;
1777    
1778          CASE_ITERATOR_PRIVATE_DATA_1
1779          if (PRIVATE_DATA(cc))
1780            {
1781            count = 1;
1782            srcw[0] = PRIVATE_DATA(cc);
1783            }
1784          cc += 2;
1785    #ifdef SUPPORT_UTF
1786          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1787    #endif
1788          break;
1789    
1790          CASE_ITERATOR_PRIVATE_DATA_2A
1791          if (PRIVATE_DATA(cc))
1792            {
1793            count = 2;
1794            srcw[0] = PRIVATE_DATA(cc);
1795            srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1796            }
1797          cc += 2;
1798    #ifdef SUPPORT_UTF
1799          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1800    #endif
1801          break;
1802    
1803          CASE_ITERATOR_PRIVATE_DATA_2B
1804          if (PRIVATE_DATA(cc))
1805            {
1806            count = 2;
1807            srcw[0] = PRIVATE_DATA(cc);
1808            srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1809            }
1810          cc += 2 + IMM2_SIZE;
1811    #ifdef SUPPORT_UTF
1812          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1813    #endif
1814          break;
1815    
1816          CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1817          if (PRIVATE_DATA(cc))
1818            {
1819            count = 1;
1820            srcw[0] = PRIVATE_DATA(cc);
1821            }
1822          cc += 1;
1823          break;
1824    
1825          CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1826          if (PRIVATE_DATA(cc))
1827            {
1828            count = 2;
1829            srcw[0] = PRIVATE_DATA(cc);
1830            srcw[1] = srcw[0] + sizeof(sljit_sw);
1831            }
1832          cc += 1;
1833          break;
1834    
1835          CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1836          if (PRIVATE_DATA(cc))
1837            {
1838            count = 2;
1839            srcw[0] = PRIVATE_DATA(cc);
1840            srcw[1] = srcw[0] + sizeof(sljit_sw);
1841            }
1842          cc += 1 + IMM2_SIZE;
1843          break;
1844    
1845          case OP_CLASS:
1846          case OP_NCLASS:
1847    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1848          case OP_XCLASS:
1849          size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
1850    #else
1851          size = 1 + 32 / (int)sizeof(pcre_uchar);
1852    #endif
1853          if (PRIVATE_DATA(cc))
1854            switch(get_class_iterator_size(cc + size))
1855              {
1856              case 1:
1857              count = 1;
1858              srcw[0] = PRIVATE_DATA(cc);
1859              break;
1860    
1861              case 2:
1862              count = 2;
1863              srcw[0] = PRIVATE_DATA(cc);
1864              srcw[1] = srcw[0] + sizeof(sljit_sw);
1865              break;
1866    
1867              default:
1868              SLJIT_ASSERT_STOP();
1869              break;
1870              }
1871          cc += size;
1872          break;
1873    
1874        default:        default:
1875        cc = next_opcode(common, cc);        cc = next_opcode(common, cc);
1876        SLJIT_ASSERT(cc != NULL);        SLJIT_ASSERT(cc != NULL);
# Line 1137  while (status != end) Line 1893  while (status != end)
1893          if (!tmp1empty)          if (!tmp1empty)
1894            {            {
1895            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1896            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1897            }            }
1898          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);
1899          tmp1empty = FALSE;          tmp1empty = FALSE;
# Line 1148  while (status != end) Line 1904  while (status != end)
1904          if (!tmp2empty)          if (!tmp2empty)
1905            {            {
1906            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1907            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1908            }            }
1909          OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);          OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);
1910          tmp2empty = FALSE;          tmp2empty = FALSE;
# Line 1165  while (status != end) Line 1921  while (status != end)
1921          if (!tmp1empty)          if (!tmp1empty)
1922            {            {
1923            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1924            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1925            }            }
1926          tmp1next = FALSE;          tmp1next = FALSE;
1927          }          }
# Line 1177  while (status != end) Line 1933  while (status != end)
1933          if (!tmp2empty)          if (!tmp2empty)
1934            {            {
1935            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1936            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1937            }            }
1938          tmp1next = TRUE;          tmp1next = TRUE;
1939          }          }
1940        }        }
1941      }      }
1942    }    }
1943    while (status != end);
1944    
1945  if (save)  if (save)
1946    {    {
# Line 1192  if (save) Line 1949  if (save)
1949      if (!tmp1empty)      if (!tmp1empty)
1950        {        {
1951        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1952        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1953        }        }
1954      if (!tmp2empty)      if (!tmp2empty)
1955        {        {
1956        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1957        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1958        }        }
1959      }      }
1960    else    else
# Line 1205  if (save) Line 1962  if (save)
1962      if (!tmp2empty)      if (!tmp2empty)
1963        {        {
1964        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1965        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1966        }        }
1967      if (!tmp1empty)      if (!tmp1empty)
1968        {        {
1969        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1970        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1971        }        }
1972      }      }
1973    }    }
1974  SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));  SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));
1975  }  }
1976    
1977  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)
1978    {
1979    pcre_uchar *end = bracketend(cc);
1980    BOOL has_alternatives = cc[GET(cc, 1)] == OP_ALT;
1981    
1982    /* Assert captures then. */
1983    if (*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT)
1984      current_offset = NULL;
1985    /* Conditional block does not. */
1986    if (*cc == OP_COND || *cc == OP_SCOND)
1987      has_alternatives = FALSE;
1988    
1989    cc = next_opcode(common, cc);
1990    if (has_alternatives)
1991      current_offset = common->then_offsets + (cc - common->start);
1992    
1993    while (cc < end)
1994      {
1995      if ((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND))
1996        cc = set_then_offsets(common, cc, current_offset);
1997      else
1998        {
1999        if (*cc == OP_ALT && has_alternatives)
2000          current_offset = common->then_offsets + (cc + 1 + LINK_SIZE - common->start);
2001        if (*cc >= OP_THEN && *cc <= OP_THEN_ARG && current_offset != NULL)
2002          *current_offset = 1;
2003        cc = next_opcode(common, cc);
2004        }
2005      }
2006    
2007    return end;
2008    }
2009    
2010    #undef CASE_ITERATOR_PRIVATE_DATA_1
2011    #undef CASE_ITERATOR_PRIVATE_DATA_2A
2012    #undef CASE_ITERATOR_PRIVATE_DATA_2B
2013    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_1
2014    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
2015    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
2016    
2017    static SLJIT_INLINE BOOL is_powerof2(unsigned int value)
2018  {  {
2019  return (value & (value - 1)) == 0;  return (value & (value - 1)) == 0;
2020  }  }
# Line 1227  static SLJIT_INLINE void set_jumps(jump_ Line 2024  static SLJIT_INLINE void set_jumps(jump_
2024  while (list)  while (list)
2025    {    {
2026    /* sljit_set_label is clever enough to do nothing    /* sljit_set_label is clever enough to do nothing
2027    if either the jump or the label is NULL */    if either the jump or the label is NULL. */
2028    sljit_set_label(list->jump, label);    SET_LABEL(list->jump, label);
2029    list = list->next;    list = list->next;
2030    }    }
2031  }  }
# Line 1244  if (list_item) Line 2041  if (list_item)
2041    }    }
2042  }  }
2043    
2044  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)
2045  {  {
2046  DEFINE_COMPILER;  DEFINE_COMPILER;
2047  stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));  stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));
2048    
2049  if (list_item)  if (list_item)
2050    {    {
   list_item->type = type;  
   list_item->data = data;  
2051    list_item->start = start;    list_item->start = start;
2052    list_item->leave = LABEL();    list_item->quit = LABEL();
2053    list_item->next = common->stubs;    list_item->next = common->stubs;
2054    common->stubs = list_item;    common->stubs = list_item;
2055    }    }
# Line 1268  stub_list* list_item = common->stubs; Line 2063  stub_list* list_item = common->stubs;
2063  while (list_item)  while (list_item)
2064    {    {
2065    JUMPHERE(list_item->start);    JUMPHERE(list_item->start);
2066    switch(list_item->type)    add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));
2067      {    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);  
2068    list_item = list_item->next;    list_item = list_item->next;
2069    }    }
2070  common->stubs = NULL;  common->stubs = NULL;
2071  }  }
2072    
2073  static SLJIT_INLINE void decrease_call_count(compiler_common *common)  static void add_label_addr(compiler_common *common)
2074    {
