/[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 792 by ph10, Wed Dec 7 16:44:48 2011 UTC revision 1303 by zherczeg, Fri Mar 29 09:01:20 2013 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2008 University of Cambridge             Copyright (c) 1997-2013 University of Cambridge
10    
11    The machine code generator part (this module) was written by Zoltan Herczeg    The machine code generator part (this module) was written by Zoltan Herczeg
12                        Copyright (c) 2010-2011                        Copyright (c) 2010-2013
13    
14  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
15  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 46  POSSIBILITY OF SUCH DAMAGE. Line 46  POSSIBILITY OF SUCH DAMAGE.
46    
47  #include "pcre_internal.h"  #include "pcre_internal.h"
48    
49  #ifdef SUPPORT_JIT  #if defined SUPPORT_JIT
50    
51  /* All-in-one: Since we use the JIT compiler only from here,  /* All-in-one: Since we use the JIT compiler only from here,
52  we just include it. This way we don't need to touch the build  we just include it. This way we don't need to touch the build
53  system files. */  system files. */
54    
55  #define SLJIT_MALLOC(size) (pcre_malloc)(size)  #define SLJIT_MALLOC(size) (PUBL(malloc))(size)
56  #define SLJIT_FREE(ptr) (pcre_free)(ptr)  #define SLJIT_FREE(ptr) (PUBL(free))(ptr)
57  #define SLJIT_CONFIG_AUTO 1  #define SLJIT_CONFIG_AUTO 1
58  #define SLJIT_CONFIG_STATIC 1  #define SLJIT_CONFIG_STATIC 1
59  #define SLJIT_VERBOSE 0  #define SLJIT_VERBOSE 0
# Line 62  system files. */ Line 62  system files. */
62  #include "sljit/sljitLir.c"  #include "sljit/sljitLir.c"
63    
64  #if defined SLJIT_CONFIG_UNSUPPORTED && SLJIT_CONFIG_UNSUPPORTED  #if defined SLJIT_CONFIG_UNSUPPORTED && SLJIT_CONFIG_UNSUPPORTED
65  #error "Unsupported architecture"  #error Unsupported architecture
66  #endif  #endif
67    
68  /* Allocate memory on the stack. Fast, but limited size. */  /* Defines for debugging purposes. */
 #define LOCAL_SPACE_SIZE 32768  
69    
70    /* 1 - Use unoptimized capturing brackets.
71       2 - Enable capture_last_ptr (includes option 1). */
72    /* #define DEBUG_FORCE_UNOPTIMIZED_CBRAS 2 */
73    
74    /* 1 - Always have a control head. */
75    /* #define DEBUG_FORCE_CONTROL_HEAD 1 */
76    
77    /* Allocate memory for the regex stack on the real machine stack.
78    Fast, but limited size. */
79    #define MACHINE_STACK_SIZE 32768
80    
81    /* Growth rate for stack allocated by the OS. Should be the multiply
82    of page size. */
83  #define STACK_GROWTH_RATE 8192  #define STACK_GROWTH_RATE 8192
84    
85  /* Enable to check that the allocation could destroy temporaries. */  /* Enable to check that the allocation could destroy temporaries. */
# Line 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 {
161    /* Pointers first. */    /* Pointers first. */
162    struct sljit_stack *stack;    struct sljit_stack *stack;
163    PCRE_SPTR str;    const pcre_uchar *str;
164    PCRE_SPTR begin;    const pcre_uchar *begin;
165    PCRE_SPTR end;    const pcre_uchar *end;
166    int *offsets;    int *offsets;
167    uschar *ptr;    pcre_uchar *uchar_ptr;
168      pcre_uchar *mark_ptr;
169      void *callout_data;
170    /* Everything else after. */    /* Everything else after. */
171    int offsetcount;    int real_offset_count;
172    int calllimit;    int offset_count;
173    uschar notbol;    int call_limit;
174    uschar noteol;    pcre_uint8 notbol;
175    uschar notempty;    pcre_uint8 noteol;
176    uschar notempty_atstart;    pcre_uint8 notempty;
177      pcre_uint8 notempty_atstart;
178  } jit_arguments;  } jit_arguments;
179    
180  typedef struct executable_function {  typedef struct executable_functions {
181    void *executable_func;    void *executable_funcs[JIT_NUMBER_OF_COMPILE_MODES];
182    pcre_jit_callback callback;    PUBL(jit_callback) callback;
183    void *userdata;    void *userdata;
184    sljit_uw executable_size;    pcre_uint32 top_bracket;
185  } executable_function;    sljit_uw executable_sizes[JIT_NUMBER_OF_COMPILE_MODES];
186    } executable_functions;
187    
188  typedef struct jump_list {  typedef struct jump_list {
189    struct sljit_jump *jump;    struct sljit_jump *jump;
190    struct jump_list *next;    struct jump_list *next;
191  } jump_list;  } jump_list;
192    
 enum stub_types { stack_alloc };  
   
193  typedef struct stub_list {  typedef struct stub_list {
   enum stub_types type;  
   int data;  
194    struct sljit_jump *start;    struct sljit_jump *start;
195    struct sljit_label *leave;    struct sljit_label *quit;
196    struct stub_list *next;    struct stub_list *next;
197  } stub_list;  } stub_list;
198    
199    enum frame_types {
200      no_frame = -1,
201      no_stack = -2
202    };
203    
204    enum control_types {
205      type_mark = 0,
206      type_then_trap = 1
207    };
208    
209  typedef int (SLJIT_CALL *jit_function)(jit_arguments *args);  typedef int (SLJIT_CALL *jit_function)(jit_arguments *args);
210    
211  /* The following structure is the key data type for the recursive  /* The following structure is the key data type for the recursive
212  code generator. It is allocated by compile_hotpath, and contains  code generator. It is allocated by compile_matchingpath, and contains
213  the aguments for compile_fallbackpath. Must be the first member  the aguments for compile_backtrackingpath. Must be the first member
214  of its descendants. */  of its descendants. */
215  typedef struct fallback_common {  typedef struct backtrack_common {
216    /* Concatenation stack. */    /* Concatenation stack. */
217    struct fallback_common *prev;    struct backtrack_common *prev;
218    jump_list *nextfallbacks;    jump_list *nextbacktracks;
219    /* Internal stack (for component operators). */    /* Internal stack (for component operators). */
220    struct fallback_common *top;    struct backtrack_common *top;
221    jump_list *topfallbacks;    jump_list *topbacktracks;
222    /* Opcode pointer. */    /* Opcode pointer. */
223    uschar *cc;    pcre_uchar *cc;
224  } fallback_common;  } backtrack_common;
225    
226  typedef struct assert_fallback {  typedef struct assert_backtrack {
227    fallback_common common;    backtrack_common common;
228    jump_list *condfailed;    jump_list *condfailed;
229    /* Less than 0 (-1) if a frame is not needed. */    /* Less than 0 if a frame is not needed. */
230    int framesize;    int framesize;
231    /* Points to our private memory word on the stack. */    /* Points to our private memory word on the stack. */
232    int localptr;    int private_data_ptr;
233    /* For iterators. */    /* For iterators. */
234    struct sljit_label *hotpath;    struct sljit_label *matchingpath;
235  } assert_fallback;  } assert_backtrack;
236    
237  typedef struct bracket_fallback {  typedef struct bracket_backtrack {
238    fallback_common common;    backtrack_common common;
239    /* Where to coninue if an alternative is successfully matched. */    /* Where to coninue if an alternative is successfully matched. */
240    struct sljit_label *althotpath;    struct sljit_label *alternative_matchingpath;
241    /* For rmin and rmax iterators. */    /* For rmin and rmax iterators. */
242    struct sljit_label *recursivehotpath;    struct sljit_label *recursive_matchingpath;
243    /* For greedy ? operator. */    /* For greedy ? operator. */
244    struct sljit_label *zerohotpath;    struct sljit_label *zero_matchingpath;
245    /* Contains the branches of a failed condition. */    /* Contains the branches of a failed condition. */
246    union {    union {
247      /* Both for OP_COND, OP_SCOND. */      /* Both for OP_COND, OP_SCOND. */
248      jump_list *condfailed;      jump_list *condfailed;
249      assert_fallback *assert;      assert_backtrack *assert;
250      /* For OP_ONCE. -1 if not needed. */      /* For OP_ONCE. Less than 0 if not needed. */
251      int framesize;      int framesize;
252    } u;    } u;
253    /* Points to our private memory word on the stack. */    /* Points to our private memory word on the stack. */
254    int localptr;    int private_data_ptr;
255  } bracket_fallback;  } bracket_backtrack;
256    
257  typedef struct bracketpos_fallback {  typedef struct bracketpos_backtrack {
258    fallback_common common;    backtrack_common common;
259    /* Points to our private memory word on the stack. */    /* Points to our private memory word on the stack. */
260    int localptr;    int private_data_ptr;
261    /* Reverting stack is needed. */    /* Reverting stack is needed. */
262    int framesize;    int framesize;
263    /* Allocated stack size. */    /* Allocated stack size. */
264    int stacksize;    int stacksize;
265  } bracketpos_fallback;  } bracketpos_backtrack;
266    
267  typedef struct braminzero_fallback {  typedef struct braminzero_backtrack {
268    fallback_common common;    backtrack_common common;
269    struct sljit_label *hotpath;    struct sljit_label *matchingpath;
270  } braminzero_fallback;  } braminzero_backtrack;
271    
272  typedef struct iterator_fallback {  typedef struct iterator_backtrack {
273    fallback_common common;    backtrack_common common;
274    /* Next iteration. */    /* Next iteration. */
275    struct sljit_label *hotpath;    struct sljit_label *matchingpath;
276  } iterator_fallback;  } iterator_backtrack;
277    
278  typedef struct recurse_entry {  typedef struct recurse_entry {
279    struct recurse_entry *next;    struct recurse_entry *next;
# Line 260  typedef struct recurse_entry { Line 282  typedef struct recurse_entry {
282    /* Collects the calls until the function is not created. */    /* Collects the calls until the function is not created. */
283    jump_list *calls;    jump_list *calls;
284    /* Points to the starting opcode. */    /* Points to the starting opcode. */
285    int start;    sljit_sw start;
286  } recurse_entry;  } recurse_entry;
287    
288  typedef struct recurse_fallback {  typedef struct recurse_backtrack {
289    fallback_common common;    backtrack_common common;
290  } recurse_fallback;    BOOL inlined_pattern;
291    } recurse_backtrack;
292    
293    #define OP_THEN_TRAP OP_TABLE_LENGTH
294    
295    typedef struct then_trap_backtrack {
296      backtrack_common common;
297      /* If then_trap is not NULL, this structure contains the real
298      then_trap for the backtracking path. */
299      struct then_trap_backtrack *then_trap;
300      /* Points to the starting opcode. */
301      sljit_sw start;
302      /* Exit point for the then opcodes of this alternative. */
303      jump_list *quit;
304      /* Frame size of the current alternative. */
305      int framesize;
306    } then_trap_backtrack;
307    
308    #define MAX_RANGE_SIZE 6
309    
310  typedef struct compiler_common {  typedef struct compiler_common {
311      /* The sljit ceneric compiler. */
312    struct sljit_compiler *compiler;    struct sljit_compiler *compiler;
313    uschar *start;    /* First byte code. */
314    int localsize;    pcre_uchar *start;
315    int *localptrs;    /* Maps private data offset to each opcode. */
316    const uschar *fcc;    int *private_data_ptrs;
317    sljit_w lcc;    /* Tells whether the capturing bracket is optimized. */
318    int cbraptr;    pcre_uint8 *optimized_cbracket;
319      /* Tells whether the starting offset is a target of then. */
320      pcre_uint8 *then_offsets;
321      /* Current position where a THEN must jump. */
322      then_trap_backtrack *then_trap;
323      /* Starting offset of private data for capturing brackets. */
324      int cbra_ptr;
325      /* Output vector starting point. Must be divisible by 2. */
326      int ovector_start;
327      /* Last known position of the requested byte. */
328      int req_char_ptr;
329      /* Head of the last recursion. */
330      int recursive_head_ptr;
331      /* First inspected character for partial matching. */
332      int start_used_ptr;
333      /* Starting pointer for partial soft matches. */
334      int hit_start;
335      /* End pointer of the first line. */
336      int first_line_end;
337      /* Points to the marked string. */
338      int mark_ptr;
339      /* Recursive control verb management chain. */
340      int control_head_ptr;
341      /* Points to the last matched capture block index. */
342      int capture_last_ptr;
343      /* Points to the starting position of the current match. */
344      int start_ptr;
345    
346      /* Flipped and lower case tables. */
347      const pcre_uint8 *fcc;
348      sljit_sw lcc;
349      /* Mode can be PCRE_STUDY_JIT_COMPILE and others. */
350      int mode;
351      /* \K is found in the pattern. */
352      BOOL has_set_som;
353      /* (*SKIP:arg) is found in the pattern. */
354      BOOL has_skip_arg;
355      /* (*THEN) is found in the pattern. */
356      BOOL has_then;
357      /* Needs to know the start position anytime. */
358      BOOL needs_start_ptr;
359      /* Currently in recurse or negative assert. */
360      BOOL local_exit;
361      /* Currently in a positive assert. */
362      BOOL positive_assert;
363      /* Newline control. */
364    int nltype;    int nltype;
365    int newline;    int newline;
366    int bsr_nltype;    int bsr_nltype;
367      /* Dollar endonly. */
368    int endonly;    int endonly;
369    sljit_w ctypes;    /* Tables. */
370      sljit_sw ctypes;
371      int digits[2 + MAX_RANGE_SIZE];
372      /* Named capturing brackets. */
373    sljit_uw name_table;    sljit_uw name_table;
374    sljit_w name_count;    sljit_sw name_count;
375    sljit_w name_entry_size;    sljit_sw name_entry_size;
376    struct sljit_label *acceptlabel;  
377      /* Labels and jump lists. */
378      struct sljit_label *partialmatchlabel;
379      struct sljit_label *quit_label;
380      struct sljit_label *forced_quit_label;
381      struct sljit_label *accept_label;
382    stub_list *stubs;    stub_list *stubs;
383    recurse_entry *entries;    recurse_entry *entries;
384    recurse_entry *currententry;    recurse_entry *currententry;
385      jump_list *partialmatch;
386      jump_list *quit;
387      jump_list *positive_assert_quit;
388      jump_list *forced_quit;
389    jump_list *accept;    jump_list *accept;
390    jump_list *calllimit;    jump_list *calllimit;
391    jump_list *stackalloc;    jump_list *stackalloc;
# Line 297  typedef struct compiler_common { Line 396  typedef struct compiler_common {
396    jump_list *vspace;    jump_list *vspace;
397    jump_list *casefulcmp;    jump_list *casefulcmp;
398    jump_list *caselesscmp;    jump_list *caselesscmp;
399      jump_list *reset_match;
400    BOOL jscript_compat;    BOOL jscript_compat;
401  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
402    BOOL utf8;    BOOL utf;
403  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
404    BOOL useucp;    BOOL use_ucp;
405    #endif
406    #ifndef COMPILE_PCRE32
407      jump_list *utfreadchar;
408  #endif  #endif
409    jump_list *utf8readchar;  #ifdef COMPILE_PCRE8
410    jump_list *utf8readtype8;    jump_list *utfreadtype8;
411  #endif  #endif
412    #endif /* SUPPORT_UTF */
413  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
414    jump_list *getucd;    jump_list *getucd;
415  #endif  #endif
# Line 317  typedef struct compare_context { Line 421  typedef struct compare_context {
421    int length;    int length;
422    int sourcereg;    int sourcereg;
423  #if defined SLJIT_UNALIGNED && SLJIT_UNALIGNED  #if defined SLJIT_UNALIGNED && SLJIT_UNALIGNED
424    int byteptr;    int ucharptr;
425    union {    union {
426      int asint;      sljit_si asint;
427      short asshort;      sljit_uh asushort;
428    #if defined COMPILE_PCRE8
429      sljit_ub asbyte;      sljit_ub asbyte;
430      sljit_ub asbytes[4];      sljit_ub asuchars[4];
431    #elif defined COMPILE_PCRE16
432        sljit_uh asuchars[2];
433    #elif defined COMPILE_PCRE32
434        sljit_ui asuchars[1];
435    #endif
436    } c;    } c;
437    union {    union {
438      int asint;      sljit_si asint;
439      short asshort;      sljit_uh asushort;
440    #if defined COMPILE_PCRE8
441      sljit_ub asbyte;      sljit_ub asbyte;
442      sljit_ub asbytes[4];      sljit_ub asuchars[4];
443    #elif defined COMPILE_PCRE16
444        sljit_uh asuchars[2];
445    #elif defined COMPILE_PCRE32
446        sljit_ui asuchars[1];
447    #endif
448    } oc;    } oc;
449  #endif  #endif
450  } compare_context;  } compare_context;
451    
452  enum {  /* Undefine sljit macros. */
453    frame_end = 0,  #undef CMP
   frame_setstrbegin = -1  
 };  
454    
455  /* Used for accessing the elements of the stack. */  /* Used for accessing the elements of the stack. */
456  #define STACK(i)      ((-(i) - 1) * (int)sizeof(sljit_w))  #define STACK(i)      ((-(i) - 1) * (int)sizeof(sljit_sw))
457    
458  #define TMP1          SLJIT_TEMPORARY_REG1  #define TMP1          SLJIT_SCRATCH_REG1
459  #define TMP2          SLJIT_TEMPORARY_REG3  #define TMP2          SLJIT_SCRATCH_REG3
460  #define TMP3          SLJIT_TEMPORARY_EREG2  #define TMP3          SLJIT_TEMPORARY_EREG2
461  #define STR_PTR       SLJIT_GENERAL_REG1  #define STR_PTR       SLJIT_SAVED_REG1
462  #define STR_END       SLJIT_GENERAL_REG2  #define STR_END       SLJIT_SAVED_REG2
463  #define STACK_TOP     SLJIT_TEMPORARY_REG2  #define STACK_TOP     SLJIT_SCRATCH_REG2
464  #define STACK_LIMIT   SLJIT_GENERAL_REG3  #define STACK_LIMIT   SLJIT_SAVED_REG3
465  #define ARGUMENTS     SLJIT_GENERAL_EREG1  #define ARGUMENTS     SLJIT_SAVED_EREG1
466  #define CALL_COUNT    SLJIT_GENERAL_EREG2  #define CALL_COUNT    SLJIT_SAVED_EREG2
467  #define RETURN_ADDR   SLJIT_TEMPORARY_EREG1  #define RETURN_ADDR   SLJIT_TEMPORARY_EREG1
468    
469  /* Locals layout. */  /* Local space layout. */
470  /* These two locals can be used by the current opcode. */  /* These two locals can be used by the current opcode. */
471  #define LOCALS0          (0 * sizeof(sljit_w))  #define LOCALS0          (0 * sizeof(sljit_sw))
472  #define LOCALS1          (1 * sizeof(sljit_w))  #define LOCALS1          (1 * sizeof(sljit_sw))
473  /* Two local variables for possessive quantifiers (char1 cannot use them). */  /* Two local variables for possessive quantifiers (char1 cannot use them). */
474  #define POSSESSIVE0      (2 * sizeof(sljit_w))  #define POSSESSIVE0      (2 * sizeof(sljit_sw))
475  #define POSSESSIVE1      (3 * sizeof(sljit_w))  #define POSSESSIVE1      (3 * sizeof(sljit_sw))
 /* Head of the last recursion. */  
 #define RECURSIVE_HEAD   (4 * sizeof(sljit_w))  
476  /* Max limit of recursions. */  /* Max limit of recursions. */
477  #define CALL_LIMIT       (5 * sizeof(sljit_w))  #define CALL_LIMIT       (4 * sizeof(sljit_sw))
 /* Last known position of the requested byte. */  
 #define REQ_BYTE_PTR     (6 * sizeof(sljit_w))  
 /* End pointer of the first line. */  
 #define FIRSTLINE_END    (7 * sizeof(sljit_w))  
478  /* The output vector is stored on the stack, and contains pointers  /* The output vector is stored on the stack, and contains pointers
