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

Diff of /code/trunk/pcre_jit_compile.c

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

revision 941 by zherczeg, Tue Feb 28 11:33:34 2012 UTC revision 1530 by zherczeg, Thu Mar 5 08:53:37 2015 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2012 University of Cambridge             Copyright (c) 1997-2013 University of Cambridge
10    
11    The machine code generator part (this module) was written by Zoltan Herczeg    The machine code generator part (this module) was written by Zoltan Herczeg
12                        Copyright (c) 2010-2012                        Copyright (c) 2010-2013
13    
14  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
15  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 46  POSSIBILITY OF SUCH DAMAGE. Line 46  POSSIBILITY OF SUCH DAMAGE.
46    
47  #include "pcre_internal.h"  #include "pcre_internal.h"
48    
49  #ifdef SUPPORT_JIT  #if defined SUPPORT_JIT
50    
51  /* All-in-one: Since we use the JIT compiler only from here,  /* All-in-one: Since we use the JIT compiler only from here,
52  we just include it. This way we don't need to touch the build  we just include it. This way we don't need to touch the build
53  system files. */  system files. */
54    
55  #define SLJIT_MALLOC(size) (PUBL(malloc))(size)  #define SLJIT_MALLOC(size, allocator_data) (PUBL(malloc))(size)
56  #define SLJIT_FREE(ptr) (PUBL(free))(ptr)  #define SLJIT_FREE(ptr, allocator_data) (PUBL(free))(ptr)
57  #define SLJIT_CONFIG_AUTO 1  #define SLJIT_CONFIG_AUTO 1
58  #define SLJIT_CONFIG_STATIC 1  #define SLJIT_CONFIG_STATIC 1
59  #define SLJIT_VERBOSE 0  #define SLJIT_VERBOSE 0
# Line 65  system files. */ Line 65  system files. */
65  #error Unsupported architecture  #error Unsupported architecture
66  #endif  #endif
67    
68  /* Allocate memory on the stack. Fast, but limited size. */  /* Defines for debugging purposes. */
 #define LOCAL_SPACE_SIZE 32768  
69    
70    /* 1 - Use unoptimized capturing brackets.
71       2 - Enable capture_last_ptr (includes option 1). */
72    /* #define DEBUG_FORCE_UNOPTIMIZED_CBRAS 2 */
73    
74    /* 1 - Always have a control head. */
75    /* #define DEBUG_FORCE_CONTROL_HEAD 1 */
76    
77    /* Allocate memory for the regex stack on the real machine stack.
78    Fast, but limited size. */
79    #define MACHINE_STACK_SIZE 32768
80    
81    /* Growth rate for stack allocated by the OS. Should be the multiply
82    of page size. */
83  #define STACK_GROWTH_RATE 8192  #define STACK_GROWTH_RATE 8192
84    
85  /* Enable to check that the allocation could destroy temporaries. */  /* Enable to check that the allocation could destroy temporaries. */
# Line 82  The code generator follows the recursive Line 94  The code generator follows the recursive
94  expressions. The basic blocks of regular expressions are condition checkers  expressions. The basic blocks of regular expressions are condition checkers
95  whose execute different commands depending on the result of the condition check.  whose execute different commands depending on the result of the condition check.
96  The relationship between the operators can be horizontal (concatenation) and  The relationship between the operators can be horizontal (concatenation) and
97  vertical (sub-expression) (See struct fallback_common for more details).  vertical (sub-expression) (See struct backtrack_common for more details).
98    
99    'ab' - 'a' and 'b' regexps are concatenated    'ab' - 'a' and 'b' regexps are concatenated
100    'a+' - 'a' is the sub-expression of the '+' operator    'a+' - 'a' is the sub-expression of the '+' operator
101    
102  The condition checkers are boolean (true/false) checkers. Machine code is generated  The condition checkers are boolean (true/false) checkers. Machine code is generated
103  for the checker itself and for the actions depending on the result of the checker.  for the checker itself and for the actions depending on the result of the checker.