2075    DEFINE_COMPILER;
2076    label_addr_list *label_addr;
2077    
2078    label_addr = sljit_alloc_memory(compiler, sizeof(label_addr_list));
2079    if (label_addr == NULL)
2080      return;
2081    label_addr->label = LABEL();
2082    label_addr->addr = common->read_only_data_ptr;
2083    label_addr->next = common->label_addrs;
2084    common->label_addrs = label_addr;
2085    common->read_only_data_ptr++;
2086    }
2087    
2088    static SLJIT_INLINE void count_match(compiler_common *common)
2089  {  {
2090  DEFINE_COMPILER;  DEFINE_COMPILER;
2091    
2092  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);
2093  add_jump(compiler, &common->calllimit, JUMP(SLJIT_C_ZERO));  add_jump(compiler, &common->calllimit, JUMP(SLJIT_C_ZERO));
2094  }  }
2095    
# Line 1293  static SLJIT_INLINE void allocate_stack( Line 2098  static SLJIT_INLINE void allocate_stack(
2098  /* May destroy all locals and registers except TMP2. */  /* May destroy all locals and registers except TMP2. */
2099  DEFINE_COMPILER;  DEFINE_COMPILER;
2100    
2101  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));
2102  #ifdef DESTROY_REGISTERS  #ifdef DESTROY_REGISTERS
2103  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);
2104  OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);  OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);
# Line 1301  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0); Line 2106  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);
2106  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);
2107  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);
2108  #endif  #endif
2109  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));
2110  }  }
2111    
2112  static SLJIT_INLINE void free_stack(compiler_common *common, int size)  static SLJIT_INLINE void free_stack(compiler_common *common, int size)
2113  {  {
2114  DEFINE_COMPILER;  DEFINE_COMPILER;
2115  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));
2116  }  }
2117    
2118  static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)  static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)
# Line 1315  static SLJIT_INLINE void reset_ovector(c Line 2120  static SLJIT_INLINE void reset_ovector(c
2120  DEFINE_COMPILER;  DEFINE_COMPILER;
2121  struct sljit_label *loop;  struct sljit_label *loop;
2122  int i;  int i;
2123    
2124  /* At this point we can freely use all temporary registers. */  /* At this point we can freely use all temporary registers. */
2125    SLJIT_ASSERT(length > 1);
2126  /* TMP1 returns with begin - 1. */  /* TMP1 returns with begin - 1. */
2127  OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG1, 0, SLJIT_MEM1(SLJIT_SAVED_REG1), SLJIT_OFFSETOF(jit_arguments, begin), SLJIT_IMM, IN_UCHARS(1));  OP2(SLJIT_SUB, SLJIT_SCRATCH_REG1, 0, SLJIT_MEM1(SLJIT_SAVED_REG1), SLJIT_OFFSETOF(jit_arguments, begin), SLJIT_IMM, IN_UCHARS(1));
2128    if (length < 8)
2129      {
2130      for (i = 1; i < length; i++)
2131        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), SLJIT_SCRATCH_REG1, 0);
2132      }
2133    else
2134      {
2135      GET_LOCAL_BASE(SLJIT_SCRATCH_REG2, 0, OVECTOR_START);
2136      OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, length - 1);
2137      loop = LABEL();
2138      OP1(SLJIT_MOVU, SLJIT_MEM1(SLJIT_SCRATCH_REG2), sizeof(sljit_sw), SLJIT_SCRATCH_REG1, 0);
2139      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 1);
2140      JUMPTO(SLJIT_C_NOT_ZERO, loop);
2141      }
2142    }
2143    
2144    static SLJIT_INLINE void do_reset_match(compiler_common *common, int length)
2145    {
2146    DEFINE_COMPILER;
2147    struct sljit_label *loop;
2148    int i;
2149    
2150    SLJIT_ASSERT(length > 1);
2151    /* OVECTOR(1) contains the "string begin - 1" constant. */
2152    if (length > 2)
2153      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));
2154  if (length < 8)  if (length < 8)
2155    {    {
2156    for (i = 0; i < length; i++)    for (i = 2; i < length; i++)
2157      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);
2158    }    }
2159  else  else
2160    {    {
2161    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));
2162    OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, length);    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_IMM, length - 2);
2163    loop = LABEL();    loop = LABEL();
2164    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);
2165    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);
2166    JUMPTO(SLJIT_C_NOT_ZERO, loop);    JUMPTO(SLJIT_C_NOT_ZERO, loop);
2167    }    }
2168    
2169    OP1(SLJIT_MOV, STACK_TOP, 0, ARGUMENTS, 0);
2170    if (common->mark_ptr != 0)
2171      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr, SLJIT_IMM, 0);
2172    if (common->control_head_ptr != 0)
2173      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->control_head_ptr, SLJIT_IMM, 0);
2174    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(jit_arguments, stack));
2175    OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_ptr);
2176    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(struct sljit_stack, base));
2177    }
2178    
2179    static sljit_sw SLJIT_CALL do_search_mark(sljit_sw *current, const pcre_uchar *skip_arg)
2180    {
2181    while (current != NULL)
2182      {
2183      switch (current[-2])
2184        {
2185        case type_then_trap:
2186        break;
2187    
2188        case type_mark:
2189        if (STRCMP_UC_UC(skip_arg, (pcre_uchar *)current[-3]) == 0)
2190          return current[-4];
2191        break;
2192    
2193        default:
2194        SLJIT_ASSERT_STOP();
2195        break;
2196        }
2197      current = (sljit_sw*)current[-1];
2198      }
2199    return -1;
2200  }  }
2201    
2202  static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)  static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)
2203  {  {
2204  DEFINE_COMPILER;  DEFINE_COMPILER;
2205  struct sljit_label *loop;  struct sljit_label *loop;
2206  struct sljit_jump *earlyexit;  struct sljit_jump *early_quit;
2207    
2208  /* At this point we can freely use all registers. */  /* At this point we can freely use all registers. */
2209  OP1(SLJIT_MOV, SLJIT_SAVED_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));  OP1(SLJIT_MOV, SLJIT_SAVED_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));
2210  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);
2211    
2212  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, ARGUMENTS, 0);  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG1, 0, ARGUMENTS, 0);
2213  if (common->mark_ptr != 0)  if (common->mark_ptr != 0)
2214    OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);
2215  OP1(SLJIT_MOV_SI, SLJIT_TEMPORARY_REG2, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, offsetcount));  OP1(SLJIT_MOV_SI, SLJIT_SCRATCH_REG2, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, offset_count));
2216  if (common->mark_ptr != 0)  if (common->mark_ptr != 0)
2217    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, mark_ptr), SLJIT_TEMPORARY_REG3, 0);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, mark_ptr), SLJIT_SCRATCH_REG3, 0);
2218  OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));  OP2(SLJIT_SUB, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));
2219  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, begin));  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG1, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, begin));
2220  OP2(SLJIT_ADD, SLJIT_SAVED_REG1, 0, SLJIT_LOCALS_REG, 0, SLJIT_IMM, OVECTOR_START);  GET_LOCAL_BASE(SLJIT_SAVED_REG1, 0, OVECTOR_START);
2221  /* Unlikely, but possible */  /* Unlikely, but possible */
2222  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);
2223  loop = LABEL();  loop = LABEL();
2224  OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, SLJIT_MEM1(SLJIT_SAVED_REG1), 0, SLJIT_TEMPORARY_REG1, 0);  OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, SLJIT_MEM1(SLJIT_SAVED_REG1), 0, SLJIT_SCRATCH_REG1, 0);
2225  OP2(SLJIT_ADD, SLJIT_SAVED_REG1, 0, SLJIT_SAVED_REG1, 0, SLJIT_IMM, sizeof(sljit_w));  OP2(SLJIT_ADD, SLJIT_SAVED_REG1, 0, SLJIT_SAVED_REG1, 0, SLJIT_IMM, sizeof(sljit_sw));
2226  /* Copy the integer value to the output buffer */  /* Copy the integer value to the output buffer */
2227  #ifdef COMPILE_PCRE16  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2228  OP2(SLJIT_ASHR, SLJIT_SAVED_REG2, 0, SLJIT_SAVED_REG2, 0, SLJIT_IMM, 1);  OP2(SLJIT_ASHR, SLJIT_SAVED_REG2, 0, SLJIT_SAVED_REG2, 0, SLJIT_IMM, UCHAR_SHIFT);
2229  #endif  #endif
2230  OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_TEMPORARY_REG3), sizeof(int), SLJIT_SAVED_REG2, 0);  OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG3), sizeof(int), SLJIT_SAVED_REG2, 0);
2231  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_TEMPORARY_REG2, 0, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, 1);  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_SCRATCH_REG2, 0, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, 1);
2232  JUMPTO(SLJIT_C_NOT_ZERO, loop);  JUMPTO(SLJIT_C_NOT_ZERO, loop);
2233  JUMPHERE(earlyexit);  JUMPHERE(early_quit);
2234    
2235  /* Calculate the return value, which is the maximum ovector value. */  /* Calculate the return value, which is the maximum ovector value. */
2236  if (topbracket > 1)  if (topbracket > 1)
2237    {    {
2238    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));
2239    OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, topbracket + 1);    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, topbracket + 1);
2240    
2241    /* OVECTOR(0) is never equal to SLJIT_SAVED_REG3. */    /* OVECTOR(0) is never equal to SLJIT_SAVED_REG3. */
2242    loop = LABEL();    loop = LABEL();
2243    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)));
2244    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);