479  to characters. The vector data is divided into two groups: the first  to characters. The vector data is divided into two groups: the first
480  group contains the start / end character pointers, and the second is  group contains the start / end character pointers, and the second is
481  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. */
482  #define OVECTOR_START    (8 * sizeof(sljit_w))  #define OVECTOR_START    (common->ovector_start)
483  #define OVECTOR(i)       (OVECTOR_START + (i) * sizeof(sljit_w))  #define OVECTOR(i)       (OVECTOR_START + (i) * sizeof(sljit_sw))
484  #define OVECTOR_PRIV(i)  (common->cbraptr + (i) * sizeof(sljit_w))  #define OVECTOR_PRIV(i)  (common->cbra_ptr + (i) * sizeof(sljit_sw))
485  #define PRIV(cc)         (common->localptrs[(cc) - common->start])  #define PRIVATE_DATA(cc) (common->private_data_ptrs[(cc) - common->start])
486    
487    #if defined COMPILE_PCRE8
488    #define MOV_UCHAR  SLJIT_MOV_UB
489    #define MOVU_UCHAR SLJIT_MOVU_UB
490    #elif defined COMPILE_PCRE16
491    #define MOV_UCHAR  SLJIT_MOV_UH
492    #define MOVU_UCHAR SLJIT_MOVU_UH
493    #elif defined COMPILE_PCRE32
494    #define MOV_UCHAR  SLJIT_MOV_UI
495    #define MOVU_UCHAR SLJIT_MOVU_UI
496    #else
497    #error Unsupported compiling mode
498    #endif
499    
500  /* Shortcuts. */  /* Shortcuts. */
501  #define DEFINE_COMPILER \  #define DEFINE_COMPILER \
# Line 391  the start pointers when the end of the c Line 512  the start pointers when the end of the c
512    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))
513  #define JUMPHERE(jump) \  #define JUMPHERE(jump) \
514    sljit_set_label((jump), sljit_emit_label(compiler))    sljit_set_label((jump), sljit_emit_label(compiler))
515    #define SET_LABEL(jump, label) \
516      sljit_set_label((jump), (label))
517  #define CMP(type, src1, src1w, src2, src2w) \  #define CMP(type, src1, src1w, src2, src2w) \
518    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))
519  #define CMPTO(type, src1, src1w, src2, src2w, label) \  #define CMPTO(type, src1, src1w, src2, src2w, label) \
520    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))
521  #define COND_VALUE(op, dst, dstw, type) \  #define OP_FLAGS(op, dst, dstw, src, srcw, type) \
522    sljit_emit_cond_value(compiler, (op), (dst), (dstw), (type))    sljit_emit_op_flags(compiler, (op), (dst), (dstw), (src), (srcw), (type))
523    #define GET_LOCAL_BASE(dst, dstw, offset) \
524      sljit_get_local_base(compiler, (dst), (dstw), (offset))
525    
526  static uschar* bracketend(uschar* cc)  static pcre_uchar* bracketend(pcre_uchar* cc)
527  {  {
528  SLJIT_ASSERT((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND));  SLJIT_ASSERT((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND));
529  do cc += GET(cc, 1); while (*cc == OP_ALT);  do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 409  return cc; Line 534  return cc;
534    
535  /* Functions whose might need modification for all new supported opcodes:  /* Functions whose might need modification for all new supported opcodes:
536   next_opcode   next_opcode
537   get_localspace   get_private_data_length
538   set_localptrs   set_private_data_ptrs
539   get_framesize   get_framesize
540   init_frame   init_frame
541   get_localsize   get_private_data_copy_length
542   copy_locals   copy_private_data
543   compile_hotpath   compile_matchingpath
544   compile_fallbackpath   compile_backtrackingpath
545  */  */
546    
547  static uschar *next_opcode(compiler_common *common, uschar *cc)  static pcre_uchar *next_opcode(compiler_common *common, pcre_uchar *cc)
548  {  {
549  SLJIT_UNUSED_ARG(common);  SLJIT_UNUSED_ARG(common);
550  switch(*cc)  switch(*cc)
# Line 437  switch(*cc) Line 562  switch(*cc)
562    case OP_WORDCHAR:    case OP_WORDCHAR:
563    case OP_ANY:    case OP_ANY:
564    case OP_ALLANY:    case OP_ALLANY:
565      case OP_NOTPROP:
566      case OP_PROP:
567    case OP_ANYNL:    case OP_ANYNL:
568    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
569    case OP_HSPACE:    case OP_HSPACE:
# Line 449  switch(*cc) Line 576  switch(*cc)
576    case OP_CIRCM:    case OP_CIRCM:
577    case OP_DOLL:    case OP_DOLL:
578    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:  
579    case OP_CRSTAR:    case OP_CRSTAR:
580    case OP_CRMINSTAR:    case OP_CRMINSTAR:
581    case OP_CRPLUS:    case OP_CRPLUS:
582    case OP_CRMINPLUS:    case OP_CRMINPLUS:
583    case OP_CRQUERY:    case OP_CRQUERY:
584    case OP_CRMINQUERY:    case OP_CRMINQUERY:
585      case OP_CRRANGE:
586      case OP_CRMINRANGE:
587      case OP_CLASS:
588      case OP_NCLASS:
589      case OP_REF:
590      case OP_REFI:
591      case OP_RECURSE:
592      case OP_CALLOUT:
593      case OP_ALT:
594      case OP_KET:
595      case OP_KETRMAX:
596      case OP_KETRMIN:
597      case OP_KETRPOS:
598      case OP_REVERSE:
599      case OP_ASSERT:
600      case OP_ASSERT_NOT:
601      case OP_ASSERTBACK:
602      case OP_ASSERTBACK_NOT:
603      case OP_ONCE:
604      case OP_ONCE_NC:
605      case OP_BRA:
606      case OP_BRAPOS:
607      case OP_CBRA:
608      case OP_CBRAPOS:
609      case OP_COND:
610      case OP_SBRA:
611      case OP_SBRAPOS:
612      case OP_SCBRA:
613      case OP_SCBRAPOS:
614      case OP_SCOND:
615      case OP_CREF:
616      case OP_NCREF:
617      case OP_RREF:
618      case OP_NRREF:
619    case OP_DEF:    case OP_DEF:
620    case OP_BRAZERO:    case OP_BRAZERO:
621    case OP_BRAMINZERO:    case OP_BRAMINZERO:
622    case OP_BRAPOSZERO:    case OP_BRAPOSZERO:
623      case OP_PRUNE:
624      case OP_SKIP:
625      case OP_THEN:
626      case OP_COMMIT:
627    case OP_FAIL:    case OP_FAIL:
628    case OP_ACCEPT:    case OP_ACCEPT:
629    case OP_ASSERT_ACCEPT:    case OP_ASSERT_ACCEPT:
630      case OP_CLOSE:
631    case OP_SKIPZERO:    case OP_SKIPZERO:
632    return cc + 1;    return cc + PRIV(OP_lengths)[*cc];
   
   case OP_ANYBYTE:  
 #ifdef SUPPORT_UTF8  
   if (common->utf8) return NULL;  
 #endif  
   return cc + 1;  
633    
634    case OP_CHAR:    case OP_CHAR:
635    case OP_CHARI:    case OP_CHARI:
636    case OP_NOT:    case OP_NOT:
637    case OP_NOTI:    case OP_NOTI:
   
638    case OP_STAR:    case OP_STAR:
639    case OP_MINSTAR:    case OP_MINSTAR:
640    case OP_PLUS:    case OP_PLUS:
641    case OP_MINPLUS:    case OP_MINPLUS:
642    case OP_QUERY:    case OP_QUERY:
643    case OP_MINQUERY:    case OP_MINQUERY:
644      case OP_UPTO:
645      case OP_MINUPTO:
646      case OP_EXACT:
647    case OP_POSSTAR:    case OP_POSSTAR:
648    case OP_POSPLUS:    case OP_POSPLUS:
649    case OP_POSQUERY:    case OP_POSQUERY:
650      case OP_POSUPTO:
651    case OP_STARI:    case OP_STARI:
652    case OP_MINSTARI:    case OP_MINSTARI:
653    case OP_PLUSI:    case OP_PLUSI:
654    case OP_MINPLUSI:    case OP_MINPLUSI:
655    case OP_QUERYI:    case OP_QUERYI:
656    case OP_MINQUERYI:    case OP_MINQUERYI:
657      case OP_UPTOI:
658      case OP_MINUPTOI:
659      case OP_EXACTI:
660    case OP_POSSTARI:    case OP_POSSTARI:
661    case OP_POSPLUSI:    case OP_POSPLUSI:
662    case OP_POSQUERYI:    case OP_POSQUERYI:
663      case OP_POSUPTOI:
664    case OP_NOTSTAR:    case OP_NOTSTAR:
665    case OP_NOTMINSTAR:    case OP_NOTMINSTAR:
666    case OP_NOTPLUS:    case OP_NOTPLUS:
667    case OP_NOTMINPLUS:    case OP_NOTMINPLUS:
668    case OP_NOTQUERY:    case OP_NOTQUERY:
669    case OP_NOTMINQUERY:    case OP_NOTMINQUERY:
670      case OP_NOTUPTO:
671      case OP_NOTMINUPTO:
672      case OP_NOTEXACT:
673    case OP_NOTPOSSTAR:    case OP_NOTPOSSTAR:
674    case OP_NOTPOSPLUS:    case OP_NOTPOSPLUS:
675    case OP_NOTPOSQUERY:    case OP_NOTPOSQUERY:
676      case OP_NOTPOSUPTO:
677    case OP_NOTSTARI:    case OP_NOTSTARI:
678    case OP_NOTMINSTARI:    case OP_NOTMINSTARI:
679    case OP_NOTPLUSI:    case OP_NOTPLUSI:
680    case OP_NOTMINPLUSI:    case OP_NOTMINPLUSI:
681    case OP_NOTQUERYI:    case OP_NOTQUERYI:
682    case OP_NOTMINQUERYI:    case OP_NOTMINQUERYI:
   case OP_NOTPOSSTARI:  
   case OP_NOTPOSPLUSI:  
   case OP_NOTPOSQUERYI:  
   cc += 2;  
 #ifdef SUPPORT_UTF8  
   if (common->utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
 #endif  
   return cc;  
   
   case OP_UPTO:  
   case OP_MINUPTO:  
   case OP_EXACT:  
   case OP_POSUPTO:  
   case OP_UPTOI:  
   case OP_MINUPTOI:  
   case OP_EXACTI:  
   case OP_POSUPTOI:  
   case OP_NOTUPTO:  
   case OP_NOTMINUPTO:  
   case OP_NOTEXACT:  
   case OP_NOTPOSUPTO:  
683    case OP_NOTUPTOI:    case OP_NOTUPTOI:
684    case OP_NOTMINUPTOI:    case OP_NOTMINUPTOI:
685    case OP_NOTEXACTI:    case OP_NOTEXACTI:
686      case OP_NOTPOSSTARI:
687      case OP_NOTPOSPLUSI:
688      case OP_NOTPOSQUERYI:
689    case OP_NOTPOSUPTOI:    case OP_NOTPOSUPTOI:
690    cc += 4;    cc += PRIV(OP_lengths)[*cc];
691  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
692    if (common->utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];    if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
693  #endif  #endif
694    return cc;    return cc;
695    
696    case OP_NOTPROP:    /* Special cases. */
697    case OP_PROP:    case OP_TYPESTAR:
698      case OP_TYPEMINSTAR:
699      case OP_TYPEPLUS:
700      case OP_TYPEMINPLUS:
701      case OP_TYPEQUERY:
702      case OP_TYPEMINQUERY:
703    case OP_TYPEUPTO:    case OP_TYPEUPTO:
704    case OP_TYPEMINUPTO:    case OP_TYPEMINUPTO:
705    case OP_TYPEEXACT:    case OP_TYPEEXACT:
706      case OP_TYPEPOSSTAR:
707      case OP_TYPEPOSPLUS:
708      case OP_TYPEPOSQUERY:
709    case OP_TYPEPOSUPTO:    case OP_TYPEPOSUPTO:
710    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 += 3;  
   return cc;  
   
   case OP_CRRANGE:  
   case OP_CRMINRANGE:  
   return cc + 5;  
711    
712    case OP_CLASS:    case OP_ANYBYTE:
713    case OP_NCLASS:  #ifdef SUPPORT_UTF
714    return cc + 33;    if (common->utf) return NULL;
715    #endif
716      return cc + 1;
717    
718  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
719    case OP_XCLASS:    case OP_XCLASS:
720    return cc + GET(cc, 1);    return cc + GET(cc, 1);
721  #endif  #endif
722    
723    case OP_RECURSE:    case OP_MARK:
724    case OP_ASSERT:    case OP_PRUNE_ARG:
725    case OP_ASSERT_NOT:    case OP_SKIP_ARG:
726    case OP_ASSERTBACK:    case OP_THEN_ARG:
727    case OP_ASSERTBACK_NOT:    return cc + 1 + 2 + cc[1];
   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 + 2;  
728    
729    default:    default:
730      /* All opcodes are supported now! */
731      SLJIT_ASSERT_STOP();
732    return NULL;    return NULL;
733    }    }
734  }  }
735    
736  static int get_localspace(compiler_common *common, uschar *cc, uschar *ccend)  #define CASE_ITERATOR_PRIVATE_DATA_1 \
737        case OP_MINSTAR: \
738        case OP_MINPLUS: \
739        case OP_QUERY: \
740        case OP_MINQUERY: \
741        case OP_MINSTARI: \
742        case OP_MINPLUSI: \
743        case OP_QUERYI: \
744        case OP_MINQUERYI: \
745        case OP_NOTMINSTAR: \
746        case OP_NOTMINPLUS: \
747        case OP_NOTQUERY: \
748        case OP_NOTMINQUERY: \
749        case OP_NOTMINSTARI: \
750        case OP_NOTMINPLUSI: \
751        case OP_NOTQUERYI: \
752        case OP_NOTMINQUERYI:
753    
754    #define CASE_ITERATOR_PRIVATE_DATA_2A \
755        case OP_STAR: \
756        case OP_PLUS: \
757        case OP_STARI: \
758        case OP_PLUSI: \
759        case OP_NOTSTAR: \
760        case OP_NOTPLUS: \
761        case OP_NOTSTARI: \
762        case OP_NOTPLUSI:
763    
764    #define CASE_ITERATOR_PRIVATE_DATA_2B \
765        case OP_UPTO: \
766        case OP_MINUPTO: \
767        case OP_UPTOI: \
768        case OP_MINUPTOI: \
769        case OP_NOTUPTO: \
770        case OP_NOTMINUPTO: \
771        case OP_NOTUPTOI: \
772        case OP_NOTMINUPTOI:
773    
774    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_1 \
775        case OP_TYPEMINSTAR: \
776        case OP_TYPEMINPLUS: \
777        case OP_TYPEQUERY: \
778        case OP_TYPEMINQUERY:
779    
780    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2A \
781        case OP_TYPESTAR: \
782        case OP_TYPEPLUS:
783    
784    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2B \
785        case OP_TYPEUPTO: \
786        case OP_TYPEMINUPTO:
787    
788    static int get_class_iterator_size(pcre_uchar *cc)
789    {
790    switch(*cc)
791      {
792      case OP_CRSTAR:
793      case OP_CRPLUS:
794      return 2;
795    
796      case OP_CRMINSTAR:
797      case OP_CRMINPLUS:
798      case OP_CRQUERY:
799      case OP_CRMINQUERY:
800      return 1;
801    
802      case OP_CRRANGE:
803      case OP_CRMINRANGE:
804      if (GET2(cc, 1) == GET2(cc, 1 + IMM2_SIZE))
805        return 0;
806      return 2;
807    
808      default:
809      return 0;
810      }
811    }
812    
813    static int get_private_data_length(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend)
814  {  {
815  int localspace = 0;  int private_data_length = 0;
816  uschar *alternative;  pcre_uchar *alternative;
817    pcre_uchar *name;
818    pcre_uchar *end = NULL;
819    int space, size, i;
820    pcre_uint32 bracketlen;
821    
822  /* 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. */
823  while (cc < ccend)  while (cc < ccend)
824    {    {
825      space = 0;
826      size = 0;
827      bracketlen = 0;
828    switch(*cc)    switch(*cc)
829      {      {
830        case OP_SET_SOM:
831        common->has_set_som = TRUE;
832        cc += 1;
833        break;
834    
835        case OP_REF:
836        case OP_REFI:
837        common->optimized_cbracket[GET2(cc, 1)] = 0;
838        cc += 1 + IMM2_SIZE;
839        break;
840    
841      case OP_ASSERT:      case OP_ASSERT:
842      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
843      case OP_ASSERTBACK:      case OP_ASSERTBACK:
# Line 628  while (cc < ccend) Line 847  while (cc < ccend)
847      case OP_BRAPOS:      case OP_BRAPOS:
848      case OP_SBRA:      case OP_SBRA:
849      case OP_SBRAPOS:      case OP_SBRAPOS:
850      case OP_SCOND:      private_data_length += sizeof(sljit_sw);
851      localspace += sizeof(sljit_w);      bracketlen = 1 + LINK_SIZE;
     cc += 1 + LINK_SIZE;  
852      break;      break;
853    
854      case OP_CBRAPOS:      case OP_CBRAPOS:
855      case OP_SCBRAPOS:      case OP_SCBRAPOS:
856      localspace += sizeof(sljit_w);      private_data_length += sizeof(sljit_sw);
857      cc += 1 + LINK_SIZE + 2;      common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] = 0;
858        bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
859      break;      break;
860    
861      case OP_COND:      case OP_COND:
862      /* Might be a hidden SCOND. */      case OP_SCOND:
863      alternative = cc + GET(cc, 1);      /* Only AUTO_CALLOUT can insert this opcode. We do
864      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)         not intend to support this case. */
865        localspace += sizeof(sljit_w);      if (cc[1 + LINK_SIZE] == OP_CALLOUT)
866          return -1;
867    
868        if (*cc == OP_COND)
869          {
870          /* Might be a hidden SCOND. */
871          alternative = cc + GET(cc, 1);
872          if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
873            private_data_length += sizeof(sljit_sw);
874          }
875        else
876          private_data_length += sizeof(sljit_sw);
877        bracketlen = 1 + LINK_SIZE;
878        break;
879    
880        case OP_CREF:
881        i = GET2(cc, 1);
882        common->optimized_cbracket[i] = 0;
883        cc += 1 + IMM2_SIZE;
884        break;
885    
886        case OP_NCREF:
887        bracketlen = GET2(cc, 1);
888        name = (pcre_uchar *)common->name_table;
889        alternative = name;
890        for (i = 0; i < common->name_count; i++)
891          {
892          if (GET2(name, 0) == bracketlen) break;
893          name += common->name_entry_size;
894          }
895        SLJIT_ASSERT(i != common->name_count);
896    
897        for (i = 0; i < common->name_count; i++)
898          {
899          if (STRCMP_UC_UC(alternative + IMM2_SIZE, name + IMM2_SIZE) == 0)
900            common->optimized_cbracket[GET2(alternative, 0)] = 0;
901          alternative += common->name_entry_size;
902          }
903        bracketlen = 0;
904        cc += 1 + IMM2_SIZE;
905        break;
906    
907        case OP_BRA:
908        bracketlen = 1 + LINK_SIZE;
909        break;
910    
911        case OP_CBRA:
912        case OP_SCBRA:
913        bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
914        break;
915    
916        CASE_ITERATOR_PRIVATE_DATA_1
917        space = 1;
918        size = -2;
919        break;
920    
921        CASE_ITERATOR_PRIVATE_DATA_2A
922        space = 2;
923        size = -2;
924        break;
925    
926        CASE_ITERATOR_PRIVATE_DATA_2B
927        space = 2;
928        size = -(2 + IMM2_SIZE);
929        break;
930    
931        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
932        space = 1;
933        size = 1;
934        break;
935    
936        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
937        if (cc[1] != OP_ANYNL && cc[1] != OP_EXTUNI)
938          space = 2;
939        size = 1;
940        break;
941    
942        CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
943        if (cc[1 + IMM2_SIZE] != OP_ANYNL && cc[1 + IMM2_SIZE] != OP_EXTUNI)
944          space = 2;
945        size = 1 + IMM2_SIZE;
946        break;
947    
948        case OP_CLASS:
949        case OP_NCLASS:
950        size += 1 + 32 / sizeof(pcre_uchar);
951        space = get_class_iterator_size(cc + size);
952        break;
953    
954    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
955        case OP_XCLASS:
956        size = GET(cc, 1);
957        space = get_class_iterator_size(cc + size);
958        break;
959    #endif
960    
961        case OP_RECURSE:
962        /* Set its value only once. */
963        if (common->recursive_head_ptr == 0)
964          {
965          common->recursive_head_ptr = common->ovector_start;
966          common->ovector_start += sizeof(sljit_sw);
967          }
968      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
969      break;      break;
970    
971        case OP_CALLOUT:
972        if (common->capture_last_ptr == 0)
973          {
974          common->capture_last_ptr = common->ovector_start;
975          common->ovector_start += sizeof(sljit_sw);
976          }
977        cc += 2 + 2 * LINK_SIZE;
978        break;
979    
980        case OP_THEN_ARG:
981        common->has_then = TRUE;
982        common->control_head_ptr = 1;
983        /* Fall through. */
984    
985        case OP_PRUNE_ARG:
986        common->needs_start_ptr = TRUE;
987        /* Fall through. */
988    
989        case OP_MARK:
990        if (common->mark_ptr == 0)
991          {
992          common->mark_ptr = common->ovector_start;
993          common->ovector_start += sizeof(sljit_sw);
994          }
995        cc += 1 + 2 + cc[1];
996        break;
997    
998        case OP_THEN:
999        common->has_then = TRUE;