104  The 'true' case is called as the hot path (expected path), and the other is called as  The 'true' case is called as the matching path (expected path), and the other is called as
105  the 'fallback' path. Branch instructions are expesive for all CPUs, so we avoid taken  the 'backtrack' path. Branch instructions are expesive for all CPUs, so we avoid taken
106  branches on the hot path.  branches on the matching path.
107    
108   Greedy star operator (*) :   Greedy star operator (*) :
109     Hot path: match happens.     Matching path: match happens.
110     Fallback path: match failed.     Backtrack path: match failed.
111   Non-greedy star operator (*?) :   Non-greedy star operator (*?) :
112     Hot path: no need to perform a match.     Matching path: no need to perform a match.
113     Fallback path: match is required.     Backtrack path: match is required.
114    
115  The following example shows how the code generated for a capturing bracket  The following example shows how the code generated for a capturing bracket
116  with two alternatives. Let A, B, C, D are arbirary regular expressions, and  with two alternatives. Let A, B, C, D are arbirary regular expressions, and
# Line 108  we have the following regular expression Line 120  we have the following regular expression
120    
121  The generated code will be the following:  The generated code will be the following:
122    
123   A hot path   A matching path
124   '(' hot path (pushing arguments to the stack)   '(' matching path (pushing arguments to the stack)
125   B hot path   B matching path
126   ')' hot path (pushing arguments to the stack)   ')' matching path (pushing arguments to the stack)
127   D hot path   D matching path
128   return with successful match   return with successful match
129    
130   D fallback path   D backtrack path
131   ')' fallback path (If we arrived from "C" jump to the fallback of "C")   ')' backtrack path (If we arrived from "C" jump to the backtrack of "C")
132   B fallback path   B backtrack path
133   C expected path   C expected path
134   jump to D hot path   jump to D matching path
135   C fallback path   C backtrack path
136   A fallback path   A backtrack path
137    
138   Notice, that the order of fallback code paths are the opposite of the fast   Notice, that the order of backtrack code paths are the opposite of the fast
139   code paths. In this way the topmost value on the stack is always belong   code paths. In this way the topmost value on the stack is always belong
140   to the current fallback code path. The fallback code path must check   to the current backtrack code path. The backtrack path must check
141   whether there is a next alternative. If so, it needs to jump back to   whether there is a next alternative. If so, it needs to jump back to
142   the hot path eventually. Otherwise it needs to clear out its own stack   the matching path eventually. Otherwise it needs to clear out its own stack
143   frame and continue the execution on the fallback code paths.   frame and continue the execution on the backtrack code paths.
144  */  */
145    
146  /*  /*
147  Saved stack frames:  Saved stack frames:
148    
149  Atomic blocks and asserts require reloading the values of local variables  Atomic blocks and asserts require reloading the values of private data
150  when the fallback mechanism performed. Because of OP_RECURSE, the locals  when the backtrack mechanism performed. Because of OP_RECURSE, the data
151  are not necessarly known in compile time, thus we need a dynamic restore  are not necessarly known in compile time, thus we need a dynamic restore
152  mechanism.  mechanism.
153    
154  The stack frames are stored in a chain list, and have the following format:  The stack frames are stored in a chain list, and have the following format:
155  ([ capturing bracket offset ][ start value ][ end value ])+ ... [ 0 ] [ previous head ]  ([ capturing bracket offset ][ start value ][ end value ])+ ... [ 0 ] [ previous head ]
156    
157  Thus we can restore the locals to a particular point in the stack.  Thus we can restore the private data to a particular point in the stack.