2245    CMPTO(SLJIT_C_EQUAL, SLJIT_TEMPORARY_REG3, 0, SLJIT_SAVED_REG3, 0, loop);    CMPTO(SLJIT_C_EQUAL, SLJIT_SCRATCH_REG3, 0, SLJIT_SAVED_REG3, 0, loop);
2246    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_TEMPORARY_REG2, 0);    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_SCRATCH_REG2, 0);
2247    }    }
2248  else  else
2249    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, 1);    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, 1);
2250  }  }
2251    
2252  static SLJIT_INLINE void return_with_partial_match(compiler_common *common, struct sljit_label *leave)  static SLJIT_INLINE void return_with_partial_match(compiler_common *common, struct sljit_label *quit)
2253  {  {
2254  DEFINE_COMPILER;  DEFINE_COMPILER;
2255    struct sljit_jump *jump;
2256    
2257  SLJIT_COMPILE_ASSERT(STR_END == SLJIT_SAVED_REG2, str_end_must_be_saved_reg2);  SLJIT_COMPILE_ASSERT(STR_END == SLJIT_SAVED_REG2, str_end_must_be_saved_reg2);
2258  SLJIT_ASSERT(common->start_used_ptr != 0 && (common->mode == JIT_PARTIAL_SOFT_COMPILE ? common->hit_start != 0 : common->hit_start == 0));  SLJIT_ASSERT(common->start_used_ptr != 0 && common->start_ptr != 0
2259      && (common->mode == JIT_PARTIAL_SOFT_COMPILE ? common->hit_start != 0 : common->hit_start == 0));
2260    
2261  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG2, 0, ARGUMENTS, 0);  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, ARGUMENTS, 0);
2262  OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, PCRE_ERROR_PARTIAL);  OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, PCRE_ERROR_PARTIAL);
2263  OP1(SLJIT_MOV_SI, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), SLJIT_OFFSETOF(jit_arguments, offsetcount));  OP1(SLJIT_MOV_SI, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, real_offset_count));
2264  CMPTO(SLJIT_C_LESS, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, 2, leave);  CMPTO(SLJIT_C_SIG_LESS, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 2, quit);
2265    
2266  /* Store match begin and end. */  /* Store match begin and end. */
2267  OP1(SLJIT_MOV, SLJIT_SAVED_REG1, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), SLJIT_OFFSETOF(jit_arguments, begin));  OP1(SLJIT_MOV, SLJIT_SAVED_REG1, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, begin));
2268  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG2, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), SLJIT_OFFSETOF(jit_arguments, offsets));  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, offsets));
2269  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mode == JIT_PARTIAL_HARD_COMPILE ? common->start_used_ptr : common->hit_start);  
2270    jump = CMP(SLJIT_C_SIG_LESS, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 3);
2271    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);
2272    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2273    OP2(SLJIT_ASHR, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, UCHAR_SHIFT);
2274    #endif
2275    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG2), 2 * sizeof(int), SLJIT_SCRATCH_REG3, 0);
2276    JUMPHERE(jump);
2277    
2278    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);
2279  OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, STR_END, 0, SLJIT_SAVED_REG1, 0);  OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, STR_END, 0, SLJIT_SAVED_REG1, 0);
2280  #ifdef COMPILE_PCRE16  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2281  OP2(SLJIT_ASHR, SLJIT_SAVED_REG2, 0, SLJIT_SAVED_REG2, 0, SLJIT_IMM, 1);  OP2(SLJIT_ASHR, SLJIT_SAVED_REG2, 0, SLJIT_SAVED_REG2, 0, SLJIT_IMM, UCHAR_SHIFT);
2282  #endif  #endif
2283  OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), sizeof(int), SLJIT_SAVED_REG2, 0);  OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG2), sizeof(int), SLJIT_SAVED_REG2, 0);
2284    
2285  OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG3, 0, SLJIT_TEMPORARY_REG3, 0, SLJIT_SAVED_REG1, 0);  OP2(SLJIT_SUB, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_SAVED_REG1, 0);
2286  #ifdef COMPILE_PCRE16  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2287  OP2(SLJIT_ASHR, SLJIT_TEMPORARY_REG3, 0, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, 1);  OP2(SLJIT_ASHR, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, UCHAR_SHIFT);
2288  #endif  #endif
2289  OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), 0, SLJIT_TEMPORARY_REG3, 0);  OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG2), 0, SLJIT_SCRATCH_REG3, 0);
2290    
2291  JUMPTO(SLJIT_JUMP, leave);  JUMPTO(SLJIT_JUMP, quit);
2292  }  }
2293    
2294  static SLJIT_INLINE void check_start_used_ptr(compiler_common *common)  static SLJIT_INLINE void check_start_used_ptr(compiler_common *common)
# Line 1523  if (c <= 127 && bit == 0x20) Line 2399  if (c <= 127 && bit == 0x20)
2399    return (0 << 8) | 0x20;    return (0 << 8) | 0x20;
2400    
2401  /* Since c != oc, they must have at least 1 bit difference. */  /* Since c != oc, they must have at least 1 bit difference. */
2402  if (!ispowerof2(bit))  if (!is_powerof2(bit))
2403    return 0;    return 0;
2404    
2405  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
2406    
2407  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
2408  if (common->utf && c > 127)  if (common->utf && c > 127)
# Line 1542  if (common->utf && c > 127) Line 2418  if (common->utf && c > 127)
2418  #endif /* SUPPORT_UTF */  #endif /* SUPPORT_UTF */
2419  return (0 << 8) | bit;  return (0 << 8) | bit;
2420    
2421  #else /* COMPILE_PCRE8 */  #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2422    
 #ifdef COMPILE_PCRE16  
2423  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
2424  if (common->utf && c > 65535)  if (common->utf && c > 65535)
2425    {    {
# Line 1555  if (common->utf && c > 65535) Line 2430  if (common->utf && c > 65535)
2430    }    }
2431  #endif /* SUPPORT_UTF */  #endif /* SUPPORT_UTF */
2432  return (bit < 256) ? ((0 << 8) | bit) : ((1 << 8) | (bit >> 8));  return (bit < 256) ? ((0 << 8) | bit) : ((1 << 8) | (bit >> 8));
 #endif /* COMPILE_PCRE16 */  
2433    
2434  #endif /* COMPILE_PCRE8 */  #endif /* COMPILE_PCRE[8|16|32] */
2435  }  }
2436    
2437  static void check_partial(compiler_common *common, BOOL force)  static void check_partial(compiler_common *common, BOOL force)
2438  {  {
2439  /* Checks whether a partial matching is occured. Does not modify registers. */  /* Checks whether a partial matching is occurred. Does not modify registers. */
2440  DEFINE_COMPILER;  DEFINE_COMPILER;
2441  struct sljit_jump *jump = NULL;  struct sljit_jump *jump = NULL;
2442    
# Line 1577  else if (common->mode == JIT_PARTIAL_SOF Line 2451  else if (common->mode == JIT_PARTIAL_SOF
2451    jump = CMP(SLJIT_C_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, SLJIT_IMM, -1);    jump = CMP(SLJIT_C_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, SLJIT_IMM, -1);
2452    
2453  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2454    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, -1);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2455  else  else
2456    {    {
2457    if (common->partialmatchlabel != NULL)    if (common->partialmatchlabel != NULL)
# Line 1590  if (jump != NULL) Line 2464  if (jump != NULL)
2464    JUMPHERE(jump);    JUMPHERE(jump);
2465  }  }
2466    
2467  static struct sljit_jump *check_str_end(compiler_common *common)  static void check_str_end(compiler_common *common, jump_list **end_reached)
2468  {  {
2469  /* Does not affect registers. Usually used in a tight spot. */  /* Does not affect registers. Usually used in a tight spot. */
2470  DEFINE_COMPILER;  DEFINE_COMPILER;
2471  struct sljit_jump *jump;  struct sljit_jump *jump;
 struct sljit_jump *nohit;  
 struct sljit_jump *return_value;  
2472    
2473  if (common->mode == JIT_COMPILE)  if (common->mode == JIT_COMPILE)
2474    return CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);    {
2475      add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2476      return;
2477      }
2478    
2479  jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);  jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);
2480  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2481    {    {
2482    nohit = CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);    add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));
2483    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, -1);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2484    JUMPHERE(nohit);    add_jump(compiler, end_reached, JUMP(SLJIT_JUMP));
   return_value = JUMP(SLJIT_JUMP);  
2485    }    }
2486  else  else
2487    {    {
2488    return_value = CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);    add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));
2489    if (common->partialmatchlabel != NULL)    if (common->partialmatchlabel != NULL)
2490      JUMPTO(SLJIT_JUMP, common->partialmatchlabel);      JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2491    else    else
2492      add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));      add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2493    }    }
2494  JUMPHERE(jump);  JUMPHERE(jump);
 return return_value;  
2495  }  }
2496    
2497  static void fallback_at_str_end(compiler_common *common, jump_list **fallbacks)  static void detect_partial_match(compiler_common *common, jump_list **backtracks)
2498  {  {
2499  DEFINE_COMPILER;  DEFINE_COMPILER;
2500  struct sljit_jump *jump;  struct sljit_jump *jump;
2501    
2502  if (common->mode == JIT_COMPILE)  if (common->mode == JIT_COMPILE)
2503    {    {
2504    add_jump(compiler, fallbacks, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));    add_jump(compiler, backtracks, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2505    return;    return;
2506    }    }
2507    