1000        common->control_head_ptr = 1;
1001        /* Fall through. */
1002    
1003        case OP_PRUNE:
1004        case OP_SKIP:
1005        common->needs_start_ptr = TRUE;
1006        cc += 1;
1007        break;
1008    
1009        case OP_SKIP_ARG:
1010        common->control_head_ptr = 1;
1011        common->has_skip_arg = TRUE;
1012        cc += 1 + 2 + cc[1];
1013        break;
1014    
1015      default:      default:
1016      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
1017      if (cc == NULL)      if (cc == NULL)
1018        return -1;        return -1;
1019      break;      break;
1020      }      }
1021    
1022      if (space > 0 && cc >= end)
1023        private_data_length += sizeof(sljit_sw) * space;
1024    
1025      if (size != 0)
1026        {
1027        if (size < 0)
1028          {
1029          cc += -size;
1030    #ifdef SUPPORT_UTF
1031          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1032    #endif
1033          }
1034        else
1035          cc += size;
1036        }
1037    
1038      if (bracketlen != 0)
1039        {
1040        if (cc >= end)
1041          {
1042          end = bracketend(cc);
1043          if (end[-1 - LINK_SIZE] == OP_KET)
1044            end = NULL;
1045          }
1046        cc += bracketlen;
1047        }
1048    }    }
1049  return localspace;  return private_data_length;
1050  }  }
1051    
1052  static void set_localptrs(compiler_common *common, int localptr, uschar *ccend)  static void set_private_data_ptrs(compiler_common *common, int private_data_ptr, pcre_uchar *ccend)
1053  {  {
1054  uschar *cc = common->start;  pcre_uchar *cc = common->start;
1055  uschar *alternative;  pcre_uchar *alternative;
1056    pcre_uchar *end = NULL;
1057    int space, size, bracketlen;
1058    
1059  while (cc < ccend)  while (cc < ccend)
1060    {    {
1061      space = 0;
1062      size = 0;
1063      bracketlen = 0;
1064    switch(*cc)    switch(*cc)
1065      {      {
1066      case OP_ASSERT:      case OP_ASSERT:
# Line 675  while (cc < ccend) Line 1073  while (cc < ccend)
1073      case OP_SBRA:      case OP_SBRA:
1074      case OP_SBRAPOS:      case OP_SBRAPOS:
1075      case OP_SCOND:      case OP_SCOND:
1076      common->localptrs[cc - common->start] = localptr;      common->private_data_ptrs[cc - common->start] = private_data_ptr;
1077      localptr += sizeof(sljit_w);      private_data_ptr += sizeof(sljit_sw);
1078      cc += 1 + LINK_SIZE;      bracketlen = 1 + LINK_SIZE;
1079      break;      break;
1080    
1081      case OP_CBRAPOS:      case OP_CBRAPOS:
1082      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1083      common->localptrs[cc - common->start] = localptr;      common->private_data_ptrs[cc - common->start] = private_data_ptr;
1084      localptr += sizeof(sljit_w);      private_data_ptr += sizeof(sljit_sw);
1085      cc += 1 + LINK_SIZE + 2;      bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
1086      break;      break;
1087    
1088      case OP_COND:      case OP_COND:
# Line 692  while (cc < ccend) Line 1090  while (cc < ccend)
1090      alternative = cc + GET(cc, 1);      alternative = cc + GET(cc, 1);
1091      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1092        {        {
1093        common->localptrs[cc - common->start] = localptr;        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1094        localptr += sizeof(sljit_w);        private_data_ptr += sizeof(sljit_sw);
1095        }        }
1096      cc += 1 + LINK_SIZE;      bracketlen = 1 + LINK_SIZE;
1097        break;
1098    
1099        case OP_BRA:
1100        bracketlen = 1 + LINK_SIZE;
1101        break;
1102    
1103        case OP_CBRA:
1104        case OP_SCBRA:
1105        bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
1106        break;
1107    
1108        CASE_ITERATOR_PRIVATE_DATA_1
1109        space = 1;
1110        size = -2;
1111        break;
1112    
1113        CASE_ITERATOR_PRIVATE_DATA_2A
1114        space = 2;
1115        size = -2;
1116        break;
1117    
1118        CASE_ITERATOR_PRIVATE_DATA_2B
1119        space = 2;
1120        size = -(2 + IMM2_SIZE);
1121        break;
1122    
1123        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1124        space = 1;
1125        size = 1;
1126        break;
1127    
1128        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1129        if (cc[1] != OP_ANYNL && cc[1] != OP_EXTUNI)
1130          space = 2;
1131        size = 1;
1132      break;      break;
1133    
1134        CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1135        if (cc[1 + IMM2_SIZE] != OP_ANYNL && cc[1 + IMM2_SIZE] != OP_EXTUNI)
1136          space = 2;
1137        size = 1 + IMM2_SIZE;
1138        break;
1139    
1140        case OP_CLASS:
1141        case OP_NCLASS:
1142        size += 1 + 32 / sizeof(pcre_uchar);
1143        space = get_class_iterator_size(cc + size);
1144        break;
1145    
1146    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1147        case OP_XCLASS:
1148        size = GET(cc, 1);
1149        space = get_class_iterator_size(cc + size);
1150        break;
1151    #endif
1152    
1153      default:      default:
1154      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
1155      SLJIT_ASSERT(cc != NULL);      SLJIT_ASSERT(cc != NULL);
1156      break;      break;
1157      }      }
1158    
1159      if (space > 0 && cc >= end)
1160        {
1161        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1162        private_data_ptr += sizeof(sljit_sw) * space;
1163        }
1164    
1165      if (size != 0)
1166        {
1167        if (size < 0)
1168          {
1169          cc += -size;
1170    #ifdef SUPPORT_UTF
1171          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1172    #endif
1173          }
1174        else
1175          cc += size;
1176        }
1177    
1178      if (bracketlen > 0)
1179        {
1180        if (cc >= end)
1181          {
1182          end = bracketend(cc);
1183          if (end[-1 - LINK_SIZE] == OP_KET)
1184            end = NULL;
1185          }
1186        cc += bracketlen;
1187        }
1188    }    }
1189  }  }
1190    
1191  /* Returns with -1 if no need for frame. */  /* Returns with a frame_types (always < 0) if no need for frame. */
1192  static int get_framesize(compiler_common *common, uschar *cc, BOOL recursive)  static int get_framesize(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, BOOL recursive, BOOL* needs_control_head)
1193  {  {
 uschar *ccend = bracketend(cc);  
1194  int length = 0;  int length = 0;
1195  BOOL possessive = FALSE;  int possessive = 0;
1196  BOOL setsom_found = FALSE;  BOOL stack_restore = FALSE;
1197    BOOL setsom_found = recursive;
1198    BOOL setmark_found = recursive;
1199    /* The last capture is a local variable even for recursions. */
1200    BOOL capture_last_found = FALSE;
1201    
1202    #if defined DEBUG_FORCE_CONTROL_HEAD && DEBUG_FORCE_CONTROL_HEAD
1203    SLJIT_ASSERT(common->control_head_ptr != 0);
1204    *needs_control_head = TRUE;
1205    #else
1206    *needs_control_head = FALSE;
1207    #endif
1208    
1209  if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))  if (ccend == NULL)
1210    {    {
1211    length = 3;    ccend = bracketend(cc) - (1 + LINK_SIZE);
1212    possessive = TRUE;    if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))
1213        {
1214        possessive = length = (common->capture_last_ptr != 0) ? 5 : 3;
1215        /* This is correct regardless of common->capture_last_ptr. */
1216        capture_last_found = TRUE;
1217        }
1218      cc = next_opcode(common, cc);
1219    }    }
1220    
 cc = next_opcode(common, cc);  
1221  SLJIT_ASSERT(cc != NULL);  SLJIT_ASSERT(cc != NULL);
1222  while (cc < ccend)  while (cc < ccend)
1223    switch(*cc)    switch(*cc)
1224      {      {
1225      case OP_SET_SOM:      case OP_SET_SOM:
1226      case OP_RECURSE:      SLJIT_ASSERT(common->has_set_som);
1227        stack_restore = TRUE;
1228      if (!setsom_found)      if (!setsom_found)
1229        {        {
1230        length += 2;        length += 2;
1231        setsom_found = TRUE;        setsom_found = TRUE;
1232        }        }
1233      cc += (*cc == OP_SET_SOM) ? 1 : 1 + LINK_SIZE;      cc += 1;
1234        break;
1235    
1236        case OP_MARK:
1237        case OP_PRUNE_ARG:
1238        case OP_THEN_ARG:
1239        SLJIT_ASSERT(common->mark_ptr != 0);
1240        stack_restore = TRUE;
1241        if (!setmark_found)
1242          {
1243          length += 2;
1244          setmark_found = TRUE;
1245          }
1246        if (common->control_head_ptr != 0)
1247          *needs_control_head = TRUE;
1248        cc += 1 + 2 + cc[1];
1249        break;
1250    
1251        case OP_RECURSE:
1252        stack_restore = TRUE;
1253        if (common->has_set_som && !setsom_found)
1254          {
1255          length += 2;
1256          setsom_found = TRUE;
1257          }
1258        if (common->mark_ptr != 0 && !setmark_found)
1259          {
1260          length += 2;
1261          setmark_found = TRUE;
1262          }
1263        if (common->capture_last_ptr != 0 && !capture_last_found)
1264          {
1265          length += 2;
1266          capture_last_found = TRUE;
1267          }
1268        cc += 1 + LINK_SIZE;
1269      break;      break;
1270    
1271      case OP_CBRA:      case OP_CBRA:
1272      case OP_CBRAPOS:      case OP_CBRAPOS:
1273      case OP_SCBRA:      case OP_SCBRA:
1274      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1275        stack_restore = TRUE;
1276        if (common->capture_last_ptr != 0 && !capture_last_found)
1277          {
1278          length += 2;
1279          capture_last_found = TRUE;
1280          }
1281      length += 3;      length += 3;
1282      cc += 1 + LINK_SIZE + 2;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1283      break;      break;
1284    
1285      default:      default:
1286        stack_restore = TRUE;
1287        /* Fall through. */
1288    
1289        case OP_NOT_WORD_BOUNDARY:
1290        case OP_WORD_BOUNDARY:
1291        case OP_NOT_DIGIT:
1292        case OP_DIGIT:
1293        case OP_NOT_WHITESPACE:
1294        case OP_WHITESPACE:
1295        case OP_NOT_WORDCHAR:
1296        case OP_WORDCHAR:
1297        case OP_ANY:
1298        case OP_ALLANY:
1299        case OP_ANYBYTE:
1300        case OP_NOTPROP:
1301        case OP_PROP:
1302        case OP_ANYNL:
1303        case OP_NOT_HSPACE:
1304        case OP_HSPACE:
1305        case OP_NOT_VSPACE:
1306        case OP_VSPACE:
1307        case OP_EXTUNI:
1308        case OP_EODN:
1309        case OP_EOD:
1310        case OP_CIRC:
1311        case OP_CIRCM:
1312        case OP_DOLL:
1313        case OP_DOLLM:
1314        case OP_CHAR:
1315        case OP_CHARI:
1316        case OP_NOT:
1317        case OP_NOTI:
1318    
1319        case OP_EXACT:
1320        case OP_POSSTAR:
1321        case OP_POSPLUS:
1322        case OP_POSQUERY:
1323        case OP_POSUPTO:
1324    
1325        case OP_EXACTI:
1326        case OP_POSSTARI:
1327        case OP_POSPLUSI:
1328        case OP_POSQUERYI:
1329        case OP_POSUPTOI:
1330    
1331        case OP_NOTEXACT:
1332        case OP_NOTPOSSTAR:
1333        case OP_NOTPOSPLUS:
1334        case OP_NOTPOSQUERY:
1335        case OP_NOTPOSUPTO:
1336    
1337        case OP_NOTEXACTI:
1338        case OP_NOTPOSSTARI:
1339        case OP_NOTPOSPLUSI:
1340        case OP_NOTPOSQUERYI:
1341        case OP_NOTPOSUPTOI:
1342    
1343        case OP_TYPEEXACT:
1344        case OP_TYPEPOSSTAR:
1345        case OP_TYPEPOSPLUS:
1346        case OP_TYPEPOSQUERY:
1347        case OP_TYPEPOSUPTO:
1348    
1349        case OP_CLASS:
1350        case OP_NCLASS:
1351        case OP_XCLASS:
1352    
1353      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
1354      SLJIT_ASSERT(cc != NULL);      SLJIT_ASSERT(cc != NULL);
1355      break;      break;
1356      }      }
1357    
1358  /* Possessive quantifiers can use a special case. */  /* Possessive quantifiers can use a special case. */
1359  if (SLJIT_UNLIKELY(possessive) && length == 3)  if (SLJIT_UNLIKELY(possessive == length))
1360    return -1;    return stack_restore ? no_frame : no_stack;
1361    
1362  if (length > 0)  if (length > 0)
1363    return length + 1;    return length + 1;
1364  return -1;  return stack_restore ? no_frame : no_stack;
1365  }  }
1366    
1367  static void init_frame(compiler_common *common, uschar *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)
1368  {  {
1369  DEFINE_COMPILER;  DEFINE_COMPILER;
1370  uschar *ccend = bracketend(cc);  BOOL setsom_found = recursive;
1371  BOOL setsom_found = FALSE;  BOOL setmark_found = recursive;
1372    /* The last capture is a local variable even for recursions. */
1373    BOOL capture_last_found = FALSE;
1374  int offset;  int offset;
1375    
1376  /* >= 1 + shortest item size (2) */  /* >= 1 + shortest item size (2) */
1377    SLJIT_UNUSED_ARG(stacktop);
1378  SLJIT_ASSERT(stackpos >= stacktop + 2);  SLJIT_ASSERT(stackpos >= stacktop + 2);
1379    
1380  stackpos = STACK(stackpos);  stackpos = STACK(stackpos);
1381  if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))  if (ccend == NULL)
1382    cc = next_opcode(common, cc);    {
1383  SLJIT_ASSERT(cc != NULL);    ccend = bracketend(cc) - (1 + LINK_SIZE);
1384  while (cc < ccend)    if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))
1385    switch(*cc)      cc = next_opcode(common, cc);
1386      {    }
1387      case OP_SET_SOM:  
1388    SLJIT_ASSERT(cc != NULL);
1389    while (cc < ccend)
1390      switch(*cc)
1391        {
1392        case OP_SET_SOM:
1393        SLJIT_ASSERT(common->has_set_som);
1394        if (!setsom_found)
1395          {
1396          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));
1397          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1398          stackpos += (int)sizeof(sljit_sw);
1399          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1400          stackpos += (int)sizeof(sljit_sw);
1401          setsom_found = TRUE;
1402          }
1403        cc += 1;
1404        break;
1405    
1406        case OP_MARK:
1407        case OP_PRUNE_ARG:
1408        case OP_THEN_ARG:
1409        SLJIT_ASSERT(common->mark_ptr != 0);
1410        if (!setmark_found)
1411          {
1412          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);
1413          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
1414          stackpos += (int)sizeof(sljit_sw);
1415          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1416          stackpos += (int)sizeof(sljit_sw);
1417          setmark_found = TRUE;
1418          }
1419        cc += 1 + 2 + cc[1];
1420        break;
1421    
1422      case OP_RECURSE:      case OP_RECURSE:
1423      if (!setsom_found)      if (common->has_set_som && !setsom_found)
1424        {        {
1425        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));
1426        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setstrbegin);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1427        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1428        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1429        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1430        setsom_found = TRUE;        setsom_found = TRUE;
1431        }        }
1432      cc += (*cc == OP_SET_SOM) ? 1 : 1 + LINK_SIZE;      if (common->mark_ptr != 0 && !setmark_found)
1433          {
1434          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);
1435          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
1436          stackpos += (int)sizeof(sljit_sw);
1437          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1438          stackpos += (int)sizeof(sljit_sw);
1439          setmark_found = TRUE;
1440          }
1441        if (common->capture_last_ptr != 0 && !capture_last_found)
1442          {
1443          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->capture_last_ptr);
1444          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1445          stackpos += (int)sizeof(sljit_sw);
1446          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1447          stackpos += (int)sizeof(sljit_sw);
1448          capture_last_found = TRUE;
1449          }
1450        cc += 1 + LINK_SIZE;
1451      break;      break;
1452    
1453      case OP_CBRA:      case OP_CBRA:
1454      case OP_CBRAPOS:      case OP_CBRAPOS:
1455      case OP_SCBRA:      case OP_SCBRA:
1456      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1457        if (common->capture_last_ptr != 0 && !capture_last_found)
1458          {
1459          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->capture_last_ptr);
1460          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1461          stackpos += (int)sizeof(sljit_sw);
1462          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1463          stackpos += (int)sizeof(sljit_sw);
1464          capture_last_found = TRUE;
1465          }
1466      offset = (GET2(cc, 1 + LINK_SIZE)) << 1;      offset = (GET2(cc, 1 + LINK_SIZE)) << 1;
1467      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));
1468      stackpos += (int)sizeof(sljit_w);      stackpos += (int)sizeof(sljit_sw);
1469      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset));      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset));
1470      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));
1471      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1472      stackpos += (int)sizeof(sljit_w);      stackpos += (int)sizeof(sljit_sw);
1473      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);
1474      stackpos += (int)sizeof(sljit_w);      stackpos += (int)sizeof(sljit_sw);
1475    
1476      cc += 1 + LINK_SIZE + 2;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1477      break;      break;
1478    
1479      default:      default:
# Line 812  while (cc < ccend) Line 1482  while (cc < ccend)
1482      break;      break;
1483      }      }
1484    
1485  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_end);  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, 0);
1486  SLJIT_ASSERT(stackpos == STACK(stacktop));  SLJIT_ASSERT(stackpos == STACK(stacktop));
1487  }  }
1488    
1489  static SLJIT_INLINE int get_localsize(compiler_common *common, uschar *cc, uschar *ccend)  static SLJIT_INLINE int get_private_data_copy_length(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, BOOL needs_control_head)
1490  {  {
1491  int localsize = 2;  int private_data_length = needs_control_head ? 3 : 2;
1492  uschar *alternative;  int size;
1493  /* Calculate the sum of the local variables. */  pcre_uchar *alternative;
1494    /* Calculate the sum of the private machine words. */
1495  while (cc < ccend)  while (cc < ccend)
1496    {    {
1497      size = 0;
1498    switch(*cc)    switch(*cc)
1499      {      {
1500      case OP_ASSERT:      case OP_ASSERT:
# Line 835  while (cc < ccend) Line 1507  while (cc < ccend)
1507      case OP_SBRA:      case OP_SBRA:
1508      case OP_SBRAPOS:      case OP_SBRAPOS:
1509      case OP_SCOND:      case OP_SCOND:
1510      localsize++;      private_data_length++;
1511      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1512      break;      break;
1513    
1514      case OP_CBRA:      case OP_CBRA:
1515      case OP_SCBRA:      case OP_SCBRA:
1516      localsize++;      if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1517      cc += 1 + LINK_SIZE + 2;        private_data_length++;
1518        cc += 1 + LINK_SIZE + IMM2_SIZE;
1519      break;      break;
1520    
1521      case OP_CBRAPOS:      case OP_CBRAPOS:
1522      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1523      localsize += 2;      private_data_length += 2;
1524      cc += 1 + LINK_SIZE + 2;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1525      break;      break;
1526    
1527      case OP_COND:      case OP_COND:
1528      /* Might be a hidden SCOND. */      /* Might be a hidden SCOND. */
1529      alternative = cc + GET(cc, 1);      alternative = cc + GET(cc, 1);
1530      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1531        localsize++;        private_data_length++;
1532      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1533      break;      break;
1534    
1535        CASE_ITERATOR_PRIVATE_DATA_1
1536        if (PRIVATE_DATA(cc))
1537          private_data_length++;
1538        cc += 2;
1539    #ifdef SUPPORT_UTF
1540        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1541    #endif
1542        break;
1543    
1544        CASE_ITERATOR_PRIVATE_DATA_2A
1545        if (PRIVATE_DATA(cc))
1546          private_data_length += 2;
1547        cc += 2;
1548    #ifdef SUPPORT_UTF
1549        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1550    #endif
1551        break;
1552    
1553        CASE_ITERATOR_PRIVATE_DATA_2B
1554        if (PRIVATE_DATA(cc))
1555          private_data_length += 2;
1556        cc += 2 + IMM2_SIZE;
1557    #ifdef SUPPORT_UTF
1558        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1559    #endif
1560        break;
1561    
1562        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1563        if (PRIVATE_DATA(cc))
1564          private_data_length++;
1565        cc += 1;
1566        break;
1567    
1568        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1569        if (PRIVATE_DATA(cc))
1570          private_data_length += 2;
1571        cc += 1;
1572        break;
1573    
1574        CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1575        if (PRIVATE_DATA(cc))
1576          private_data_length += 2;
1577        cc += 1 + IMM2_SIZE;
1578        break;
1579    
1580        case OP_CLASS:
1581        case OP_NCLASS:
1582    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1583        case OP_XCLASS:
1584        size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
1585    #else
1586        size = 1 + 32 / (int)sizeof(pcre_uchar);
1587    #endif
1588        if (PRIVATE_DATA(cc))
1589          private_data_length += get_class_iterator_size(cc + size);
1590        cc += size;
1591        break;
1592    
1593      default:      default:
1594      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
1595      SLJIT_ASSERT(cc != NULL);      SLJIT_ASSERT(cc != NULL);
# Line 866  while (cc < ccend) Line 1597  while (cc < ccend)
1597      }      }
1598    }    }
1599  SLJIT_ASSERT(cc == ccend);  SLJIT_ASSERT(cc == ccend);
1600  return localsize;  return private_data_length;
1601  }  }
1602    
1603  static void copy_locals(compiler_common *common, uschar *cc, uschar *ccend,  static void copy_private_data(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend,
1604    BOOL save, int stackptr, int stacktop)    BOOL save, int stackptr, int stacktop, BOOL needs_control_head)
1605  {  {
1606  DEFINE_COMPILER;  DEFINE_COMPILER;
1607  int srcw[2];  int srcw[2];
1608  int count;  int count, size;
1609  BOOL tmp1next = TRUE;  BOOL tmp1next = TRUE;
1610  BOOL tmp1empty = TRUE;  BOOL tmp1empty = TRUE;
1611  BOOL tmp2empty = TRUE;  BOOL tmp2empty = TRUE;
1612  uschar *alternative;  pcre_uchar *alternative;
1613  enum {  enum {
1614    start,    start,
1615    loop,    loop,
# Line 891  stacktop = STACK(stacktop - 1); Line 1622  stacktop = STACK(stacktop - 1);
1622    
1623  if (!save)  if (!save)
1624    {    {
1625    stackptr += sizeof(sljit_w);    stackptr += (needs_control_head ? 2 : 1) * sizeof(sljit_sw);
1626    if (stackptr < stacktop)    if (stackptr < stacktop)
1627      {      {
1628      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1629      stackptr += sizeof(sljit_w);      stackptr += sizeof(sljit_sw);
1630      tmp1empty = FALSE;      tmp1empty = FALSE;
1631      }      }
1632    if (stackptr < stacktop)    if (stackptr < stacktop)
1633      {      {
1634      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1635      stackptr += sizeof(sljit_w);      stackptr += sizeof(sljit_sw);
1636      tmp2empty = FALSE;      tmp2empty = FALSE;
1637      }      }
1638    /* The tmp1next must be TRUE in either way. */    /* The tmp1next must be TRUE in either way. */
1639    }    }
1640    
1641  while (status != end)  do
1642    {    {
1643    count = 0;    count = 0;
1644    switch(status)    switch(status)
1645      {      {
1646      case start:      case start:
1647      SLJIT_ASSERT(save);      SLJIT_ASSERT(save && common->recursive_head_ptr != 0);
1648      count = 1;      count = 1;
1649      srcw[0] = RECURSIVE_HEAD;      srcw[0] = common->recursive_head_ptr;
1650        if (needs_control_head)
1651          {
1652          SLJIT_ASSERT(common->control_head_ptr != 0);
1653          count = 2;
1654          srcw[1] = common->control_head_ptr;
1655          }
1656      status = loop;      status = loop;
1657      break;      break;
1658    
# Line 939  while (status != end) Line 1676  while (status != end)
1676        case OP_SBRAPOS:        case OP_SBRAPOS:
1677        case OP_SCOND:        case OP_SCOND:
1678        count = 1;        count = 1;
1679        srcw[0] = PRIV(cc);        srcw[0] = PRIVATE_DATA(cc);
1680        SLJIT_ASSERT(srcw[0] != 0);        SLJIT_ASSERT(srcw[0] != 0);
1681        cc += 1 + LINK_SIZE;        cc += 1 + LINK_SIZE;
1682        break;        break;
1683    
1684        case OP_CBRA:        case OP_CBRA:
1685        case OP_SCBRA:        case OP_SCBRA:
1686        count = 1;        if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1687        srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));          {
1688        cc += 1 + LINK_SIZE + 2;          count = 1;
1689            srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1690            }
1691          cc += 1 + LINK_SIZE + IMM2_SIZE;
1692        break;        break;
1693    
1694        case OP_CBRAPOS:        case OP_CBRAPOS:
1695        case OP_SCBRAPOS:        case OP_SCBRAPOS:
1696        count = 2;        count = 2;
1697          srcw[0] = PRIVATE_DATA(cc);
1698        srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));        srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1699        srcw[0] = PRIV(cc);        SLJIT_ASSERT(srcw[0] != 0 && srcw[1] != 0);
1700        SLJIT_ASSERT(srcw[0] != 0);        cc += 1 + LINK_SIZE + IMM2_SIZE;
       cc += 1 + LINK_SIZE + 2;  
1701        break;        break;
1702    
1703        case OP_COND:        case OP_COND:
# Line 966  while (status != end) Line 1706  while (status != end)
1706        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1707          {          {
1708          count = 1;          count = 1;
1709          srcw[0] = PRIV(cc);          srcw[0] = PRIVATE_DATA(cc);
1710          SLJIT_ASSERT(srcw[0] != 0);          SLJIT_ASSERT(srcw[0] != 0);
1711          }          }
1712        cc += 1 + LINK_SIZE;        cc += 1 + LINK_SIZE;
1713        break;        break;
1714    
1715          CASE_ITERATOR_PRIVATE_DATA_1
1716          if (PRIVATE_DATA(cc))
1717            {
1718            count = 1;
1719            srcw[0] = PRIVATE_DATA(cc);
1720            }
1721          cc += 2;
1722    #ifdef SUPPORT_UTF
1723          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1724    #endif
1725          break;
1726    
1727          CASE_ITERATOR_PRIVATE_DATA_2A
1728          if (PRIVATE_DATA(cc))
1729            {
1730            count = 2;
1731            srcw[0] = PRIVATE_DATA(cc);
1732            srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1733            }
1734          cc += 2;
1735    #ifdef SUPPORT_UTF
1736          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1737    #endif
1738          break;
1739    
1740          CASE_ITERATOR_PRIVATE_DATA_2B
1741          if (PRIVATE_DATA(cc))
1742            {
1743            count = 2;
1744            srcw[0] = PRIVATE_DATA(cc);
1745            srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1746            }
1747          cc += 2 + IMM2_SIZE;
1748    #ifdef SUPPORT_UTF
1749          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1750    #endif
1751          break;
1752    
1753          CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1754          if (PRIVATE_DATA(cc))
1755            {
1756            count = 1;
1757            srcw[0] = PRIVATE_DATA(cc);
1758            }
1759          cc += 1;
1760          break;
1761    
1762          CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1763          if (PRIVATE_DATA(cc))
1764            {
1765            count = 2;
1766            srcw[0] = PRIVATE_DATA(cc);
1767            srcw[1] = srcw[0] + sizeof(sljit_sw);
1768            }
1769          cc += 1;
1770          break;
1771    
1772          CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1773          if (PRIVATE_DATA(cc))
1774            {
1775            count = 2;
1776            srcw[0] = PRIVATE_DATA(cc);
1777            srcw[1] = srcw[0] + sizeof(sljit_sw);
1778            }
1779          cc += 1 + IMM2_SIZE;
1780          break;
1781    
1782          case OP_CLASS:
1783          case OP_NCLASS:
1784    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1785          case OP_XCLASS:
1786          size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
1787    #else
1788          size = 1 + 32 / (int)sizeof(pcre_uchar);
1789    #endif
1790          if (PRIVATE_DATA(cc))
1791            switch(get_class_iterator_size(cc + size))
1792              {
1793              case 1:
1794              count = 1;
1795              srcw[0] = PRIVATE_DATA(cc);
1796              break;
1797    
1798              case 2:
1799              count = 2;
1800              srcw[0] = PRIVATE_DATA(cc);
1801              srcw[1] = srcw[0] + sizeof(sljit_sw);
1802              break;
1803    
1804              default:
1805              SLJIT_ASSERT_STOP();
1806              break;
1807              }
1808          cc += size;
1809          break;
1810    
1811        default:        default:
1812        cc = next_opcode(common, cc);        cc = next_opcode(common, cc);
1813        SLJIT_ASSERT(cc != NULL);        SLJIT_ASSERT(cc != NULL);
# Line 994  while (status != end) Line 1830  while (status != end)
1830          if (!tmp1empty)          if (!tmp1empty)
1831            {            {
1832            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1833            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1834            }            }
1835          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);
1836          tmp1empty = FALSE;          tmp1empty = FALSE;
# Line 1005  while (status != end) Line 1841  while (status != end)
1841          if (!tmp2empty)          if (!tmp2empty)
1842            {            {
1843            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1844            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1845            }            }
1846          OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);          OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);
1847          tmp2empty = FALSE;          tmp2empty = FALSE;
# Line 1022  while (status != end) Line 1858  while (status != end)
1858          if (!tmp1empty)          if (!tmp1empty)
1859            {            {
1860            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1861            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1862            }            }
1863          tmp1next = FALSE;          tmp1next = FALSE;
1864          }          }
# Line 1034  while (status != end) Line 1870  while (status != end)
1870          if (!tmp2empty)          if (!tmp2empty)
1871            {            {
1872            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1873            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1874            }            }
1875          tmp1next = TRUE;          tmp1next = TRUE;
1876          }          }
1877        }        }
1878      }      }
1879    }    }
1880    while (status != end);
1881    
1882  if (save)  if (save)
1883    {    {
# Line 1049  if (save) Line 1886  if (save)
1886      if (!tmp1empty)      if (!tmp1empty)
1887        {        {
1888        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1889        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1890        }        }
1891      if (!tmp2empty)      if (!tmp2empty)
1892        {        {
1893        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1894        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1895        }        }
1896      }      }
1897    else    else
# Line 1062  if (save) Line 1899  if (save)
1899      if (!tmp2empty)      if (!tmp2empty)
1900        {        {
1901        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1902        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1903        }        }
1904      if (!tmp1empty)      if (!tmp1empty)
1905        {        {
1906        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1907        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1908        }        }
1909      }      }
1910    }    }
1911  SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));  SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));
1912  }  }
1913    
1914  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)
1915    {
1916    pcre_uchar *end = bracketend(cc);
1917    BOOL has_alternatives = cc[GET(cc, 1)] == OP_ALT;
1918    
1919    /* Assert captures then. */
1920    if (*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT)
1921      current_offset = NULL;
1922    /* Conditional block does not. */
1923    if (*cc == OP_COND || *cc == OP_SCOND)
1924      has_alternatives = FALSE;
1925    
1926    cc = next_opcode(common, cc);
1927    if (has_alternatives)
1928      current_offset = common->then_offsets + (cc - common->start);
1929    
1930    while (cc < end)
1931      {
1932      if ((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND))
1933        cc = set_then_offsets(common, cc, current_offset);
1934      else
1935        {
1936        if (*cc == OP_ALT && has_alternatives)
1937          current_offset = common->then_offsets + (cc + 1 + LINK_SIZE - common->start);
1938        if (*cc >= OP_THEN && *cc <= OP_THEN_ARG && current_offset != NULL)
1939          *current_offset = 1;
1940        cc = next_opcode(common, cc);
1941        }
1942      }
1943    
1944    return end;
1945    }
1946    
1947    #undef CASE_ITERATOR_PRIVATE_DATA_1
1948    #undef CASE_ITERATOR_PRIVATE_DATA_2A
1949    #undef CASE_ITERATOR_PRIVATE_DATA_2B
1950    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1951    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1952    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1953    
1954    static SLJIT_INLINE BOOL is_powerof2(unsigned int value)
1955  {  {
1956  return (value & (value - 1)) == 0;  return (value & (value - 1)) == 0;
1957  }  }
# Line 1084  static SLJIT_INLINE void set_jumps(jump_ Line 1961  static SLJIT_INLINE void set_jumps(jump_
1961  while (list)  while (list)
1962    {    {
1963    /* sljit_set_label is clever enough to do nothing    /* sljit_set_label is clever enough to do nothing
1964    if either the jump or the label is NULL */    if either the jump or the label is NULL. */
1965    sljit_set_label(list->jump, label);    SET_LABEL(list->jump, label);
1966    list = list->next;    list = list->next;
1967    }    }
1968  }  }
# Line 1101  if (list_item) Line 1978  if (list_item)
1978    }    }
1979  }  }
1980    
1981  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)
1982  {  {
1983  DEFINE_COMPILER;  DEFINE_COMPILER;
1984  stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));  stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));
1985    
1986  if (list_item)  if (list_item)
1987    {    {
   list_item->type = type;  
   list_item->data = data;  
1988    list_item->start = start;    list_item->start = start;
1989    list_item->leave = LABEL();    list_item->quit = LABEL();
1990    list_item->next = common->stubs;    list_item->next = common->stubs;
1991    common->stubs = list_item;    common->stubs = list_item;
1992    }    }
# Line 1125  stub_list* list_item = common->stubs; Line 2000  stub_list* list_item = common->stubs;
2000  while (list_item)  while (list_item)
2001    {    {
2002    JUMPHERE(list_item->start);    JUMPHERE(list_item->start);
2003    switch(list_item->type)    add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));
2004      {    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);  
2005    list_item = list_item->next;    list_item = list_item->next;
2006    }    }
2007  common->stubs = NULL;  common->stubs = NULL;
# Line 1150  static SLJIT_INLINE void allocate_stack( Line 2020  static SLJIT_INLINE void allocate_stack(
2020  /* May destroy all locals and registers except TMP2. */  /* May destroy all locals and registers except TMP2. */
2021  DEFINE_COMPILER;  DEFINE_COMPILER;
2022    
2023  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));
2024  #ifdef DESTROY_REGISTERS  #ifdef DESTROY_REGISTERS
2025  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);
2026  OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);  OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);
# Line 1158  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0); Line 2028  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);
2028  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);
2029  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);
2030  #endif  #endif
2031  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));
2032  }  }
2033    
2034  static SLJIT_INLINE void free_stack(compiler_common *common, int size)  static SLJIT_INLINE void free_stack(compiler_common *common, int size)
2035  {  {
2036  DEFINE_COMPILER;  DEFINE_COMPILER;
2037  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));
2038  }  }
2039    
2040  static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)  static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)
# Line 1172  static SLJIT_INLINE void reset_ovector(c Line 2042  static SLJIT_INLINE void reset_ovector(c
2042  DEFINE_COMPILER;  DEFINE_COMPILER;
2043  struct sljit_label *loop;  struct sljit_label *loop;
2044  int i;  int i;