158  */  */
159    
160  typedef struct jit_arguments {  typedef struct jit_arguments {
# Line 154  typedef struct jit_arguments { Line 166  typedef struct jit_arguments {
166    int *offsets;    int *offsets;
167    pcre_uchar *uchar_ptr;    pcre_uchar *uchar_ptr;
168    pcre_uchar *mark_ptr;    pcre_uchar *mark_ptr;
169      void *callout_data;
170    /* Everything else after. */    /* Everything else after. */
171    int offsetcount;    pcre_uint32 limit_match;
172    int calllimit;    int real_offset_count;
173      int offset_count;
174    pcre_uint8 notbol;    pcre_uint8 notbol;
175    pcre_uint8 noteol;    pcre_uint8 noteol;
176    pcre_uint8 notempty;    pcre_uint8 notempty;
# Line 165  typedef struct jit_arguments { Line 179  typedef struct jit_arguments {
179    
180  typedef struct executable_functions {  typedef struct executable_functions {
181    void *executable_funcs[JIT_NUMBER_OF_COMPILE_MODES];    void *executable_funcs[JIT_NUMBER_OF_COMPILE_MODES];
182      void *read_only_data_heads[JIT_NUMBER_OF_COMPILE_MODES];
183      sljit_uw executable_sizes[JIT_NUMBER_OF_COMPILE_MODES];
184    PUBL(jit_callback) callback;    PUBL(jit_callback) callback;
185    void *userdata;    void *userdata;
186    sljit_uw executable_sizes[JIT_NUMBER_OF_COMPILE_MODES];    pcre_uint32 top_bracket;
187      pcre_uint32 limit_match;
188  } executable_functions;  } executable_functions;
189    
190  typedef struct jump_list {  typedef struct jump_list {
# Line 175  typedef struct jump_list { Line 192  typedef struct jump_list {
192    struct jump_list *next;    struct jump_list *next;
193  } jump_list;  } jump_list;
194    
 enum stub_types { stack_alloc };  
   
195  typedef struct stub_list {  typedef struct stub_list {
   enum stub_types type;  
   int data;  
196    struct sljit_jump *start;    struct sljit_jump *start;
197    struct sljit_label *leave;    struct sljit_label *quit;
198    struct stub_list *next;    struct stub_list *next;
199  } stub_list;  } stub_list;
200    
201    typedef struct label_addr_list {
202      struct sljit_label *label;
203      sljit_uw *update_addr;
204      struct label_addr_list *next;
205    } label_addr_list;
206    
207    enum frame_types {
208      no_frame = -1,
209      no_stack = -2
210    };
211    
212    enum control_types {
213      type_mark = 0,
214      type_then_trap = 1
215    };
216    
217  typedef int (SLJIT_CALL *jit_function)(jit_arguments *args);  typedef int (SLJIT_CALL *jit_function)(jit_arguments *args);
218    
219  /* The following structure is the key data type for the recursive  /* The following structure is the key data type for the recursive
220  code generator. It is allocated by compile_hotpath, and contains  code generator. It is allocated by compile_matchingpath, and contains
221  the aguments for compile_fallbackpath. Must be the first member  the arguments for compile_backtrackingpath. Must be the first member
222  of its descendants. */  of its descendants. */
223  typedef struct fallback_common {  typedef struct backtrack_common {
224    /* Concatenation stack. */    /* Concatenation stack. */
225    struct fallback_common *prev;    struct backtrack_common *prev;
226    jump_list *nextfallbacks;    jump_list *nextbacktracks;
227    /* Internal stack (for component operators). */    /* Internal stack (for component operators). */
228    struct fallback_common *top;    struct backtrack_common *top;
229    jump_list *topfallbacks;    jump_list *topbacktracks;
230    /* Opcode pointer. */    /* Opcode pointer. */
231    pcre_uchar *cc;    pcre_uchar *cc;
232  } fallback_common;  } backtrack_common;
233    
234  typedef struct assert_fallback {  typedef struct assert_backtrack {
235    fallback_common common;    backtrack_common common;
236    jump_list *condfailed;    jump_list *condfailed;
237    /* Less than 0 (-1) if a frame is not needed. */    /* Less than 0 if a frame is not needed. */
238    int framesize;    int framesize;
239    /* Points to our private memory word on the stack. */    /* Points to our private memory word on the stack. */
240    int localptr;    int private_data_ptr;
241    /* For iterators. */    /* For iterators. */