2508  /* Partial matching mode. */  /* Partial matching mode. */
2509  jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);  jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);
2510  add_jump(compiler, fallbacks, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));  add_jump(compiler, backtracks, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));
2511  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2512    {    {
2513    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, -1);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2514    add_jump(compiler, fallbacks, JUMP(SLJIT_JUMP));    add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));
2515    }    }
2516  else  else
2517    {    {
# Line 1650  else Line 2523  else
2523  JUMPHERE(jump);  JUMPHERE(jump);
2524  }  }
2525    
2526  static void read_char(compiler_common *common)  static void peek_char(compiler_common *common, pcre_uint32 max)
2527  {  {
2528  /* Reads the character into TMP1, updates STR_PTR.  /* Reads the character into TMP1, keeps STR_PTR.
2529  Does not check STR_END. TMP2 Destroyed. */  Does not check STR_END. TMP2 Destroyed. */
2530  DEFINE_COMPILER;  DEFINE_COMPILER;
2531  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2532  struct sljit_jump *jump;  struct sljit_jump *jump;
2533  #endif  #endif
2534    
2535    SLJIT_UNUSED_ARG(max);
2536    
2537  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2538  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2539  if (common->utf)  if (common->utf)
2540    {    {
2541  #ifdef COMPILE_PCRE8    if (max < 128) return;
2542    
2543    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2544  #else    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
 #ifdef COMPILE_PCRE16  
   jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);  
 #endif  
 #endif /* COMPILE_PCRE8 */  
2545    add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2546      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2547      JUMPHERE(jump);
2548      }
2549    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2550    
2551    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2552    if (common->utf)
2553      {
2554      if (max < 0xd800) return;
2555    
2556      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2557      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2558      /* TMP2 contains the high surrogate. */
2559      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2560      OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2561      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2562      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2563      OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2564    JUMPHERE(jump);    JUMPHERE(jump);
2565    }    }
2566  #endif  #endif
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  
2567  }  }
2568    
2569  static void peek_char(compiler_common *common)  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2570    
2571    static BOOL is_char7_bitset(const pcre_uint8 *bitset, BOOL nclass)
2572  {  {
2573  /* Reads the character into TMP1, keeps STR_PTR.  /* Tells whether the character codes below 128 are enough
2574  Does not check STR_END. TMP2 Destroyed. */  to determine a match. */
2575    const pcre_uint8 value = nclass ? 0xff : 0;
2576    const pcre_uint8* end = bitset + 32;
2577    
2578    bitset += 16;
2579    do
2580      {
2581      if (*bitset++ != value)
2582        return FALSE;
2583      }
2584    while (bitset < end);
2585    return TRUE;
2586    }
2587    
2588    static void read_char7_type(compiler_common *common, BOOL full_read)
2589    {
2590    /* Reads the precise character type of a character into TMP1, if the character
2591    is less than 128. Otherwise it returns with zero. Does not check STR_END. The
2592    full_read argument tells whether characters above max are accepted or not. */
2593  DEFINE_COMPILER;  DEFINE_COMPILER;
 #ifdef SUPPORT_UTF  
2594  struct sljit_jump *jump;  struct sljit_jump *jump;
 #endif  
2595    
2596  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  SLJIT_ASSERT(common->utf);
2597  #ifdef SUPPORT_UTF  
2598  if (common->utf)  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2599    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2600    
2601    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2602    
2603    if (full_read)
2604    {    {
2605  #ifdef COMPILE_PCRE8    jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2606    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);    OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2607  #else    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
 #ifdef COMPILE_PCRE16  
   jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);  
 #endif  
 #endif /* COMPILE_PCRE8 */  
   add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));  
   OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);  
2608    JUMPHERE(jump);    JUMPHERE(jump);
2609    }    }
 #endif  
2610  }  }
2611    
2612  static void read_char8_type(compiler_common *common)  #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2613    
2614    static void read_char_range(compiler_common *common, pcre_uint32 min, pcre_uint32 max, BOOL update_str_ptr)
2615  {  {
2616  /* Reads the character type into TMP1, updates STR_PTR. Does not check STR_END. */  /* Reads the precise value of a character into TMP1, if the character is
2617    between min and max (c >= min && c <= max). Otherwise it returns with a value
2618    outside the range. Does not check STR_END. */
2619  DEFINE_COMPILER;  DEFINE_COMPILER;
2620  #if defined SUPPORT_UTF || defined COMPILE_PCRE16  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2621  struct sljit_jump *jump;  struct sljit_jump *jump;
2622  #endif  #endif
2623    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2624    struct sljit_jump *jump2;
2625    #endif
2626    
2627  #ifdef SUPPORT_UTF  SLJIT_UNUSED_ARG(update_str_ptr);
2628    SLJIT_UNUSED_ARG(min);
2629    SLJIT_UNUSED_ARG(max);
2630    SLJIT_ASSERT(min <= max);
2631    
2632    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2633    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2634    
2635    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2636  if (common->utf)  if (common->utf)
2637    {    {
2638    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);    if (max < 128 && !update_str_ptr) return;
2639    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  
2640  #ifdef COMPILE_PCRE8    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2641    /* This can be an extra read in some situations, but hopefully    if (min >= 0x10000)
2642    it is needed in most cases. */      {
2643    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xf0);
2644    jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);      if (update_str_ptr)
2645    add_jump(compiler, &common->utfreadtype8, JUMP(SLJIT_FAST_CALL));        OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2646    JUMPHERE(jump);      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2647  #else      jump2 = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0x7);
2648  #ifdef COMPILE_PCRE16      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2649    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2650    jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);      OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2651    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);      OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2652    JUMPHERE(jump);      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2653    /* Skip low surrogate if necessary. */      OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2654    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0xfc00);      OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2655    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0xd800);      OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2656    COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_EQUAL);      if (!update_str_ptr)
2657    OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 1);        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2658    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2659        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2660        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2661        JUMPHERE(jump2);
2662        if (update_str_ptr)
2663          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2664        }
2665      else if (min >= 0x800 && max <= 0xffff)
2666        {
2667        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xe0);
2668        if (update_str_ptr)
2669          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2670        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2671        jump2 = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xf);
2672        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2673        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2674        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2675        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2676        if (!update_str_ptr)
2677          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2678        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2679        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2680        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2681        JUMPHERE(jump2);
2682        if (update_str_ptr)
2683          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2684        }
2685      else if (max >= 0x800)
2686        add_jump(compiler, (max < 0x10000) ? &common->utfreadchar16 : &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2687      else if (max < 128)
2688        {