2045    
2046  /* At this point we can freely use all temporary registers. */  /* At this point we can freely use all temporary registers. */
2047    SLJIT_ASSERT(length > 1);
2048  /* TMP1 returns with begin - 1. */  /* TMP1 returns with begin - 1. */
2049  OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG1, 0, SLJIT_MEM1(SLJIT_GENERAL_REG1), SLJIT_OFFSETOF(jit_arguments, begin), SLJIT_IMM, 1);  OP2(SLJIT_SUB, SLJIT_SCRATCH_REG1, 0, SLJIT_MEM1(SLJIT_SAVED_REG1), SLJIT_OFFSETOF(jit_arguments, begin), SLJIT_IMM, IN_UCHARS(1));
2050    if (length < 8)
2051      {
2052      for (i = 1; i < length; i++)
2053        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), SLJIT_SCRATCH_REG1, 0);
2054      }
2055    else
2056      {
2057      GET_LOCAL_BASE(SLJIT_SCRATCH_REG2, 0, OVECTOR_START);
2058      OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, length - 1);
2059      loop = LABEL();
2060      OP1(SLJIT_MOVU, SLJIT_MEM1(SLJIT_SCRATCH_REG2), sizeof(sljit_sw), SLJIT_SCRATCH_REG1, 0);
2061      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 1);
2062      JUMPTO(SLJIT_C_NOT_ZERO, loop);
2063      }
2064    }
2065    
2066    static SLJIT_INLINE void do_reset_match(compiler_common *common, int length)
2067    {
2068    DEFINE_COMPILER;
2069    struct sljit_label *loop;
2070    int i;
2071    
2072    SLJIT_ASSERT(length > 1);
2073    /* OVECTOR(1) contains the "string begin - 1" constant. */
2074    if (length > 2)
2075      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));
2076  if (length < 8)  if (length < 8)
2077    {    {
2078    for (i = 0; i < length; i++)    for (i = 2; i < length; i++)
2079      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);
2080    }    }
2081  else  else
2082    {    {
2083    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));
2084    OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, length);    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_IMM, length - 2);
2085    loop = LABEL();    loop = LABEL();
2086    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);
2087    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);
2088    JUMPTO(SLJIT_C_NOT_ZERO, loop);    JUMPTO(SLJIT_C_NOT_ZERO, loop);
2089    }    }
2090    
2091    OP1(SLJIT_MOV, STACK_TOP, 0, ARGUMENTS, 0);
2092    if (common->mark_ptr != 0)
2093      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr, SLJIT_IMM, 0);
2094    if (common->control_head_ptr != 0)
2095      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->control_head_ptr, SLJIT_IMM, 0);
2096    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(jit_arguments, stack));
2097    OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_ptr);
2098    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(struct sljit_stack, base));
2099    }
2100    
2101    static sljit_sw SLJIT_CALL do_search_mark(sljit_sw *current, const pcre_uchar *skip_arg)
2102    {
2103    while (current != NULL)
2104      {
2105      switch (current[-2])
2106        {
2107        case type_then_trap:
2108        break;
2109    
2110        case type_mark:
2111        if (STRCMP_UC_UC(skip_arg, (pcre_uchar *)current[-3]) == 0)
2112          return current[-4];
2113        break;
2114    
2115        default:
2116        SLJIT_ASSERT_STOP();
2117        break;
2118        }
2119      current = (sljit_sw*)current[-1];
2120      }
2121    return -1;
2122  }  }
2123    
2124  static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)  static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)
2125  {  {
2126  DEFINE_COMPILER;  DEFINE_COMPILER;
2127  struct sljit_label *loop;  struct sljit_label *loop;
2128  struct sljit_jump *earlyexit;  struct sljit_jump *early_quit;
2129    
2130  /* At this point we can freely use all registers. */  /* At this point we can freely use all registers. */
2131  OP1(SLJIT_MOV, SLJIT_GENERAL_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));  OP1(SLJIT_MOV, SLJIT_SAVED_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));
2132  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);
2133    
2134  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, ARGUMENTS, 0);  OP1(SLJIT_MOV, SLJIT_SCRATCH_REG1, 0, ARGUMENTS, 0);
2135  OP1(SLJIT_MOV_SI, SLJIT_TEMPORARY_REG2, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, offsetcount));  if (common->mark_ptr != 0)
2136  OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);
2137  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, begin));  OP1(SLJIT_MOV_SI, SLJIT_SCRATCH_REG2, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, offset_count));
2138  OP2(SLJIT_ADD, SLJIT_GENERAL_REG1, 0, SLJIT_LOCALS_REG, 0, SLJIT_IMM, OVECTOR_START);  if (common->mark_ptr != 0)
2139      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, mark_ptr), SLJIT_SCRATCH_REG3, 0);
2140    OP2(SLJIT_SUB, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));
2141    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG1, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, begin));
2142    GET_LOCAL_BASE(SLJIT_SAVED_REG1, 0, OVECTOR_START);
2143  /* Unlikely, but possible */  /* Unlikely, but possible */
2144  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);
2145  loop = LABEL();  loop = LABEL();
2146  OP2(SLJIT_SUB, SLJIT_GENERAL_REG2, 0, SLJIT_MEM1(SLJIT_GENERAL_REG1), 0, SLJIT_TEMPORARY_REG1, 0);  OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, SLJIT_MEM1(SLJIT_SAVED_REG1), 0, SLJIT_SCRATCH_REG1, 0);
2147  OP2(SLJIT_ADD, SLJIT_GENERAL_REG1, 0, SLJIT_GENERAL_REG1, 0, SLJIT_IMM, sizeof(sljit_w));  OP2(SLJIT_ADD, SLJIT_SAVED_REG1, 0, SLJIT_SAVED_REG1, 0, SLJIT_IMM, sizeof(sljit_sw));
2148  /* Copy the integer value to the output buffer */  /* Copy the integer value to the output buffer */
2149  OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_TEMPORARY_REG3), sizeof(int), SLJIT_GENERAL_REG2, 0);  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2150  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_TEMPORARY_REG2, 0, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, 1);  OP2(SLJIT_ASHR, SLJIT_SAVED_REG2, 0, SLJIT_SAVED_REG2, 0, SLJIT_IMM, UCHAR_SHIFT);
2151    #endif
2152    OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG3), sizeof(int), SLJIT_SAVED_REG2, 0);
2153    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_SCRATCH_REG2, 0, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, 1);
2154  JUMPTO(SLJIT_C_NOT_ZERO, loop);  JUMPTO(SLJIT_C_NOT_ZERO, loop);
2155  JUMPHERE(earlyexit);  JUMPHERE(early_quit);
2156    
2157  /* Calculate the return value, which is the maximum ovector value. */  /* Calculate the return value, which is the maximum ovector value. */
2158  if (topbracket > 1)  if (topbracket > 1)
2159    {    {
2160    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));
2161    OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, topbracket + 1);    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, topbracket + 1);
2162    
2163    /* OVECTOR(0) is never equal to SLJIT_GENERAL_REG3. */    /* OVECTOR(0) is never equal to SLJIT_SAVED_REG3. */
2164    loop = LABEL();    loop = LABEL();
2165    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)));
2166    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);
2167    CMPTO(SLJIT_C_EQUAL, SLJIT_TEMPORARY_REG3, 0, SLJIT_GENERAL_REG3, 0, loop);    CMPTO(SLJIT_C_EQUAL, SLJIT_SCRATCH_REG3, 0, SLJIT_SAVED_REG3, 0, loop);
2168    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_TEMPORARY_REG2, 0);    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_SCRATCH_REG2, 0);
2169    }    }
2170  else  else
2171    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, 1);    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, 1);
2172  }  }
2173    
2174  static SLJIT_INLINE BOOL char_has_othercase(compiler_common *common, uschar* cc)  static SLJIT_INLINE void return_with_partial_match(compiler_common *common, struct sljit_label *quit)
2175    {
2176    DEFINE_COMPILER;
2177    struct sljit_jump *jump;
2178    
2179    SLJIT_COMPILE_ASSERT(STR_END == SLJIT_SAVED_REG2, str_end_must_be_saved_reg2);
2180    SLJIT_ASSERT(common->start_used_ptr != 0 && common->start_ptr != 0
2181      && (common->mode == JIT_PARTIAL_SOFT_COMPILE ? common->hit_start != 0 : common->hit_start == 0));
2182    
2183    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, ARGUMENTS, 0);
2184    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, PCRE_ERROR_PARTIAL);
2185    OP1(SLJIT_MOV_SI, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, real_offset_count));
2186    CMPTO(SLJIT_C_SIG_LESS, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 2, quit);
2187    
2188    /* Store match begin and end. */
2189    OP1(SLJIT_MOV, SLJIT_SAVED_REG1, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, begin));
2190    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, offsets));
2191    
2192    jump = CMP(SLJIT_C_SIG_LESS, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 3);
2193    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);
2194    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2195    OP2(SLJIT_ASHR, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, UCHAR_SHIFT);
2196    #endif
2197    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG2), 2 * sizeof(int), SLJIT_SCRATCH_REG3, 0);
2198    JUMPHERE(jump);
2199    
2200    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);
2201    OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, STR_END, 0, SLJIT_SAVED_REG1, 0);
2202    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2203    OP2(SLJIT_ASHR, SLJIT_SAVED_REG2, 0, SLJIT_SAVED_REG2, 0, SLJIT_IMM, UCHAR_SHIFT);
2204    #endif
2205    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG2), sizeof(int), SLJIT_SAVED_REG2, 0);
2206    
2207    OP2(SLJIT_SUB, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_SAVED_REG1, 0);
2208    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2209    OP2(SLJIT_ASHR, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, UCHAR_SHIFT);
2210    #endif
2211    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG2), 0, SLJIT_SCRATCH_REG3, 0);
2212    
2213    JUMPTO(SLJIT_JUMP, quit);
2214    }
2215    
2216    static SLJIT_INLINE void check_start_used_ptr(compiler_common *common)
2217    {
2218    /* May destroy TMP1. */
2219    DEFINE_COMPILER;
2220    struct sljit_jump *jump;
2221    
2222    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2223      {
2224      /* The value of -1 must be kept for start_used_ptr! */
2225      OP2(SLJIT_ADD, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, SLJIT_IMM, 1);
2226      /* Jumps if start_used_ptr < STR_PTR, or start_used_ptr == -1. Although overwriting
2227      is not necessary if start_used_ptr == STR_PTR, it does not hurt as well. */
2228      jump = CMP(SLJIT_C_LESS_EQUAL, TMP1, 0, STR_PTR, 0);
2229      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);
2230      JUMPHERE(jump);
2231      }
2232    else if (common->mode == JIT_PARTIAL_HARD_COMPILE)
2233      {
2234      jump = CMP(SLJIT_C_LESS_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);
2235      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);
2236      JUMPHERE(jump);
2237      }
2238    }
2239    
2240    static SLJIT_INLINE BOOL char_has_othercase(compiler_common *common, pcre_uchar* cc)
2241  {  {
2242  /* Detects if the character has an othercase. */  /* Detects if the character has an othercase. */
2243  unsigned int c;  unsigned int c;
2244    
2245  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2246  if (common->utf8)  if (common->utf)
2247    {    {
2248    GETCHAR(c, cc);    GETCHAR(c, cc);
2249    if (c > 127)    if (c > 127)
# Line 1251  if (common->utf8) Line 2254  if (common->utf8)
2254      return FALSE;      return FALSE;
2255  #endif  #endif
2256      }      }
2257    #ifndef COMPILE_PCRE8
2258      return common->fcc[c] != c;
2259    #endif
2260    }    }
2261  else  else
2262  #endif  #endif
2263    c = *cc;    c = *cc;
2264  return common->fcc[c] != c;  return MAX_255(c) ? common->fcc[c] != c : FALSE;
2265  }  }
2266    
2267  static SLJIT_INLINE unsigned int char_othercase(compiler_common *common, unsigned int c)  static SLJIT_INLINE unsigned int char_othercase(compiler_common *common, unsigned int c)
2268  {  {
2269  /* Returns with the othercase. */  /* Returns with the othercase. */
2270  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2271  if (common->utf8 && c > 127)  if (common->utf && c > 127)
2272    {    {
2273  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2274    return UCD_OTHERCASE(c);    return UCD_OTHERCASE(c);
# Line 1271  if (common->utf8 && c > 127) Line 2277  if (common->utf8 && c > 127)
2277  #endif  #endif
2278    }    }
2279  #endif  #endif
2280  return common->fcc[c];  return TABLE_GET(c, common->fcc, c);
2281  }  }
2282    
2283  static unsigned int char_get_othercase_bit(compiler_common *common, uschar* cc)  static unsigned int char_get_othercase_bit(compiler_common *common, pcre_uchar* cc)
2284  {  {
2285  /* Detects if the character and its othercase has only 1 bit difference. */  /* Detects if the character and its othercase has only 1 bit difference. */
2286  unsigned int c, oc, bit;  unsigned int c, oc, bit;
2287  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2288  int n;  int n;
2289  #endif  #endif
2290    
2291  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2292  if (common->utf8)  if (common->utf)
2293    {    {
2294    GETCHAR(c, cc);    GETCHAR(c, cc);
2295    if (c <= 127)    if (c <= 127)
# Line 1300  if (common->utf8) Line 2306  if (common->utf8)
2306  else  else
2307    {    {
2308    c = *cc;    c = *cc;
2309    oc = common->fcc[c];    oc = TABLE_GET(c, common->fcc, c);
2310    }    }
2311  #else  #else
2312  c = *cc;  c = *cc;
2313  oc = common->fcc[c];  oc = TABLE_GET(c, common->fcc, c);
2314  #endif  #endif
2315    
2316  SLJIT_ASSERT(c != oc);  SLJIT_ASSERT(c != oc);
# Line 1315  if (c <= 127 && bit == 0x20) Line 2321  if (c <= 127 && bit == 0x20)
2321    return (0 << 8) | 0x20;    return (0 << 8) | 0x20;
2322    
2323  /* Since c != oc, they must have at least 1 bit difference. */  /* Since c != oc, they must have at least 1 bit difference. */
2324  if (!ispowerof2(bit))  if (!is_powerof2(bit))
2325    return 0;    return 0;
2326    
2327  #ifdef SUPPORT_UTF8  #if defined COMPILE_PCRE8
2328  if (common->utf8 && c > 127)  
2329    #ifdef SUPPORT_UTF
2330    if (common->utf && c > 127)
2331    {    {
2332    n = _pcre_utf8_table4[*cc & 0x3f];    n = GET_EXTRALEN(*cc);
2333    while ((bit & 0x3f) == 0)    while ((bit & 0x3f) == 0)
2334      {      {
2335      n--;      n--;
# Line 1329  if (common->utf8 && c > 127) Line 2337  if (common->utf8 && c > 127)
2337      }      }
2338    return (n << 8) | bit;    return (n << 8) | bit;
2339    }    }
2340  #endif  #endif /* SUPPORT_UTF */
2341  return (0 << 8) | bit;  return (0 << 8) | bit;
2342    
2343    #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2344    
2345    #ifdef SUPPORT_UTF
2346    if (common->utf && c > 65535)
2347      {
2348      if (bit >= (1 << 10))
2349        bit >>= 10;
2350      else
2351        return (bit < 256) ? ((2 << 8) | bit) : ((3 << 8) | (bit >> 8));
2352      }
2353    #endif /* SUPPORT_UTF */
2354    return (bit < 256) ? ((0 << 8) | bit) : ((1 << 8) | (bit >> 8));
2355    
2356    #endif /* COMPILE_PCRE[8|16|32] */
2357    }
2358    
2359    static void check_partial(compiler_common *common, BOOL force)
2360    {
2361    /* Checks whether a partial matching is occured. Does not modify registers. */
2362    DEFINE_COMPILER;
2363    struct sljit_jump *jump = NULL;
2364    
2365    SLJIT_ASSERT(!force || common->mode != JIT_COMPILE);
2366    
2367    if (common->mode == JIT_COMPILE)
2368      return;
2369    
2370    if (!force)
2371      jump = CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);
2372    else if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2373      jump = CMP(SLJIT_C_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, SLJIT_IMM, -1);
2374    
2375    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2376      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2377    else
2378      {
2379      if (common->partialmatchlabel != NULL)
2380        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2381      else
2382        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2383      }
2384    
2385    if (jump != NULL)
2386      JUMPHERE(jump);
2387    }
2388    
2389    static void check_str_end(compiler_common *common, jump_list **end_reached)
2390    {
2391    /* Does not affect registers. Usually used in a tight spot. */
2392    DEFINE_COMPILER;
2393    struct sljit_jump *jump;
2394    
2395    if (common->mode == JIT_COMPILE)
2396      {
2397      add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2398      return;
2399      }
2400    
2401    jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);
2402    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2403      {
2404      add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));
2405      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2406      add_jump(compiler, end_reached, JUMP(SLJIT_JUMP));
2407      }
2408    else
2409      {
2410      add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));
2411      if (common->partialmatchlabel != NULL)
2412        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2413      else
2414        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2415      }
2416    JUMPHERE(jump);
2417  }  }
2418    
2419  static SLJIT_INLINE void check_input_end(compiler_common *common, jump_list **fallbacks)  static void detect_partial_match(compiler_common *common, jump_list **backtracks)
2420  {  {
2421  DEFINE_COMPILER;  DEFINE_COMPILER;
2422  add_jump(compiler, fallbacks, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));  struct sljit_jump *jump;
2423    
2424    if (common->mode == JIT_COMPILE)
2425      {
2426      add_jump(compiler, backtracks, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2427      return;
2428      }
2429    
2430    /* Partial matching mode. */
2431    jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);
2432    add_jump(compiler, backtracks, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));
2433    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2434      {
2435      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2436      add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));
2437      }
2438    else
2439      {
2440      if (common->partialmatchlabel != NULL)
2441        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2442      else
2443        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2444      }
2445    JUMPHERE(jump);
2446  }  }
2447    
2448  static void read_char(compiler_common *common)  static void read_char(compiler_common *common)
# Line 1344  static void read_char(compiler_common *c Line 2450  static void read_char(compiler_common *c
2450  /* Reads the character into TMP1, updates STR_PTR.  /* Reads the character into TMP1, updates STR_PTR.