242    struct sljit_label *hotpath;    struct sljit_label *matchingpath;
243  } assert_fallback;  } assert_backtrack;
244    
245  typedef struct bracket_fallback {  typedef struct bracket_backtrack {
246    fallback_common common;    backtrack_common common;
247    /* Where to coninue if an alternative is successfully matched. */    /* Where to coninue if an alternative is successfully matched. */
248    struct sljit_label *althotpath;    struct sljit_label *alternative_matchingpath;
249    /* For rmin and rmax iterators. */    /* For rmin and rmax iterators. */
250    struct sljit_label *recursivehotpath;    struct sljit_label *recursive_matchingpath;
251    /* For greedy ? operator. */    /* For greedy ? operator. */
252    struct sljit_label *zerohotpath;    struct sljit_label *zero_matchingpath;
253    /* Contains the branches of a failed condition. */    /* Contains the branches of a failed condition. */
254    union {    union {
255      /* Both for OP_COND, OP_SCOND. */      /* Both for OP_COND, OP_SCOND. */
256      jump_list *condfailed;      jump_list *condfailed;
257      assert_fallback *assert;      assert_backtrack *assert;
258      /* For OP_ONCE. -1 if not needed. */      /* For OP_ONCE. Less than 0 if not needed. */
259      int framesize;      int framesize;
260    } u;    } u;
261    /* Points to our private memory word on the stack. */    /* Points to our private memory word on the stack. */
262    int localptr;    int private_data_ptr;
263  } bracket_fallback;  } bracket_backtrack;
264    
265  typedef struct bracketpos_fallback {  typedef struct bracketpos_backtrack {
266    fallback_common common;    backtrack_common common;
267    /* Points to our private memory word on the stack. */    /* Points to our private memory word on the stack. */
268    int localptr;    int private_data_ptr;
269    /* Reverting stack is needed. */    /* Reverting stack is needed. */
270    int framesize;    int framesize;
271    /* Allocated stack size. */    /* Allocated stack size. */
272    int stacksize;    int stacksize;
273  } bracketpos_fallback;  } bracketpos_backtrack;
274    
275  typedef struct braminzero_fallback {  typedef struct braminzero_backtrack {
276    fallback_common common;    backtrack_common common;
277    struct sljit_label *hotpath;    struct sljit_label *matchingpath;
278  } braminzero_fallback;  } braminzero_backtrack;
279    
280  typedef struct iterator_fallback {  typedef struct iterator_backtrack {
281    fallback_common common;    backtrack_common common;
282    /* Next iteration. */    /* Next iteration. */
283    struct sljit_label *hotpath;    struct sljit_label *matchingpath;
284  } iterator_fallback;  } iterator_backtrack;
285    
286  typedef struct recurse_entry {  typedef struct recurse_entry {
287    struct recurse_entry *next;    struct recurse_entry *next;
# Line 261  typedef struct recurse_entry { Line 290  typedef struct recurse_entry {
290    /* Collects the calls until the function is not created. */    /* Collects the calls until the function is not created. */
291    jump_list *calls;    jump_list *calls;
292    /* Points to the starting opcode. */    /* Points to the starting opcode. */
293    int start;    sljit_sw start;
294  } recurse_entry;  } recurse_entry;
295    
296  typedef struct recurse_fallback {  typedef struct recurse_backtrack {
297    fallback_common common;    backtrack_common common;
298  } recurse_fallback;    BOOL inlined_pattern;
299    } recurse_backtrack;
300    
301    #define OP_THEN_TRAP OP_TABLE_LENGTH
302    
303    typedef struct then_trap_backtrack {
304      backtrack_common common;
305      /* If then_trap is not NULL, this structure contains the real
306      then_trap for the backtracking path. */
307      struct then_trap_backtrack *then_trap;
308      /* Points to the starting opcode. */
309      sljit_sw start;
310      /* Exit point for the then opcodes of this alternative. */
311      jump_list *quit;
312      /* Frame size of the current alternative. */
313      int framesize;
314    } then_trap_backtrack;
315    
316    #define MAX_RANGE_SIZE 4
317    
318  typedef struct compiler_common {  typedef struct compiler_common {