2689        OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2690        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2691        }
2692      else
2693        {
2694        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2695        if (!update_str_ptr)
2696          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2697        else
2698          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2699        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2700        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2701        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2702        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2703        if (update_str_ptr)
2704          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2705        }
2706      JUMPHERE(jump);
2707      }
2708  #endif  #endif
2709  #endif /* COMPILE_PCRE8 */  
2710    return;  #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2711    if (common->utf)
2712      {
2713      if (max >= 0x10000)
2714        {
2715        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2716        jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2717        /* TMP2 contains the high surrogate. */
2718        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2719        OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2720        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2721        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2722        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2723        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2724        JUMPHERE(jump);
2725        return;
2726        }
2727    
2728      if (max < 0xd800 && !update_str_ptr) return;
2729    
2730      /* Skip low surrogate if necessary. */
2731      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2732      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2733      if (update_str_ptr)
2734        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2735      if (max >= 0xd800)
2736        OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0x10000);
2737      JUMPHERE(jump);
2738    }    }
2739  #endif  #endif
2740    }
2741    
2742    static SLJIT_INLINE void read_char(compiler_common *common)
2743    {
2744    read_char_range(common, 0, READ_CHAR_MAX, TRUE);
2745    }
2746    
2747    static void read_char8_type(compiler_common *common, BOOL update_str_ptr)
2748    {
2749    /* Reads the character type into TMP1, updates STR_PTR. Does not check STR_END. */
2750    DEFINE_COMPILER;
2751    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2752    struct sljit_jump *jump;
2753    #endif
2754    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2755    struct sljit_jump *jump2;
2756    #endif
2757    
2758    SLJIT_UNUSED_ARG(update_str_ptr);
2759    
2760  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2761  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2762  #ifdef COMPILE_PCRE16  
2763    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2764    if (common->utf)
2765      {
2766      /* This can be an extra read in some situations, but hopefully
2767      it is needed in most cases. */
2768      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2769      jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2770      if (!update_str_ptr)
2771        {
2772        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2773        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2774        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2775        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2776        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2777        OP2(SLJIT_OR, TMP2, 0, TMP2, 0, TMP1, 0);
2778        OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2779        jump2 = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);
2780        OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2781        JUMPHERE(jump2);
2782        }
2783      else
2784        add_jump(compiler, &common->utfreadtype8, JUMP(SLJIT_FAST_CALL));
2785      JUMPHERE(jump);
2786      return;
2787      }
2788    #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2789    
2790    #if !defined COMPILE_PCRE8
2791  /* The ctypes array contains only 256 values. */  /* The ctypes array contains only 256 values. */
2792  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2793  jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);  jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);
2794  #endif  #endif
2795  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2796  #ifdef COMPILE_PCRE16  #if !defined COMPILE_PCRE8
2797  JUMPHERE(jump);  JUMPHERE(jump);
2798  #endif  #endif
2799    
2800    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2801    if (common->utf && update_str_ptr)
2802      {
2803      /* Skip low surrogate if necessary. */
2804      OP2(SLJIT_SUB, TMP2, 0, TMP2, 0, SLJIT_IMM, 0xd800);
2805      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2806      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2807      JUMPHERE(jump);
2808      }
2809    #endif /* SUPPORT_UTF && COMPILE_PCRE16 */
2810  }  }
2811    
2812  static void skip_char_back(compiler_common *common)  static void skip_char_back(compiler_common *common)
2813  {  {
2814  /* Goes one character back. Affects STR_PTR and TMP1. Does not check begin. */  /* Goes one character back. Affects STR_PTR and TMP1. Does not check begin. */
2815  DEFINE_COMPILER;  DEFINE_COMPILER;
2816  #if defined SUPPORT_UTF && defined COMPILE_PCRE8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2817    #if defined COMPILE_PCRE8
2818  struct sljit_label *label;  struct sljit_label *label;
2819    
2820  if (common->utf)  if (common->utf)
# Line 1770  if (common->utf) Line 2826  if (common->utf)
2826    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, 0x80, label);    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, 0x80, label);
2827    return;    return;
2828    }    }
2829  #endif  #elif defined COMPILE_PCRE16
 #if defined SUPPORT_UTF && defined COMPILE_PCRE16  
2830  if (common->utf)  if (common->utf)
2831    {    {
2832    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
# Line 1779  if (common->utf) Line 2834  if (common->utf)
2834    /* Skip low surrogate if necessary. */    /* Skip low surrogate if necessary. */
2835    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
2836    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xdc00);    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xdc00);
2837    COND_VALUE(SLJIT_MOV, TMP1, 0, SLJIT_C_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
2838    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
2839    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2840    return;    return;
2841    }    }
2842  #endif  #endif /* COMPILE_PCRE[8|16] */
2843    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2844  OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2845  }  }
2846    
2847  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)
2848  {  {
2849  /* 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. */
2850  DEFINE_COMPILER;  DEFINE_COMPILER;
2851    struct sljit_jump *jump;
2852    
2853  if (nltype == NLTYPE_ANY)  if (nltype == NLTYPE_ANY)
2854    {    {
2855    add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));
2856    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));
2857    }    }
2858  else if (nltype == NLTYPE_ANYCRLF)  else if (nltype == NLTYPE_ANYCRLF)
2859    {    {
2860    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_CR);    if (jumpifmatch)
2861    COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_EQUAL);      {
2862    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));
2863    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));
2864    add_jump(compiler, fallbacks, JUMP(jumpiftrue ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));      }
2865      else
2866        {
2867        jump = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR);
2868        add_jump(compiler, backtracks, CMP(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL));
2869        JUMPHERE(jump);
2870        }
2871    }    }
2872  else  else
2873    {    {
2874    SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline < 256);    SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline < 256);
2875    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));
2876    }    }
2877  }  }
2878    
2879  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
2880    
2881  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
2882  static void do_utfreadchar(compiler_common *common)  static void do_utfreadchar(compiler_common *common)
2883  {  {
2884  /* Fast decoding a UTF-8 character. TMP1 contains the first byte  /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2885  of the character (>= 0xc0). Return char value in TMP1, length - 1 in TMP2. */  of the character (>= 0xc0). Return char value in TMP1, length in TMP2. */
2886  DEFINE_COMPILER;  DEFINE_COMPILER;
2887  struct sljit_jump *jump;  struct sljit_jump *jump;
2888    
2889  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2890    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2891    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2892    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2893    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2894    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2895    
2896  /* Searching for the first zero. */  /* Searching for the first zero. */
2897  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x20);  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
2898  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_C_NOT_ZERO);
2899  /* Two byte sequence. */  /* Two byte sequence. */
 OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  