2451  Does not check STR_END. TMP2 Destroyed. */  Does not check STR_END. TMP2 Destroyed. */
2452  DEFINE_COMPILER;  DEFINE_COMPILER;
2453  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2454  struct sljit_jump *jump;  struct sljit_jump *jump;
2455  #endif  #endif
2456    
2457  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2458  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2459  if (common->utf8)  if (common->utf)
2460    {    {
2461    #if defined COMPILE_PCRE8
2462    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2463    add_jump(compiler, &common->utf8readchar, JUMP(SLJIT_FAST_CALL));  #elif defined COMPILE_PCRE16
2464      jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);
2465    #endif /* COMPILE_PCRE[8|16] */
2466      add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2467    JUMPHERE(jump);    JUMPHERE(jump);
2468    }    }
2469  #endif  #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2470  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2471  }  }
2472    
2473  static void peek_char(compiler_common *common)  static void peek_char(compiler_common *common)
# Line 1365  static void peek_char(compiler_common *c Line 2475  static void peek_char(compiler_common *c
2475  /* Reads the character into TMP1, keeps STR_PTR.  /* Reads the character into TMP1, keeps STR_PTR.
2476  Does not check STR_END. TMP2 Destroyed. */  Does not check STR_END. TMP2 Destroyed. */
2477  DEFINE_COMPILER;  DEFINE_COMPILER;
2478  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2479  struct sljit_jump *jump;  struct sljit_jump *jump;
2480  #endif  #endif
2481    
2482  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2483  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2484  if (common->utf8)  if (common->utf)
2485    {    {
2486    #if defined COMPILE_PCRE8
2487    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2488    add_jump(compiler, &common->utf8readchar, JUMP(SLJIT_FAST_CALL));  #elif defined COMPILE_PCRE16
2489      jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);
2490    #endif /* COMPILE_PCRE[8|16] */
2491      add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2492    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2493    JUMPHERE(jump);    JUMPHERE(jump);
2494    }    }
2495  #endif  #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2496  }  }
2497    
2498  static void read_char8_type(compiler_common *common)  static void read_char8_type(compiler_common *common)
2499  {  {
2500  /* Reads the character type into TMP1, updates STR_PTR. Does not check STR_END. */  /* Reads the character type into TMP1, updates STR_PTR. Does not check STR_END. */
2501  DEFINE_COMPILER;  DEFINE_COMPILER;
2502  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2503  struct sljit_jump *jump;  struct sljit_jump *jump;
2504  #endif  #endif
2505    
2506  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2507  if (common->utf8)  if (common->utf)
2508    {    {
2509    OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2510    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2511    #if defined COMPILE_PCRE8
2512    /* This can be an extra read in some situations, but hopefully    /* This can be an extra read in some situations, but hopefully
2513    it is a clever early read in most cases. */    it is needed in most cases. */
2514    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2515    jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);    jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2516    add_jump(compiler, &common->utf8readtype8, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->utfreadtype8, JUMP(SLJIT_FAST_CALL));
2517      JUMPHERE(jump);
2518    #elif defined COMPILE_PCRE16
2519      OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2520      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);
2521      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2522    JUMPHERE(jump);    JUMPHERE(jump);
2523      /* Skip low surrogate if necessary. */
2524      OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0xfc00);
2525      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0xd800);
2526      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
2527      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 1);
2528      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2529    #elif defined COMPILE_PCRE32
2530      OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2531      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);
2532      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2533      JUMPHERE(jump);
2534    #endif /* COMPILE_PCRE[8|16|32] */
2535    return;    return;
2536    }    }
2537    #endif /* SUPPORT_UTF */
2538    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2539    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2540    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2541    /* The ctypes array contains only 256 values. */
2542    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2543    jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);
2544    #endif
2545    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2546    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2547    JUMPHERE(jump);
2548  #endif  #endif
 OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);  
 OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), common->ctypes);  
2549  }  }
2550    
2551  static void skip_char_back(compiler_common *common)  static void skip_char_back(compiler_common *common)
2552  {  {
2553  /* Goes one character back. Only affects STR_PTR. Does not check begin. */  /* Goes one character back. Affects STR_PTR and TMP1. Does not check begin. */
2554  DEFINE_COMPILER;  DEFINE_COMPILER;
2555  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2556    #if defined COMPILE_PCRE8
2557  struct sljit_label *label;  struct sljit_label *label;
2558    
2559  if (common->utf8)  if (common->utf)
2560    {    {
2561    label = LABEL();    label = LABEL();
2562    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
2563    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2564    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xc0);    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xc0);
2565    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, 0x80, label);    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, 0x80, label);
2566    return;    return;
2567    }    }
2568  #endif  #elif defined COMPILE_PCRE16
2569  OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);  if (common->utf)
2570      {
2571      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
2572      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2573      /* Skip low surrogate if necessary. */
2574      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
2575      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xdc00);
2576      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
2577      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
2578      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2579      return;
2580      }
2581    #endif /* COMPILE_PCRE[8|16] */
2582    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2583    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2584  }  }
2585    
2586  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 jumpiftrue)
2587  {  {
2588  /* 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. */
2589  DEFINE_COMPILER;  DEFINE_COMPILER;
# Line 1436  DEFINE_COMPILER; Line 2591  DEFINE_COMPILER;
2591  if (nltype == NLTYPE_ANY)  if (nltype == NLTYPE_ANY)
2592    {    {
2593    add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));
2594    add_jump(compiler, fallbacks, JUMP(jumpiftrue ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));    add_jump(compiler, backtracks, JUMP(jumpiftrue ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));
2595    }    }
2596  else if (nltype == NLTYPE_ANYCRLF)  else if (nltype == NLTYPE_ANYCRLF)
2597    {    {
2598    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_CR);    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_CR);
2599    COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
2600    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_NL);    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_NL);
2601    COND_VALUE(SLJIT_OR | SLJIT_SET_E, TMP2, 0, SLJIT_C_EQUAL);    OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);
2602    add_jump(compiler, fallbacks, JUMP(jumpiftrue ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));    add_jump(compiler, backtracks, JUMP(jumpiftrue ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));
2603    }    }
2604  else  else
2605    {    {
2606    SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline <= 255);    SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline < 256);
2607    add_jump(compiler, fallbacks, CMP(jumpiftrue ? SLJIT_C_EQUAL : SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, common->newline));    add_jump(compiler, backtracks, CMP(jumpiftrue ? SLJIT_C_EQUAL : SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, common->newline));
2608    }    }
2609  }  }
2610    
2611  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2612  static void do_utf8readchar(compiler_common *common)  
2613    #if defined COMPILE_PCRE8
2614    static void do_utfreadchar(compiler_common *common)
2615  {  {
2616  /* Fast decoding an utf8 character. TMP1 contains the first byte  /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2617  of the character (>= 0xc0). Return char value in TMP1, length - 1 in TMP2. */  of the character (>= 0xc0). Return char value in TMP1, length - 1 in TMP2. */
2618  DEFINE_COMPILER;  DEFINE_COMPILER;
2619  struct sljit_jump *jump;  struct sljit_jump *jump;
2620    
2621  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2622  /* Searching for the first zero. */  /* Searching for the first zero. */
2623  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, 0x20);
2624  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_C_NOT_ZERO);
2625  /* 2 byte sequence */  /* Two byte sequence. */
2626  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 1);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2627  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2628  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x1f);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x1f);
2629  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2630  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2631  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2632  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 1);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(1));
2633  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2634  JUMPHERE(jump);  JUMPHERE(jump);
2635    
2636  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, 0x10);
2637  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_C_NOT_ZERO);
2638  /* 3 byte sequence */  /* Three byte sequence. */
2639  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 1);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2640  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x0f);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x0f);
2641  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 12);  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 12);
2642  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2643  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2644  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2645  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 2);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2646  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 2);  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2647  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2648  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2649  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 2);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(2));
2650  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2651  JUMPHERE(jump);  JUMPHERE(jump);
2652    
2653  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x08);  /* Four byte sequence. */
2654  jump = JUMP(SLJIT_C_NOT_ZERO);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2655  /* 4 byte sequence */  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x07);
2656  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 1);  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 18);
 OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x07);  
 OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 18);  
 OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  
 OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 12);  
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
 OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 2);  
 OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  
 OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
 OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 3);  
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 3);  
 OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
 OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 3);  
 sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  
 JUMPHERE(jump);  
   
 /* 5 byte sequence */  
 OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 1);  
 OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x03);  
 OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 24);  
 OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  
 OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 18);  
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
 OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 2);  
2657  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2658  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 12);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 12);
2659  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2660  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 3);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2661  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2662  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2663  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2664  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 4);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(3));
2665  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 4);  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2666  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2667  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2668  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 4);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(3));
2669  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2670  }  }
2671    
2672  static void do_utf8readtype8(compiler_common *common)  static void do_utfreadtype8(compiler_common *common)
2673  {  {
2674  /* Fast decoding an utf8 character type. TMP2 contains the first byte  /* Fast decoding a UTF-8 character type. TMP2 contains the first byte
2675  of the character (>= 0xc0) and TMP1 is destroyed. Return value in TMP1. */  of the character (>= 0xc0). Return value in TMP1. */
2676  DEFINE_COMPILER;  DEFINE_COMPILER;
2677  struct sljit_jump *jump;  struct sljit_jump *jump;
2678  struct sljit_jump *compare;  struct sljit_jump *compare;
2679    
2680  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2681    
2682  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);
2683  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_C_NOT_ZERO);
2684  /* 2 byte sequence */  /* Two byte sequence. */
2685  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2686  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2687  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);
2688  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2689  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
# Line 1566  sljit_emit_fast_return(compiler, RETURN_ Line 2698  sljit_emit_fast_return(compiler, RETURN_
2698  JUMPHERE(jump);  JUMPHERE(jump);
2699    
2700  /* We only have types for characters less than 256. */  /* We only have types for characters less than 256. */
2701  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), (sljit_w)_pcre_utf8_char_sizes - 0xc0);  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2702  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2703  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2704  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2705  }  }
2706    
2707  #endif  #elif defined COMPILE_PCRE16
2708    
2709    static void do_utfreadchar(compiler_common *common)
2710    {
2711    /* Fast decoding a UTF-16 character. TMP1 contains the first 16 bit char
2712    of the character (>= 0xd800). Return char value in TMP1, length - 1 in TMP2. */
2713    DEFINE_COMPILER;
2714    struct sljit_jump *jump;
2715    
2716    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2717    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xdc00);
2718    /* Do nothing, only return. */
2719    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2720    
2721    JUMPHERE(jump);
2722    /* Combine two 16 bit characters. */
2723    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2724    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2725    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2726    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 10);
2727    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3ff);
2728    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2729    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(1));
2730    OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2731    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2732    }
2733    
2734    #endif /* COMPILE_PCRE[8|16] */
2735    
2736    #endif /* SUPPORT_UTF */
2737    
2738  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2739    
# Line 1588  DEFINE_COMPILER; Line 2749  DEFINE_COMPILER;
2749    
2750  SLJIT_ASSERT(UCD_BLOCK_SIZE == 128 && sizeof(ucd_record) == 8);  SLJIT_ASSERT(UCD_BLOCK_SIZE == 128 && sizeof(ucd_record) == 8);
2751    
2752  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2753  OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);  OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
2754  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_w)_pcre_ucd_stage1);  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(ucd_stage1));
2755  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);
2756  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
2757  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, TMP2, 0);
2758  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, (sljit_w)_pcre_ucd_stage2);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_stage2));
2759  OP1(SLJIT_MOV_UH, TMP2, 0, SLJIT_MEM2(TMP2, TMP1), 1);  OP1(SLJIT_MOV_UH, TMP2, 0, SLJIT_MEM2(TMP2, TMP1), 1);
2760  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, (sljit_w)_pcre_ucd_records + SLJIT_OFFSETOF(ucd_record, chartype));  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_records) + SLJIT_OFFSETOF(ucd_record, chartype));
2761  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM2(TMP1, TMP2), 3);  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM2(TMP1, TMP2), 3);
2762  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2763  }  }
# Line 1610  struct sljit_label *newlinelabel = NULL; Line 2771  struct sljit_label *newlinelabel = NULL;
2771  struct sljit_jump *start;  struct sljit_jump *start;
2772  struct sljit_jump *end = NULL;  struct sljit_jump *end = NULL;
2773  struct sljit_jump *nl = NULL;  struct sljit_jump *nl = NULL;
2774  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2775  struct sljit_jump *singlebyte;  struct sljit_jump *singlechar;
2776  #endif  #endif
2777  jump_list *newline = NULL;  jump_list *newline = NULL;
2778  BOOL newlinecheck = FALSE;  BOOL newlinecheck = FALSE;
2779  BOOL readbyte = FALSE;  BOOL readuchar = FALSE;
2780    
2781  if (!(hascrorlf || firstline) && (common->nltype == NLTYPE_ANY ||  if (!(hascrorlf || firstline) && (common->nltype == NLTYPE_ANY ||
2782      common->nltype == NLTYPE_ANYCRLF || common->newline > 255))      common->nltype == NLTYPE_ANYCRLF || common->newline > 255))
# Line 1624  if (!(hascrorlf || firstline) && (common Line 2785  if (!(hascrorlf || firstline) && (common
2785  if (firstline)  if (firstline)
2786    {    {
2787    /* Search for the end of the first line. */    /* Search for the end of the first line. */
2788    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, STR_PTR, 0);    SLJIT_ASSERT(common->first_line_end != 0);
2789    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), FIRSTLINE_END, STR_END, 0);    OP1(SLJIT_MOV, TMP3, 0, STR_PTR, 0);
2790    
2791    if (common->nltype == NLTYPE_FIXED && common->newline > 255)    if (common->nltype == NLTYPE_FIXED && common->newline > 255)
2792      {      {
2793      mainloop = LABEL();      mainloop = LABEL();
2794      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2795      end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);      end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
2796      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), -1);      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(-1));
2797      OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);      OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2798      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);
2799      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);
2800      OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_LOCALS_REG), FIRSTLINE_END, STR_PTR, 0, SLJIT_IMM, 1);      JUMPHERE(end);
2801        OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2802      }      }
2803    else    else
2804      {      {
2805      end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);      end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
2806      mainloop = LABEL();      mainloop = LABEL();
2807      /* Continual stores does not cause data dependency. */      /* Continual stores does not cause data dependency. */
2808      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), FIRSTLINE_END, STR_PTR, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);
2809      read_char(common);      read_char(common);
2810      check_newlinechar(common, common->nltype, &newline, TRUE);      check_newlinechar(common, common->nltype, &newline, TRUE);
2811      CMPTO(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0, mainloop);      CMPTO(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0, mainloop);
2812      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), FIRSTLINE_END, STR_PTR, 0);      JUMPHERE(end);
2813        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);
2814      set_jumps(newline, LABEL());      set_jumps(newline, LABEL());
2815      }      }
2816    
2817    JUMPHERE(end);    OP1(SLJIT_MOV, STR_PTR, 0, TMP3, 0);
   OP1(SLJIT_MOV, STR_PTR, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0);  
2818    }    }
2819    
2820  start = JUMP(SLJIT_JUMP);  start = JUMP(SLJIT_JUMP);
# Line 1660  start = JUMP(SLJIT_JUMP); Line 2822  start = JUMP(SLJIT_JUMP);
2822  if (newlinecheck)  if (newlinecheck)
2823    {    {
2824    newlinelabel = LABEL();    newlinelabel = LABEL();
2825    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2826    end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);    end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
2827    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2828    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);
2829    COND_VALUE(SLJIT_MOV, TMP1, 0, SLJIT_C_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
2830    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2831      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, UCHAR_SHIFT);
2832    #endif
2833    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2834    nl = JUMP(SLJIT_JUMP);    nl = JUMP(SLJIT_JUMP);
2835    }    }
# Line 1672  if (newlinecheck) Line 2837  if (newlinecheck)
2837  mainloop = LABEL();  mainloop = LABEL();
2838    
2839  /* Increasing the STR_PTR here requires one less jump in the most common case. */  /* Increasing the STR_PTR here requires one less jump in the most common case. */
2840  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
2841  if (common->utf8) readbyte = TRUE;  if (common->utf) readuchar = TRUE;
2842  #endif  #endif
2843  if (newlinecheck) readbyte = TRUE;  if (newlinecheck) readuchar = TRUE;
2844    
2845  if (readbyte)  if (readuchar)
2846    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2847    
2848  if (newlinecheck)  if (newlinecheck)
2849    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);
2850    
2851  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2852  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2853  if (common->utf8)  #if defined COMPILE_PCRE8
2854    if (common->utf)
2855    {    {
2856    singlebyte = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);    singlechar = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2857    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_w)_pcre_utf8_char_sizes - 0xc0);    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2858    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2859    JUMPHERE(singlebyte);    JUMPHERE(singlechar);
2860    }    }
2861  #endif  #elif defined COMPILE_PCRE16
2862    if (common->utf)
2863      {
2864      singlechar = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);
2865      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
2866      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2867      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
2868      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
2869      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2870      JUMPHERE(singlechar);
2871      }
2872    #endif /* COMPILE_PCRE[8|16] */
2873    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2874  JUMPHERE(start);  JUMPHERE(start);
2875    
2876  if (newlinecheck)  if (newlinecheck)
# Line 1704  if (newlinecheck) Line 2882  if (newlinecheck)
2882  return mainloop;  return mainloop;
2883  }  }
2884    
2885  static SLJIT_INLINE void fast_forward_first_byte(compiler_common *common, pcre_uint16 firstbyte, BOOL firstline)  #define MAX_N_CHARS 3
2886    
2887    static SLJIT_INLINE BOOL fast_forward_first_n_chars(compiler_common *common, BOOL firstline)
2888    {
2889    DEFINE_COMPILER;
2890    struct sljit_label *start;
2891    struct sljit_jump *quit;
2892    pcre_uint32 chars[MAX_N_CHARS * 2];
2893    pcre_uchar *cc = common->start + 1 + LINK_SIZE;
2894    int location = 0;
2895    pcre_int32 len, c, bit, caseless;
2896    int must_stop;
2897    
2898    /* We do not support alternatives now. */
2899    if (*(common->start + GET(common->start, 1)) == OP_ALT)
2900      return FALSE;
2901    
2902    while (TRUE)
2903      {
2904      caseless = 0;
2905      must_stop = 1;
2906      switch(*cc)
2907        {
2908        case OP_CHAR:
2909        must_stop = 0;
2910        cc++;
2911        break;
2912    
2913        case OP_CHARI:
2914        caseless = 1;
2915        must_stop = 0;
2916        cc++;
2917        break;
2918    
2919        case OP_SOD:
2920        case OP_SOM:
2921        case OP_SET_SOM:
2922        case OP_NOT_WORD_BOUNDARY:
2923        case OP_WORD_BOUNDARY:
2924        case OP_EODN:
2925        case OP_EOD:
2926        case OP_CIRC:
2927        case OP_CIRCM:
2928        case OP_DOLL:
2929        case OP_DOLLM:
2930        /* Zero width assertions. */
2931        cc++;
2932        continue;
2933    
2934        case OP_PLUS:
2935        case OP_MINPLUS:
2936        case OP_POSPLUS:
2937        cc++;
2938        break;
2939    
2940        case OP_EXACT:
2941        cc += 1 + IMM2_SIZE;
2942        break;
2943    
2944        case OP_PLUSI:
2945        case OP_MINPLUSI:
2946        case OP_POSPLUSI:
2947        caseless = 1;
2948        cc++;
2949        break;
2950    
2951        case OP_EXACTI:
2952        caseless = 1;
2953        cc += 1 + IMM2_SIZE;
2954        break;
2955    
2956        default:
2957        must_stop = 2;
2958        break;
2959        }
2960    
2961      if (must_stop == 2)
2962          break;
2963    
2964      len = 1;
2965    #ifdef SUPPORT_UTF
2966      if (common->utf && HAS_EXTRALEN(cc[0])) len += GET_EXTRALEN(cc[0]);
2967    #endif
2968    
2969      if (caseless && char_has_othercase(common, cc))
2970        {
2971        caseless = char_get_othercase_bit(common, cc);
2972        if (caseless == 0)
2973          return FALSE;
2974    #ifdef COMPILE_PCRE8
2975        caseless = ((caseless & 0xff) << 8) | (len - (caseless >> 8));
2976    #else
2977        if ((caseless & 0x100) != 0)
2978          caseless = ((caseless & 0xff) << 16) | (len - (caseless >> 9));
2979        else
2980          caseless = ((caseless & 0xff) << 8) | (len - (caseless >> 9));
2981    #endif
2982        }
2983      else
2984        caseless = 0;
2985    
2986      while (len > 0 && location < MAX_N_CHARS * 2)
2987        {
2988        c = *cc;
2989        bit = 0;
2990        if (len == (caseless & 0xff))
2991          {
2992          bit = caseless >> 8;
2993          c |= bit;
2994          }
2995    
2996        chars[location] = c;
2997        chars[location + 1] = bit;
2998    
2999        len--;
3000        location += 2;
3001        cc++;
3002        }
3003    
3004      if (location >= MAX_N_CHARS * 2 || must_stop != 0)
3005        break;
3006      }
3007    
3008    /* At least two characters are required. */
3009    if (location < 2 * 2)
3010        return FALSE;
3011    
3012    if (firstline)
3013      {
3014      SLJIT_ASSERT(common->first_line_end != 0);
3015      OP1(SLJIT_MOV, TMP3, 0, STR_END, 0);
3016      OP2(SLJIT_SUB, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, SLJIT_IMM, IN_UCHARS((location >> 1) - 1));
3017      }
3018    else
3019      OP2(SLJIT_SUB, STR_END, 0, STR_END, 0, SLJIT_IMM, IN_UCHARS((location >> 1) - 1));
3020    
3021    start = LABEL();
3022    quit = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3023    
3024    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
3025    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
3026    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3027    if (chars[1] != 0)
3028      OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, chars[1]);
3029    CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, chars[0], start);
3030    if (location > 2 * 2)
3031      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
3032    if (chars[3] != 0)
3033      OP2(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_IMM, chars[3]);
3034    CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, chars[2], start);
3035    if (location > 2 * 2)
3036      {
3037      if (chars[5] != 0)
3038        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, chars[5]);
3039      CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, chars[4], start);
3040      }
3041    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3042    
3043    JUMPHERE(quit);
3044    
3045    if (firstline)
3046      OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);
3047    else
3048      OP2(SLJIT_ADD, STR_END, 0, STR_END, 0, SLJIT_IMM, IN_UCHARS((location >> 1) - 1));
3049    return TRUE;
3050    }
3051    
3052    #undef MAX_N_CHARS
3053    
3054    static SLJIT_INLINE void fast_forward_first_char(compiler_common *common, pcre_uchar first_char, BOOL caseless, BOOL firstline)
3055  {  {
3056  DEFINE_COMPILER;  DEFINE_COMPILER;
3057  struct sljit_label *start;  struct sljit_label *start;
3058  struct sljit_jump *leave;  struct sljit_jump *quit;
3059  struct sljit_jump *found;  struct sljit_jump *found;
3060  pcre_uint16 oc, bit;  pcre_uchar oc, bit;
3061    
3062  if (firstline)  if (firstline)
3063    {    {
3064    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), POSSESSIVE0, STR_END, 0);    SLJIT_ASSERT(common->first_line_end != 0);
3065    OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), FIRSTLINE_END);    OP1(SLJIT_MOV, TMP3, 0, STR_END, 0);
3066      OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end);
3067    }    }
3068    
3069  start = LABEL();  start = LABEL();
3070  leave = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);  quit = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3071  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3072    
3073  if ((firstbyte & REQ_CASELESS) == 0)  oc = first_char;
3074    found = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, firstbyte & 0xff);  if (caseless)
3075      {
3076      oc = TABLE_GET(first_char, common->fcc, first_char);
3077    #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8)
3078      if (first_char > 127 && common->utf)
3079        oc = UCD_OTHERCASE(first_char);
3080    #endif
3081      }
3082    if (first_char == oc)
3083      found = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, first_char);
3084  else  else
3085    {    {
3086    firstbyte &= 0xff;    bit = first_char ^ oc;
3087    oc = common->fcc[firstbyte];    if (is_powerof2(bit))
   bit = firstbyte ^ oc;  
   if (ispowerof2(bit))  
3088      {      {
3089      OP2(SLJIT_OR, TMP2, 0, TMP1, 0, SLJIT_IMM, bit);      OP2(SLJIT_OR, TMP2, 0, TMP1, 0, SLJIT_IMM, bit);
3090      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, firstbyte | bit);      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, first_char | bit);
3091      }      }
3092    else    else
3093      {      {
3094      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, firstbyte);      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, first_char);
3095      COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_EQUAL);      OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
3096      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, oc);      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, oc);
3097      COND_VALUE(SLJIT_OR | SLJIT_SET_E, TMP2, 0, SLJIT_C_EQUAL);      OP_FLAGS(SLJIT_OR | SLJIT_SET_E, TMP2, 0, TMP2, 0, SLJIT_C_EQUAL);
3098      found = JUMP(SLJIT_C_NOT_ZERO);      found = JUMP(SLJIT_C_NOT_ZERO);
3099      }      }
3100    }    }
3101    
3102  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
 #ifdef SUPPORT_UTF8  
 if (common->utf8)  
   {  
   CMPTO(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0, start);  
   OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_w)_pcre_utf8_char_sizes - 0xc0);  
   OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);  
   }  
 #endif  
3103  JUMPTO(SLJIT_JUMP, start);  JUMPTO(SLJIT_JUMP, start);
3104  JUMPHERE(found);  JUMPHERE(found);
3105  JUMPHERE(leave);  JUMPHERE(quit);
3106    
3107  if (firstline)  if (firstline)
3108    OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), POSSESSIVE0);    OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);
3109  }  }
3110    
3111  static SLJIT_INLINE void fast_forward_newline(compiler_common *common, BOOL firstline)  static SLJIT_INLINE void fast_forward_newline(compiler_common *common, BOOL firstline)
# Line 1767  DEFINE_COMPILER; Line 3114  DEFINE_COMPILER;
3114  struct sljit_label *loop;  struct sljit_label *loop;
3115  struct sljit_jump *lastchar;  struct sljit_jump *lastchar;
3116  struct sljit_jump *firstchar;  struct sljit_jump *firstchar;
3117  struct sljit_jump *leave;  struct sljit_jump *quit;
3118  struct sljit_jump *foundcr = NULL;  struct sljit_jump *foundcr = NULL;
3119  struct sljit_jump *notfoundnl;  struct sljit_jump *notfoundnl;
3120  jump_list *newline = NULL;  jump_list *newline = NULL;
3121    
3122  if (firstline)  if (firstline)
3123    {    {
3124    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), POSSESSIVE0, STR_END, 0);    SLJIT_ASSERT(common->first_line_end != 0);
3125    OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), FIRSTLINE_END);    OP1(SLJIT_MOV, TMP3, 0, STR_END, 0);
3126      OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end);
3127    }    }
3128    
3129  if (common->nltype == NLTYPE_FIXED && common->newline > 255)  if (common->nltype == NLTYPE_FIXED && common->newline > 255)
# Line 1786  if (common->nltype == NLTYPE_FIXED && co Line 3134  if (common->nltype == NLTYPE_FIXED && co
3134    OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(TMP1), SLJIT_OFFSETOF(jit_arguments, begin));    OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(TMP1), SLJIT_OFFSETOF(jit_arguments, begin));
3135    firstchar = CMP(SLJIT_C_LESS_EQUAL, STR_PTR, 0, TMP2, 0);    firstchar = CMP(SLJIT_C_LESS_EQUAL, STR_PTR, 0, TMP2, 0);
3136    
3137    OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 2);    OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, IN_UCHARS(2));
3138    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, STR_PTR, 0, TMP1, 0);
3139    COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_GREATER_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_GREATER_EQUAL);
3140    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
3141      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCHAR_SHIFT);
3142    #endif
3143    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
3144    
3145    loop = LABEL();    loop = LABEL();
3146    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3147    leave = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);    quit = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3148    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), -2);    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(-2));
3149    OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(STR_PTR), -1);    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(-1));
3150    CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, loop);    CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, loop);
3151    CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, loop);    CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, loop);
3152    
3153    JUMPHERE(leave);    JUMPHERE(quit);
3154    JUMPHERE(firstchar);    JUMPHERE(firstchar);
3155    JUMPHERE(lastchar);    JUMPHERE(lastchar);
3156    
# Line 1823  set_jumps(newline, loop); Line 3174  set_jumps(newline, loop);
3174    
3175  if (common->nltype == NLTYPE_ANY || common->nltype == NLTYPE_ANYCRLF)  if (common->nltype == NLTYPE_ANY || common->nltype == NLTYPE_ANYCRLF)
3176    {    {
3177    leave = JUMP(SLJIT_JUMP);    quit = JUMP(SLJIT_JUMP);
3178    JUMPHERE(foundcr);    JUMPHERE(foundcr);
3179    notfoundnl = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);    notfoundnl = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3180    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3181    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_NL);    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_NL);
3182    COND_VALUE(SLJIT_MOV, TMP1, 0, SLJIT_C_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
3183    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
3184      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, UCHAR_SHIFT);
3185    #endif
3186    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3187    JUMPHERE(notfoundnl);    JUMPHERE(notfoundnl);
3188    JUMPHERE(leave);    JUMPHERE(quit);
3189    }    }
3190  JUMPHERE(lastchar);  JUMPHERE(lastchar);
3191  JUMPHERE(firstchar);  JUMPHERE(firstchar);
3192    
3193  if (firstline)  if (firstline)
3194    OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), POSSESSIVE0);    OP1(SLJIT_MOV, STR_END, 0, TMP3, 0);
3195  }  }
3196    
3197    static BOOL check_class_ranges(compiler_common *common, const pcre_uint8 *bits, BOOL nclass, jump_list **backtracks);
3198    
3199  static SLJIT_INLINE void fast_forward_start_bits(compiler_common *common, sljit_uw start_bits, BOOL firstline)  static SLJIT_INLINE void fast_forward_start_bits(compiler_common *common, sljit_uw start_bits, BOOL firstline)
3200  {  {
3201  DEFINE_COMPILER;  DEFINE_COMPILER;
3202  struct sljit_label *start;  struct sljit_label *start;
3203  struct sljit_jump *leave;  struct sljit_jump *quit;
3204  struct sljit_jump *found;  struct sljit_jump *found = NULL;
3205    jump_list *matches = NULL;
3206    pcre_uint8 inverted_start_bits[32];
3207    int i;
3208    #ifndef COMPILE_PCRE8
3209    struct sljit_jump *jump;
3210    #endif
3211    
3212    for (i = 0; i < 32; ++i)
3213      inverted_start_bits[i] = ~(((pcre_uint8*)start_bits)[i]);
3214    
3215  if (firstline)  if (firstline)
3216    {    {
3217    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), POSSESSIVE0, STR_END, 0);    SLJIT_ASSERT(common->first_line_end != 0);
3218    OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), FIRSTLINE_END);    OP1(SLJIT_MOV, RETURN_ADDR, 0, STR_END, 0);
3219      OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end);
3220    }    }
3221    
3222  start = LABEL();  start = LABEL();