2900  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2901  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x1f);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(2));
2902    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2903    
2904    JUMPHERE(jump);
2905    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2906    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
2907  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2908  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2909  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, IN_UCHARS(1));  
 sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  
 JUMPHERE(jump);  
2910    
2911  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x10);  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2912  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_C_NOT_ZERO);
2913  /* Three byte sequence. */  /* Three byte sequence. */
 OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  
 OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x0f);  
 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(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));  
2914  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2915  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(3));
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
 OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(2));  
2916  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
 JUMPHERE(jump);  
2917    
2918  /* Four byte sequence. */  /* Four byte sequence. */
2919  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  JUMPHERE(jump);
2920  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x07);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2921  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 18);  OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2922    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2923    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2924  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);  
2925  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2926  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(4));
2927    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2928    }
2929    
2930    static void do_utfreadchar16(compiler_common *common)
2931    {
2932    /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2933    of the character (>= 0xc0). Return value in TMP1. */
2934    DEFINE_COMPILER;
2935    struct sljit_jump *jump;
2936    
2937    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2938    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2939    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2940    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2941  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);  
2942  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2943  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(3));  
2944  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));  /* Searching for the first zero. */
2945    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
2946    jump = JUMP(SLJIT_C_NOT_ZERO);
2947    /* Two byte sequence. */
2948    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2949    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2950    
2951    JUMPHERE(jump);
2952    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x400);
2953    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_NOT_ZERO);
2954    /* This code runs only in 8 bit mode. No need to shift the value. */
2955    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2956    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2957    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
2958    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2959  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2960  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2961  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(3));  /* Three byte sequence. */
2962    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2963  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2964  }  }
2965    
# Line 1882  DEFINE_COMPILER; Line 2971  DEFINE_COMPILER;
2971  struct sljit_jump *jump;  struct sljit_jump *jump;
2972  struct sljit_jump *compare;  struct sljit_jump *compare;
2973    
2974  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2975    
2976  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);
2977  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_C_NOT_ZERO);
# Line 1890  jump = JUMP(SLJIT_C_NOT_ZERO); Line 2979  jump = JUMP(SLJIT_C_NOT_ZERO);
2979  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2980  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2981  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);
2982    /* The upper 5 bits are known at this point. */
2983    compare = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0x3);
2984  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2985  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2986  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);  
2987  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2988  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2989    
2990  JUMPHERE(compare);  JUMPHERE(compare);
2991  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2992  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
 JUMPHERE(jump);  
2993    
2994  /* We only have types for characters less than 256. */  /* We only have types for characters less than 256. */
 OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), (sljit_w)PRIV(utf8_table4) - 0xc0);  
 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);  
 }  
   
 #else /* COMPILE_PCRE8 */  
   
 #ifdef COMPILE_PCRE16  
 static void do_utfreadchar(compiler_common *common)  
 {  
 /* Fast decoding a UTF-16 character. TMP1 contains the first 16 bit char  
 of the character (>= 0xd800). Return char value in TMP1, length - 1 in TMP2. */  
 DEFINE_COMPILER;  
 struct sljit_jump *jump;  
   
 sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  
 jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xdc00);  
 /* Do nothing, only return. */  
 sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  
   
2995  JUMPHERE(jump);  JUMPHERE(jump);
2996  /* Combine two 16 bit characters. */  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2997  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2998  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
 OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);  
 OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 10);  
 OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3ff);  
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
 OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(1));  
 OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x10000);  
2999  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3000  }  }
 #endif /* COMPILE_PCRE16 */  
3001    
3002  #endif /* COMPILE_PCRE8 */  #endif /* COMPILE_PCRE8 */
3003    
# Line 1956  DEFINE_COMPILER; Line 3017  DEFINE_COMPILER;
3017    
3018  SLJIT_ASSERT(UCD_BLOCK_SIZE == 128 && sizeof(ucd_record) == 8);  SLJIT_ASSERT(UCD_BLOCK_SIZE == 128 && sizeof(ucd_record) == 8);
3019    
3020  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
3021  OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);  OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
3022  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_w)PRIV(ucd_stage1));  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(ucd_stage1));
3023  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);
3024  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
3025  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, TMP2, 0);
3026  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, (sljit_w)PRIV(ucd_stage2));  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_stage2));
3027  OP1(SLJIT_MOV_UH, TMP2, 0, SLJIT_MEM2(TMP2, TMP1), 1);  OP1(SLJIT_MOV_UH, TMP2, 0, SLJIT_MEM2(TMP2, TMP1), 1);
3028  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, (sljit_w)PRIV(ucd_records) + SLJIT_OFFSETOF(ucd_record, chartype));  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_records) + SLJIT_OFFSETOF(ucd_record, chartype));
3029  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM2(TMP1, TMP2), 3);  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM2(TMP1, TMP2), 3);
3030  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3031  }  }
# Line 1978  struct sljit_label *newlinelabel = NULL; Line 3039  struct sljit_label *newlinelabel = NULL;
3039  struct sljit_jump *start;  struct sljit_jump *start;
3040  struct sljit_jump *end = NULL;  struct sljit_jump *end = NULL;
3041  struct sljit_jump *nl = NULL;  struct sljit_jump *nl = NULL;
3042  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3043  struct sljit_jump *singlechar;  struct sljit_jump *singlechar;
3044  #endif  #endif
3045  jump_list *newline = NULL;  jump_list *newline = NULL;
# Line 1993  if (firstline) Line 3054  if (firstline)
3054    {    {
3055    /* Search for the end of the first line. */    /* Search for the end of the first line. */
3056    SLJIT_ASSERT(common->first_line_end != 0);    SLJIT_ASSERT(common->first_line_end != 0);
3057    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, STR_PTR, 0);    OP1(SLJIT_MOV, TMP3, 0, STR_PTR, 0);
   OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_END, 0);  
3058    
3059    if (common->nltype == NLTYPE_FIXED && common->newline > 255)    if (common->nltype == NLTYPE_FIXED && common->newline > 255)
3060      {      {
# Line 2005  if (firstline) Line 3065  if (firstline)
3065      OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));      OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
3066      CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, mainloop);      CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, mainloop);