3223  leave = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);  quit = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3224  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3225  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3226  if (common->utf8)  if (common->utf)
3227    OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);    OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);
3228  #endif  #endif
 OP2(SLJIT_AND, TMP2, 0, TMP1, 0, SLJIT_IMM, 0x7);  
 OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 3);  
 OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), start_bits);  
 OP2(SLJIT_SHL, TMP2, 0, SLJIT_IMM, 1, TMP2, 0);  
 OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);  
 found = JUMP(SLJIT_C_NOT_ZERO);  
3229    
3230  #ifdef SUPPORT_UTF8  if (!check_class_ranges(common, inverted_start_bits, (inverted_start_bits[31] & 0x80) != 0, &matches))
3231  if (common->utf8)    {
3232    #ifndef COMPILE_PCRE8
3233      jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 255);
3234      OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 255);
3235      JUMPHERE(jump);
3236    #endif
3237      OP2(SLJIT_AND, TMP2, 0, TMP1, 0, SLJIT_IMM, 0x7);
3238      OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 3);
3239      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), start_bits);
3240      OP2(SLJIT_SHL, TMP2, 0, SLJIT_IMM, 1, TMP2, 0);
3241      OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, TMP2, 0);
3242      found = JUMP(SLJIT_C_NOT_ZERO);
3243      }
3244    
3245    #ifdef SUPPORT_UTF
3246    if (common->utf)
3247    OP1(SLJIT_MOV, TMP1, 0, TMP3, 0);    OP1(SLJIT_MOV, TMP1, 0, TMP3, 0);
3248  #endif  #endif
3249  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3250  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
3251  if (common->utf8)  #if defined COMPILE_PCRE8
3252    if (common->utf)
3253    {    {
3254    CMPTO(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0, start);    CMPTO(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0, start);
3255    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_w)_pcre_utf8_char_sizes - 0xc0);    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
3256    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3257    }    }
3258  #endif  #elif defined COMPILE_PCRE16
3259    if (common->utf)
3260      {
3261      CMPTO(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800, start);
3262      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
3263      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);
3264      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
3265      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
3266      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3267      }
3268    #endif /* COMPILE_PCRE[8|16] */
3269    #endif /* SUPPORT_UTF */
3270  JUMPTO(SLJIT_JUMP, start);  JUMPTO(SLJIT_JUMP, start);
3271  JUMPHERE(found);  if (found != NULL)
3272  JUMPHERE(leave);    JUMPHERE(found);
3273    if (matches != NULL)
3274      set_jumps(matches, LABEL());
3275    JUMPHERE(quit);
3276    
3277  if (firstline)  if (firstline)
3278    OP1(SLJIT_MOV, STR_END, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), POSSESSIVE0);    OP1(SLJIT_MOV, STR_END, 0, RETURN_ADDR, 0);
3279  }  }
3280    
3281  static SLJIT_INLINE struct sljit_jump *search_requested_char(compiler_common *common, pcre_uint16 reqbyte, BOOL has_firstbyte)  static SLJIT_INLINE struct sljit_jump *search_requested_char(compiler_common *common, pcre_uchar req_char, BOOL caseless, BOOL has_firstchar)
3282  {  {
3283  DEFINE_COMPILER;  DEFINE_COMPILER;
3284  struct sljit_label *loop;  struct sljit_label *loop;
# Line 1897  struct sljit_jump *alreadyfound; Line 3287  struct sljit_jump *alreadyfound;
3287  struct sljit_jump *found;  struct sljit_jump *found;
3288  struct sljit_jump *foundoc = NULL;  struct sljit_jump *foundoc = NULL;
3289  struct sljit_jump *notfound;  struct sljit_jump *notfound;
3290  pcre_uint16 oc, bit;  pcre_uint32 oc, bit;
3291    
3292  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), REQ_BYTE_PTR);  SLJIT_ASSERT(common->req_char_ptr != 0);
3293    OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->req_char_ptr);
3294  OP2(SLJIT_ADD, TMP1, 0, STR_PTR, 0, SLJIT_IMM, REQ_BYTE_MAX);  OP2(SLJIT_ADD, TMP1, 0, STR_PTR, 0, SLJIT_IMM, REQ_BYTE_MAX);
3295  toolong = CMP(SLJIT_C_LESS, TMP1, 0, STR_END, 0);  toolong = CMP(SLJIT_C_LESS, TMP1, 0, STR_END, 0);
3296  alreadyfound = CMP(SLJIT_C_LESS, STR_PTR, 0, TMP2, 0);  alreadyfound = CMP(SLJIT_C_LESS, STR_PTR, 0, TMP2, 0);
3297    
3298  if (has_firstbyte)  if (has_firstchar)
3299    OP2(SLJIT_ADD, TMP1, 0, STR_PTR, 0, SLJIT_IMM, 1);    OP2(SLJIT_ADD, TMP1, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3300  else  else
3301    OP1(SLJIT_MOV, TMP1, 0, STR_PTR, 0);    OP1(SLJIT_MOV, TMP1, 0, STR_PTR, 0);
3302    
3303  loop = LABEL();  loop = LABEL();
3304  notfound = CMP(SLJIT_C_GREATER_EQUAL, TMP1, 0, STR_END, 0);  notfound = CMP(SLJIT_C_GREATER_EQUAL, TMP1, 0, STR_END, 0);
3305    
3306  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP1), 0);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(TMP1), 0);
3307  if ((reqbyte & REQ_CASELESS) == 0)  oc = req_char;
3308    found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, reqbyte & 0xff);  if (caseless)
3309      {
3310      oc = TABLE_GET(req_char, common->fcc, req_char);
3311    #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8)
3312      if (req_char > 127 && common->utf)
3313        oc = UCD_OTHERCASE(req_char);
3314    #endif
3315      }
3316    if (req_char == oc)
3317      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, req_char);
3318  else  else
3319    {    {
3320    reqbyte &= 0xff;    bit = req_char ^ oc;
3321    oc = common->fcc[reqbyte];    if (is_powerof2(bit))
   bit = reqbyte ^ oc;  
   if (ispowerof2(bit))  
3322      {      {
3323      OP2(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_IMM, bit);      OP2(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_IMM, bit);
3324      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, reqbyte | bit);      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, req_char | bit);
3325      }      }
3326    else    else
3327      {      {
3328      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, reqbyte);      found = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, req_char);
3329      foundoc = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, oc);      foundoc = CMP(SLJIT_C_EQUAL, TMP2, 0, SLJIT_IMM, oc);
3330      }      }
3331    }    }
3332  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, IN_UCHARS(1));
3333  JUMPTO(SLJIT_JUMP, loop);  JUMPTO(SLJIT_JUMP, loop);
3334    
3335  JUMPHERE(found);  JUMPHERE(found);
3336  if (foundoc)  if (foundoc)
3337    JUMPHERE(foundoc);    JUMPHERE(foundoc);
3338  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), REQ_BYTE_PTR, TMP1, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->req_char_ptr, TMP1, 0);
3339  JUMPHERE(alreadyfound);  JUMPHERE(alreadyfound);
3340  JUMPHERE(toolong);  JUMPHERE(toolong);
3341  return notfound;  return notfound;
# Line 1949  DEFINE_COMPILER; Line 3347  DEFINE_COMPILER;
3347  struct sljit_jump *jump;  struct sljit_jump *jump;
3348  struct sljit_label *mainloop;  struct sljit_label *mainloop;
3349    
3350  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
3351  OP1(SLJIT_MOV, TMP1, 0, STACK_TOP, 0);  OP1(SLJIT_MOV, TMP1, 0, STACK_TOP, 0);
3352    GET_LOCAL_BASE(TMP3, 0, 0);
3353    
3354  /* Drop frames until we reach STACK_TOP. */  /* Drop frames until we reach STACK_TOP. */
3355  mainloop = LABEL();  mainloop = LABEL();
3356  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(TMP1), 0);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(TMP1), 0);
3357  jump = CMP(SLJIT_C_SIG_LESS_EQUAL, TMP2, 0, SLJIT_IMM, frame_end);  OP2(SLJIT_SUB | SLJIT_SET_S, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0);
3358  OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_LOCALS_REG, 0);  jump = JUMP(SLJIT_C_SIG_LESS_EQUAL);
3359  OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), 0, SLJIT_MEM1(TMP1), sizeof(sljit_w));  
3360  OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), sizeof(sljit_w), SLJIT_MEM1(TMP1), 2 * sizeof(sljit_w));  OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, TMP3, 0);
3361  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 3 * sizeof(sljit_w));  OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), 0, SLJIT_MEM1(TMP1), sizeof(sljit_sw));
3362    OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), sizeof(sljit_sw), SLJIT_MEM1(TMP1), 2 * sizeof(sljit_sw));
3363    OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 3 * sizeof(sljit_sw));
3364  JUMPTO(SLJIT_JUMP, mainloop);  JUMPTO(SLJIT_JUMP, mainloop);
3365    
3366  JUMPHERE(jump);  JUMPHERE(jump);
3367  jump = CMP(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, frame_end);  jump = JUMP(SLJIT_C_SIG_LESS);
3368  /* End of dropping frames. */  /* End of dropping frames. */
3369  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3370    
3371  JUMPHERE(jump);  JUMPHERE(jump);
3372  jump = CMP(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, frame_setstrbegin);  OP1(SLJIT_NEG, TMP2, 0, TMP2, 0);
3373  /* Set string begin. */  OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, TMP3, 0);
3374  OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(TMP1), sizeof(sljit_w));  OP1(SLJIT_MOV, SLJIT_MEM1(TMP2), 0, SLJIT_MEM1(TMP1), sizeof(sljit_sw));
3375  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 2 * sizeof(sljit_w));  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 2 * sizeof(sljit_sw));
 OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0), TMP2, 0);  
 JUMPTO(SLJIT_JUMP, mainloop);  
   
 JUMPHERE(jump);  
 /* Unknown command. */  
 OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 2 * sizeof(sljit_w));  
3376  JUMPTO(SLJIT_JUMP, mainloop);  JUMPTO(SLJIT_JUMP, mainloop);
3377  }  }
3378    
3379  static void check_wordboundary(compiler_common *common)  static void check_wordboundary(compiler_common *common)
3380  {  {
3381  DEFINE_COMPILER;  DEFINE_COMPILER;
3382  struct sljit_jump *beginend;  struct sljit_jump *skipread;
3383  #ifdef SUPPORT_UTF8  jump_list *skipread_list = NULL;
3384    #if !(defined COMPILE_PCRE8) || defined SUPPORT_UTF
3385  struct sljit_jump *jump;  struct sljit_jump *jump;
3386  #endif  #endif
3387    
3388  SLJIT_COMPILE_ASSERT(ctype_word == 0x10, ctype_word_must_be_16);  SLJIT_COMPILE_ASSERT(ctype_word == 0x10, ctype_word_must_be_16);
3389    
3390  sljit_emit_fast_enter(compiler, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, 1, 5, 5, common->localsize);  sljit_emit_fast_enter(compiler, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0);
3391  /* Get type of the previous char, and put it to LOCALS1. */  /* Get type of the previous char, and put it to LOCALS1. */
3392  OP1(SLJIT_MOV, TMP1, 0, ARGUMENTS, 0);  OP1(SLJIT_MOV, TMP1, 0, ARGUMENTS, 0);
3393  OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(TMP1), SLJIT_OFFSETOF(jit_arguments, begin));  OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(TMP1), SLJIT_OFFSETOF(jit_arguments, begin));
3394  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, SLJIT_IMM, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, SLJIT_IMM, 0);
3395  beginend = CMP(SLJIT_C_LESS_EQUAL, STR_PTR, 0, TMP1, 0);  skipread = CMP(SLJIT_C_LESS_EQUAL, STR_PTR, 0, TMP1, 0);
3396  skip_char_back(common);  skip_char_back(common);
3397    check_start_used_ptr(common);
3398  read_char(common);  read_char(common);
3399    
3400  /* Testing char type. */  /* Testing char type. */
3401  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3402  if (common->useucp)  if (common->use_ucp)
3403    {    {
3404    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 1);    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 1);
3405    jump = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_UNDERSCORE);    jump = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_UNDERSCORE);
3406    add_jump(compiler, &common->getucd, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->getucd, JUMP(SLJIT_FAST_CALL));
3407    OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Ll);    OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Ll);
3408    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_Lu - ucp_Ll);    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_Lu - ucp_Ll);
3409    COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_LESS_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_LESS_EQUAL);
3410    OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Nd - ucp_Ll);    OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Nd - ucp_Ll);
3411    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_No - ucp_Nd);    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_No - ucp_Nd);
3412    COND_VALUE(SLJIT_OR, TMP2, 0, SLJIT_C_LESS_EQUAL);    OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_C_LESS_EQUAL);
3413    JUMPHERE(jump);    JUMPHERE(jump);
3414    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP2, 0);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP2, 0);
3415    }    }
3416  else  else
3417  #endif  #endif
3418    {    {
3419  #ifdef SUPPORT_UTF8  #ifndef COMPILE_PCRE8
3420      jump = CMP(SLJIT_C_GREATER, TMP1, 0, SLJIT_IMM, 255);
3421    #elif defined SUPPORT_UTF
3422    /* Here LOCALS1 has already been zeroed. */    /* Here LOCALS1 has already been zeroed. */
3423    jump = NULL;    jump = NULL;
3424    if (common->utf8)    if (common->utf)
3425      jump = CMP(SLJIT_C_GREATER, TMP1, 0, SLJIT_IMM, 255);      jump = CMP(SLJIT_C_GREATER, TMP1, 0, SLJIT_IMM, 255);
3426  #endif  #endif /* COMPILE_PCRE8 */
3427    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), common->ctypes);    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), common->ctypes);
3428    OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 4 /* ctype_word */);    OP2(SLJIT_LSHR, TMP1, 0, TMP1, 0, SLJIT_IMM, 4 /* ctype_word */);
3429    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
3430    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);
3431  #ifdef SUPPORT_UTF8  #ifndef COMPILE_PCRE8
3432      JUMPHERE(jump);
3433    #elif defined SUPPORT_UTF
3434    if (jump != NULL)    if (jump != NULL)
3435      JUMPHERE(jump);      JUMPHERE(jump);
3436  #endif  #endif /* COMPILE_PCRE8 */
3437    }    }
3438  JUMPHERE(beginend);  JUMPHERE(skipread);
3439    
3440  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 0);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 0);
3441  beginend = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);  check_str_end(common, &skipread_list);
3442  peek_char(common);  peek_char(common);
3443    
3444  /* Testing char type. This is a code duplication. */  /* Testing char type. This is a code duplication. */
3445  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3446  if (common->useucp)  if (common->use_ucp)
3447    {    {
3448    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 1);    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 1);
3449    jump = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_UNDERSCORE);    jump = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_UNDERSCORE);
3450    add_jump(compiler, &common->getucd, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->getucd, JUMP(SLJIT_FAST_CALL));
3451    OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Ll);    OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Ll);
3452    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_Lu - ucp_Ll);    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_Lu - ucp_Ll);
3453    COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_LESS_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_LESS_EQUAL);
3454    OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Nd - ucp_Ll);    OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ucp_Nd - ucp_Ll);
3455    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_No - ucp_Nd);    OP2(SLJIT_SUB | SLJIT_SET_U, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, ucp_No - ucp_Nd);
3456    COND_VALUE(SLJIT_OR, TMP2, 0, SLJIT_C_LESS_EQUAL);    OP_FLAGS(SLJIT_OR, TMP2, 0, TMP2, 0, SLJIT_C_LESS_EQUAL);
3457    JUMPHERE(jump);    JUMPHERE(jump);
3458    }    }
3459  else  else
3460  #endif  #endif
3461    {    {
3462  #ifdef SUPPORT_UTF8  #ifndef COMPILE_PCRE8
3463      /* TMP2 may be destroyed by peek_char. */
3464      OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 0);
3465      jump = CMP(SLJIT_C_GREATER, TMP1, 0, SLJIT_IMM, 255);
3466    #elif defined SUPPORT_UTF
3467    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 0);    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, 0);
3468    jump = NULL;    jump = NULL;
3469    if (common->utf8)    if (common->utf)
3470      jump = CMP(SLJIT_C_GREATER, TMP1, 0, SLJIT_IMM, 255);      jump = CMP(SLJIT_C_GREATER, TMP1, 0, SLJIT_IMM, 255);
3471  #endif  #endif
3472    OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP1), common->ctypes);    OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP1), common->ctypes);
3473    OP2(SLJIT_LSHR, TMP2, 0, TMP2, 0, SLJIT_IMM, 4 /* ctype_word */);    OP2(SLJIT_LSHR, TMP2, 0, TMP2, 0, SLJIT_IMM, 4 /* ctype_word */);
3474    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 1);    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 1);
3475  #ifdef SUPPORT_UTF8  #ifndef COMPILE_PCRE8
3476      JUMPHERE(jump);
3477    #elif defined SUPPORT_UTF
3478    if (jump != NULL)    if (jump != NULL)
3479      JUMPHERE(jump);      JUMPHERE(jump);
3480  #endif  #endif /* COMPILE_PCRE8 */
3481    }    }
3482  JUMPHERE(beginend);  set_jumps(skipread_list, LABEL());
3483    
3484  OP2(SLJIT_XOR | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1);  OP2(SLJIT_XOR | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1);
3485  sljit_emit_fast_return(compiler, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0);  sljit_emit_fast_return(compiler, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0);
3486  }  }
3487    
3488    /*
3489      range format:
3490    
3491      ranges[0] = length of the range (max MAX_RANGE_SIZE, -1 means invalid range).
3492      ranges[1] = first bit (0 or 1)
3493      ranges[2-length] = position of the bit change (when the current bit is not equal to the previous)
3494    */
3495    
3496    static BOOL check_ranges(compiler_common *common, int *ranges, jump_list **backtracks, BOOL readch)
3497    {
3498    DEFINE_COMPILER;
3499    struct sljit_jump *jump;
3500    
3501    if (ranges[0] < 0)
3502      return FALSE;
3503    
3504    switch(ranges[0])
3505      {
3506      case 1:
3507      if (readch)
3508        read_char(common);
3509      add_jump(compiler, backtracks, CMP(ranges[1] == 0 ? SLJIT_C_LESS : SLJIT_C_GREATER_EQUAL, TMP1, 0, SLJIT_IMM, ranges[2]));
3510      return TRUE;
3511    
3512      case 2:
3513      if (readch)
3514        read_char(common);
3515      OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[2]);
3516      add_jump(compiler, backtracks, CMP(ranges[1] != 0 ? SLJIT_C_LESS : SLJIT_C_GREATER_EQUAL, TMP1, 0, SLJIT_IMM, ranges[3] - ranges[2]));
3517      return TRUE;
3518    
3519      case 4:
3520      if (ranges[2] + 1 == ranges[3] && ranges[4] + 1 == ranges[5])
3521        {
3522        if (readch)
3523          read_char(common);
3524        if (ranges[1] != 0)
3525          {
3526          add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, ranges[2]));
3527          add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, ranges[4]));
3528          }
3529        else
3530          {
3531          jump = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, ranges[2]);
3532          add_jump(compiler, backtracks, CMP(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, ranges[4]));
3533          JUMPHERE(jump);
3534          }
3535        return TRUE;
3536        }
3537      if ((ranges[3] - ranges[2]) == (ranges[5] - ranges[4]) && is_powerof2(ranges[4] - ranges[2]))
3538        {
3539        if (readch)
3540          read_char(common);
3541        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[4] - ranges[2]);
3542        OP2(SLJIT_SUB, TMP1, 0, TMP1, 0, SLJIT_IMM, ranges[4]);
3543        add_jump(compiler, backtracks, CMP(ranges[1] != 0 ? SLJIT_C_LESS : SLJIT_C_GREATER_EQUAL, TMP1, 0, SLJIT_IMM, ranges[5] - ranges[4]));
3544