3067      CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, mainloop);      CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, mainloop);
3068        JUMPHERE(end);
3069      OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));      OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3070      }      }
3071    else    else
# Line 2013  if (firstline) Line 3074  if (firstline)
3074      mainloop = LABEL();      mainloop = LABEL();
3075      /* Continual stores does not cause data dependency. */      /* Continual stores does not cause data dependency. */
3076      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);
3077      read_char(common);      read_char_range(common, common->nlmin, common->nlmax, TRUE);
3078      check_newlinechar(common, common->nltype, &newline, TRUE);      check_newlinechar(common, common->nltype, &newline, TRUE);
3079      CMPTO(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0, mainloop);      CMPTO(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0, mainloop);
3080        JUMPHERE(end);
3081      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);
3082      set_jumps(newline, LABEL());      set_jumps(newline, LABEL());
3083      }      }
3084    
3085    JUMPHERE(end);    OP1(SLJIT_MOV, STR_PTR, 0, TMP3, 0);
   OP1(SLJIT_MOV, STR_PTR, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0);  
3086    }    }
3087    
3088  start = JUMP(SLJIT_JUMP);  start = JUMP(SLJIT_JUMP);
# Line 2033  if (newlinecheck) Line 3094  if (newlinecheck)
3094    end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);    end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3095    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3096    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, common->newline & 0xff);    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, common->newline & 0xff);
3097    COND_VALUE(SLJIT_MOV, TMP1, 0, SLJIT_C_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
3098  #ifdef COMPILE_PCRE16  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
3099    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, UCHAR_SHIFT);
3100  #endif  #endif
3101    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3102    nl = JUMP(SLJIT_JUMP);    nl = JUMP(SLJIT_JUMP);
# Line 2056  if (newlinecheck) Line 3117  if (newlinecheck)
3117    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, newlinelabel);    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, newlinelabel);
3118    
3119  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3120  #if defined SUPPORT_UTF && defined COMPILE_PCRE8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3121    #if defined COMPILE_PCRE8
3122  if (common->utf)  if (common->utf)
3123    {    {
3124    singlechar = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);    singlechar = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
3125    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_w)PRIV(utf8_table4) - 0xc0);    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
3126    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3127    JUMPHERE(singlechar);    JUMPHERE(singlechar);
3128    }    }
3129    #elif defined COMPILE_PCRE16
3130    if (common->utf)
3131      {
3132      singlechar = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);
3133      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
3134      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);
3135      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
3136      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
3137      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3138      JUMPHERE(singlechar);
3139      }
3140    #endif /* COMPILE_PCRE[8|16] */
3141    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
3142    JUMPHERE(start);
3143    
3144    if (newlinecheck)
3145      {
3146      JUMPHERE(end);
3147      JUMPHERE(nl);
3148      }
3149    
3150    return mainloop;
3151    }
3152    
3153    #define MAX_N_CHARS 16
3154    #define MAX_N_BYTES 8
3155    
3156    static SLJIT_INLINE void add_prefix_byte(pcre_uint8 byte, pcre_uint8 *bytes)
3157    {
3158    pcre_uint8 len = bytes[0];
3159    int i;
3160    
3161    if (len == 255)
3162      return;
3163    
3164    if (len == 0)
3165      {
3166      bytes[0] = 1;
3167      bytes[1] = byte;
3168      return;
3169      }
3170    
3171    for (i = len; i > 0; i--)
3172      if (bytes[i] == byte)
3173        return;
3174    
3175    if (len >= MAX_N_BYTES - 1)
3176      {
3177      bytes[0] = 255;
3178      return;
3179      }
3180    
3181    len++;
3182    bytes[len] = byte;
3183    bytes[0] = len;
3184    }
3185    
3186    static int scan_prefix(compiler_common *common, pcre_uchar *cc, pcre_uint32 *chars, pcre_uint8 *bytes, int max_chars)
3187    {
3188    /* Recursive function, which scans prefix literals. */
3189    BOOL last, any, caseless;
3190    int len, repeat, len_save, consumed = 0;
3191    pcre_uint32 chr, mask;
3192    pcre_uchar *alternative, *cc_save, *oc;
3193    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3194    pcre_uchar othercase[8];
3195    #elif defined SUPPORT_UTF && defined COMPILE_PCRE16
3196    pcre_uchar othercase[2];
3197    #else
3198    pcre_uchar othercase[1];
3199    #endif
3200    
3201    repeat = 1;
3202    while (TRUE)
3203      {
3204      last = TRUE;
3205      any = FALSE;
3206      caseless = FALSE;
3207      switch (*cc)
3208        {
3209        case OP_CHARI:
3210        caseless = TRUE;
3211        case OP_CHAR:
3212        last = FALSE;
3213        cc++;
3214        break;
3215    
3216        case OP_SOD:
3217        case OP_SOM:
3218        case OP_SET_SOM:
3219        case OP_NOT_WORD_BOUNDARY:
3220        case OP_WORD_BOUNDARY:
3221        case OP_EODN:
3222        case OP_EOD:
3223        case OP_CIRC:
3224        case OP_CIRCM:
3225        case OP_DOLL:
3226        case OP_DOLLM:
3227        /* Zero width assertions. */
3228        cc++;
3229        continue;
3230    
3231        case OP_ASSERT:
3232        case OP_ASSERT_NOT:
3233        case OP_ASSERTBACK:
3234        case OP_ASSERTBACK_NOT:
3235        cc = bracketend(cc);
3236        continue;
3237    
3238        case OP_PLUSI:
3239        case OP_MINPLUSI:
3240        case OP_POSPLUSI:
3241        caseless = TRUE;
3242        case OP_PLUS:
3243        case OP_MINPLUS:
3244        case OP_POSPLUS:
3245        cc++;
3246        break;
3247    
3248        case OP_EXACTI:
3249        caseless = TRUE;
3250        case OP_EXACT:
3251        repeat = GET2(cc, 1);
3252        last = FALSE;
3253        cc += 1 + IMM2_SIZE;
3254        break;
3255    
3256        case OP_QUERYI:
3257        case OP_MINQUERYI:
3258        case OP_POSQUERYI:
3259        caseless = TRUE;
3260        case OP_QUERY:
3261        case OP_MINQUERY:
3262        case OP_POSQUERY:
3263        len = 1;
3264        cc++;
3265    #ifdef SUPPORT_UTF
3266        if (common->utf && HAS_EXTRALEN(*cc)) len += GET_EXTRALEN(*cc);
3267    #endif
3268        max_chars = scan_prefix(common, cc + len, chars, bytes, max_chars);
3269        if (max_chars == 0)
3270          return consumed;
3271        last = FALSE;
3272        break;
3273    
3274        case OP_KET:
3275        cc += 1 + LINK_SIZE;
3276        continue;
3277    
3278        case OP_ALT:
3279        cc += GET(cc, 1);
3280        continue;
3281    
3282        case OP_ONCE:
3283        case OP_ONCE_NC:
3284        case OP_BRA:
3285        case OP_BRAPOS:
3286        case OP_CBRA:
3287        case OP_CBRAPOS:
3288        alternative = cc + GET(cc, 1);
3289        while (*alternative == OP_ALT)
3290          {
3291          max_chars = scan_prefix(common, alternative + 1 + LINK_SIZE, chars, bytes, max_chars);
3292          if (max_chars == 0)
3293            return consumed;
3294          alternative += GET(alternative, 1);
3295          }
3296    
3297        if (*cc == OP_CBRA || *cc == OP_CBRAPOS)
3298          cc += IMM2_SIZE;
3299        cc += 1 + LINK_SIZE;
3300        continue;
3301    
3302        case OP_CLASS:
3303    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3304        if (common->utf && !is_char7_bitset((const pcre_uint8 *)(cc + 1), FALSE)) return consumed;
3305    #endif
3306        any = TRUE;
3307        cc += 1 + 32 / sizeof(pcre_uchar);
3308        break;
3309    
3310        case OP_NCLASS:
3311    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3312        if (common->utf) return consumed;
3313    #endif
3314        any = TRUE;
3315        cc += 1 + 32 / sizeof(pcre_uchar);
3316        break;
3317    
3318    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3319        case OP_XCLASS:
3320    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3321        if (common->utf) return consumed;
3322    #endif
3323        any = TRUE;
3324        cc += GET(cc, 1);
3325        break;
3326    #endif
3327    
3328        case OP_DIGIT:
3329    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3330        if (common->utf && !is_char7_bitset((const pcre_uint8 *)common->ctypes - cbit_length + cbit_digit, FALSE))
3331          return consumed;
3332    #endif
3333        any = TRUE;
3334        cc++;
3335        break;
3336    
3337        case OP_WHITESPACE:
3338    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3339        if (common->utf && !is_char7_bitset((const pcre_uint8 *)common->ctypes - cbit_length + cbit_space, FALSE))
3340          return consumed;
3341    #endif
3342        any = TRUE;
3343        cc++;
3344        break;
3345    
3346        case OP_WORDCHAR:
3347    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3348        if (common->utf && !is_char7_bitset((const pcre_uint8 *)common->ctypes - cbit_length + cbit_word, FALSE))
3349          return consumed;
3350    #endif
3351        any = TRUE;
3352        cc++;
3353        break;
3354    
3355        case OP_NOT:
3356        case OP_NOTI:
3357        cc++;
3358        /* Fall through. */
3359        case OP_NOT_DIGIT:
3360        case OP_NOT_WHITESPACE:
3361        case OP_NOT_WORDCHAR:
3362        case OP_ANY:
3363        case OP_ALLANY:
3364    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3365        if (common->utf) return consumed;
3366    #endif
3367        any = TRUE;
3368        cc++;
3369        break;
3370    
3371    #ifdef SUPPORT_UCP
3372        case OP_NOTPROP:
3373        case OP_PROP:
3374    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3375        if (common->utf) return consumed;
3376    #endif
3377        any = TRUE;
3378        cc += 1 + 2;
3379        break;
3380    #endif
3381    
3382        case OP_TYPEEXACT:
3383        repeat = GET2(cc, 1);
3384        cc += 1 + IMM2_SIZE;
3385        continue;
3386    
3387        case OP_NOTEXACT:
3388        case OP_NOTEXACTI:
3389    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3390        if (common->utf) return consumed;
3391    #endif
3392        any = TRUE;
3393        repeat = GET2(cc, 1);
3394        cc += 1 + IMM2_SIZE + 1;
3395        break;
3396    
3397        default:
3398        return consumed;
3399        }
3400    
3401      if (any)
3402        {
3403    #if defined COMPILE_PCRE8
3404        mask = 0xff;
3405    #elif defined COMPILE_PCRE16
3406        mask = 0xffff;
3407    #elif defined COMPILE_PCRE32
3408        mask = 0xffffffff;
3409    #else
3410        SLJIT_ASSERT_STOP();
3411    #endif
3412    
3413        do
3414          {
3415          chars[0] = mask;
3416          chars[1] = mask;
3417          bytes[0] = 255;
3418    
3419          consumed++;
3420          if (--max_chars == 0)
3421            return consumed;
3422          chars += 2;
3423          bytes += MAX_N_BYTES;
3424          }
3425        while (--repeat > 0);
3426    
3427        repeat = 1;
3428        continue;
3429        }
3430    
3431      len = 1;
3432    #ifdef SUPPORT_UTF
3433      if (common->utf && HAS_EXTRALEN(*cc)) len += GET_EXTRALEN(*cc);
3434    #endif
3435    
3436      if (caseless && char_has_othercase(common, cc))
3437        {
3438    #ifdef SUPPORT_UTF
3439        if (common->utf)
3440          {
3441          GETCHAR(chr, cc);
3442          if ((int)PRIV(ord2utf)(char_othercase(common, chr), othercase) != len)
3443            return consumed;
3444          }
3445        else
3446    #endif
3447          {
3448          chr = *cc;
3449          othercase[0] = TABLE_GET(chr, common->fcc, chr);
3450          }
3451        }
3452      else
3453        caseless = FALSE;
3454    
3455      len_save = len;
3456      cc_save = cc;
3457      while (TRUE)
3458        {
3459        oc = othercase;
3460        do
3461          {
3462          chr = *cc;
3463    #ifdef COMPILE_PCRE32
3464          if (SLJIT_UNLIKELY(chr == NOTACHAR))
3465            return consumed;
3466    #endif
3467          add_prefix_byte((pcre_uint8)chr, bytes);
3468    
3469          mask = 0;
3470          if (caseless)
3471            {
3472            add_prefix_byte((pcre_uint8)*oc, bytes);
3473            mask = *cc ^ *oc;
3474            chr |= mask;
3475            }
3476    
3477    #ifdef COMPILE_PCRE32
3478          if (chars[0] == NOTACHAR && chars[1] == 0)
3479    #else
3480          if (chars[0] == NOTACHAR)
3481    #endif
3482            {
3483            chars[0] = chr;
3484            chars[1] = mask;
3485            }
3486          else
3487            {
3488            mask |= chars[0] ^ chr;
3489            chr |= mask;
3490            chars[0] = chr;
3491            chars[1] |= mask;
3492            }
3493    
3494          len--;
3495          consumed++;
3496          if (--max_chars == 0)
3497            return consumed;
3498          chars += 2;
3499          bytes += MAX_N_BYTES;
3500          cc++;
3501          oc++;
3502          }
3503        while (len > 0);
3504    
3505        if (--repeat == 0)
3506          break;
3507    
3508        len = len_save;
3509        cc = cc_save;
3510        }
3511    
3512      repeat = 1;
3513      if (last)
3514        return consumed;
3515      }
3516    }
3517    
3518    static SLJIT_INLINE BOOL fast_forward_first_n_chars(compiler_common *common, BOOL firstline)
3519    {
3520    DEFINE_COMPILER;
3521    struct sljit_label *start;
3522    struct sljit_jump *quit;
3523    pcre_uint32 chars[MAX_N_CHARS * 2];
3524    pcre_uint8 bytes[MAX_N_CHARS * MAX_N_BYTES];
3525    pcre_uint8 ones[MAX_N_CHARS];
3526    int offsets[3];
3527    pcre_uint32 mask;
3528    pcre_uint8 *byte_set, *byte_set_end;
3529    int i, max, from;
3530    int range_right = -1, range_len = 3 - 1;
3531    sljit_ub *update_table = NULL;
3532    BOOL in_range;
3533    
3534    /* This is even TRUE, if both are NULL. */
3535    SLJIT_ASSERT(common->read_only_data_ptr == common->read_only_data);
3536    
3537    for (i = 0; i < MAX_N_CHARS; i++)
3538      {
3539      chars[i << 1] = NOTACHAR;
3540      chars[(i << 1) + 1] = 0;
3541      bytes[i * MAX_N_BYTES] = 0;
3542      }
3543    
3544    max = scan_prefix(common, common->start, chars, bytes, MAX_N_CHARS);
3545    
3546    if (max <= 1)
3547      return FALSE;
3548    
3549    for (i = 0; i < max; i++)
3550      {
3551      mask = chars[(i << 1) + 1];
3552      ones[i] = ones_in_half_byte[mask & 0xf];
3553      mask >>= 4;
3554      while (mask != 0)
3555        {
3556        ones[i] += ones_in_half_byte[mask & 0xf];
3557        mask >>= 4;
3558        }
3559      }
3560    
3561    in_range = FALSE;
3562    from = 0;   /* Prevent compiler "uninitialized" warning */
3563    for (i = 0; i <= max; i++)
3564      {
3565      if (in_range && (i - from) > range_len && (bytes[(i - 1) * MAX_N_BYTES] <= 4))
3566        {
3567        range_len = i - from;
3568        range_right = i - 1;
3569        }
3570    
3571      if (i < max && bytes[i * MAX_N_BYTES] < 255)
3572        {
3573        if (!in_range)
3574          {
3575          in_range = TRUE;
3576          from = i;
3577          }
3578        }
3579      else if (in_range)
3580        in_range = FALSE;
3581      }
3582    
3583    if (range_right >= 0)
3584      {
3585      /* Since no data is consumed (see the assert in the beginning
3586      of this function), this space can be reallocated. */
3587      if (common->read_only_data)
3588        SLJIT_FREE(common->read_only_data);
3589    
3590      common->read_only_data_size += 256;
3591      common->read_only_data = (sljit_uw *)SLJIT_MALLOC(common->read_only_data_size);
3592      if (common->read_only_data == NULL)
3593        return TRUE;
3594    
3595      update_table = (sljit_ub *)common->read_only_data;
3596      common->read_only_data_ptr = (sljit_uw *)(update_table + 256);
3597      memset(update_table, IN_UCHARS(range_len), 256);
3598    
3599      for (i = 0; i < range_len; i++)
3600        {
3601        byte_set = bytes + ((range_right - i) * MAX_N_BYTES);
3602        SLJIT_ASSERT(byte_set[0] > 0 && byte_set[0] < 255);
3603        byte_set_end = byte_set + byte_set[0];
3604        byte_set++;
3605        while (byte_set <= byte_set_end)
3606          {
3607          if (update_table[*byte_set] > IN_UCHARS(i))
3608            update_table[*byte_set] = IN_UCHARS(i);
3609          byte_set++;
3610          }
3611        }
3612      }
3613    
3614    offsets[0] = -1;
3615    /* Scan forward. */
3616    for (i = 0; i < max; i++)
3617      if (ones[i] <= 2) {
3618        offsets[0] = i;
3619        break;
3620      }
3621    
3622    if (offsets[0] < 0 && range_right < 0)
3623      return FALSE;
3624    
3625    if (offsets[0] >= 0)
3626      {
3627      /* Scan backward. */
3628      offsets[1] = -1;
3629      for (i = max - 1; i > offsets[0]; i--)
3630        if (ones[i] <= 2 && i != range_right)
3631          {
3632          offsets[1] = i;
3633          break;
3634          }
3635    
3636      /* This case is handled better by fast_forward_first_char. */
3637      if (offsets[1] == -1 && offsets[0] == 0 && range_right < 0)
3638        return FALSE;
3639    
3640      offsets[2] = -1;
3641      /* We only search for a middle character if there is no range check. */
3642      if (offsets[1] >= 0 && range_right == -1)
3643        {
3644        /* Scan from middle. */
3645        for (i = (offsets[0] + offsets[1]) / 2 + 1; i < offsets[1]; i++)
3646          if (ones[i] <= 2)
3647            {
3648            offsets[2] = i;
3649            break;
3650            }
3651    
3652        if (offsets[2] == -1)
3653          {
3654          for (i = (offsets[0] + offsets[1]) / 2; i > offsets[0]; i--)
3655            if (ones[i] <= 2)
3656              {
3657              offsets[2] = i;
3658              break;
3659              }
3660          }
3661        }
3662    
3663      SLJIT_ASSERT(offsets[1] == -1 || (offsets[0] < offsets[1]));
3664      SLJIT_ASSERT(offsets[2] == -1 || (offsets[0] < offsets[2] && offsets[1] > offsets[2]));
3665    
3666      chars[0] = chars[offsets[0] << 1];
3667      chars[1] = chars[(offsets[0] << 1) + 1];
3668      if (offsets[2] >= 0)
3669        {
3670        chars[2] = chars[offsets[2] << 1];
3671        chars[3] = chars[(offsets[2] << 1) + 1];
3672        }
3673      if (offsets[1] >= 0)
3674        {
3675        chars[4] = chars[offsets[1] << 1];
3676        chars[5] = chars[(offsets[1] << 1) + 1];
3677        }
3678      }
3679    
3680    max -= 1;
3681    if (firstline)
3682      {
3683      SLJIT_ASSERT(common->first_line_end != 0);
3684      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end);
3685      OP1(SLJIT_MOV, TMP3, 0, STR_END, 0);
3686      OP2(SLJIT_SUB, STR_END, 0, STR_END, 0, SLJIT_IMM, IN_UCHARS(max));
3687      quit = CMP(SLJIT_C_LESS_EQUAL, STR_END, 0, TMP1, 0);
3688      OP1(SLJIT_MOV, STR_END, 0, TMP1, 0);
3689      JUMPHERE(quit);
3690      }
3691    else
3692      OP2(SLJIT_SUB, STR_END, 0, STR_END, 0, SLJIT_IMM, IN_UCHARS(max));
3693    
3694    #if !(defined SLJIT_CONFIG_X86_32 && SLJIT_CONFIG_X86_32)
3695    if (range_right >= 0)
3696