319      /* The sljit ceneric compiler. */
320    struct sljit_compiler *compiler;    struct sljit_compiler *compiler;
321      /* First byte code. */
322    pcre_uchar *start;    pcre_uchar *start;
323      /* Maps private data offset to each opcode. */
324    /* Local stack area size and variable pointers. */    sljit_si *private_data_ptrs;
325    int localsize;    /* Chain list of read-only data ptrs. */
326    int *localptrs;    void *read_only_data_head;
327    int cbraptr;    /* Tells whether the capturing bracket is optimized. */
328    /* OVector starting point. Must be divisible by 2. */    pcre_uint8 *optimized_cbracket;
329      /* Tells whether the starting offset is a target of then. */
330      pcre_uint8 *then_offsets;
331      /* Current position where a THEN must jump. */
332      then_trap_backtrack *then_trap;
333      /* Starting offset of private data for capturing brackets. */
334      int cbra_ptr;
335      /* Output vector starting point. Must be divisible by 2. */
336    int ovector_start;    int ovector_start;
337    /* Last known position of the requested byte. */    /* Last known position of the requested byte. */
338    int req_char_ptr;    int req_char_ptr;
339    /* Head of the last recursion. */    /* Head of the last recursion. */
340    int recursive_head;    int recursive_head_ptr;
341    /* First inspected character for partial matching. */    /* First inspected character for partial matching. */
342    int start_used_ptr;    int start_used_ptr;
343    /* Starting pointer for partial soft matches. */    /* Starting pointer for partial soft matches. */
# Line 290  typedef struct compiler_common { Line 346  typedef struct compiler_common {
346    int first_line_end;    int first_line_end;
347    /* Points to the marked string. */    /* Points to the marked string. */
348    int mark_ptr;    int mark_ptr;
349      /* Recursive control verb management chain. */
350      int control_head_ptr;
351      /* Points to the last matched capture block index. */
352      int capture_last_ptr;
353      /* Points to the starting position of the current match. */
354      int start_ptr;
355    
356    /* Other  */    /* Flipped and lower case tables. */
357    const pcre_uint8 *fcc;    const pcre_uint8 *fcc;
358    sljit_w lcc;    sljit_sw lcc;
359      /* Mode can be PCRE_STUDY_JIT_COMPILE and others. */
360    int mode;    int mode;
361      /* TRUE, when minlength is greater than 0. */
362      BOOL might_be_empty;
363      /* \K is found in the pattern. */
364      BOOL has_set_som;
365      /* (*SKIP:arg) is found in the pattern. */
366      BOOL has_skip_arg;
367      /* (*THEN) is found in the pattern. */
368      BOOL has_then;
369      /* Needs to know the start position anytime. */
370      BOOL needs_start_ptr;
371      /* Currently in recurse or negative assert. */
372      BOOL local_exit;
373      /* Currently in a positive assert. */
374      BOOL positive_assert;
375      /* Newline control. */
376    int nltype;    int nltype;
377      pcre_uint32 nlmax;
378      pcre_uint32 nlmin;
379    int newline;    int newline;
380    int bsr_nltype;    int bsr_nltype;
381      pcre_uint32 bsr_nlmax;
382      pcre_uint32 bsr_nlmin;
383      /* Dollar endonly. */
384    int endonly;    int endonly;
385    BOOL has_set_som;    /* Tables. */
386    sljit_w ctypes;    sljit_sw ctypes;
387    sljit_uw name_table;    /* Named capturing brackets. */
388    sljit_w name_count;    pcre_uchar *name_table;
389    sljit_w name_entry_size;    sljit_sw name_count;
390      sljit_sw name_entry_size;
391    
392    /* Labels and jump lists. */    /* Labels and jump lists. */
393    struct sljit_label *partialmatchlabel;    struct sljit_label *partialmatchlabel;
394    struct sljit_label *leavelabel;    struct sljit_label *quit_label;
395    struct sljit_label *acceptlabel;    struct sljit_label *forced_quit_label;
396      struct sljit_label *accept_label;
397      struct sljit_label *ff_newline_shortcut;
398    stub_list *stubs;    stub_list *stubs;
399      label_addr_list *label_addrs;
400    recurse_entry *entries;    recurse_entry *entries;
401    recurse_entry *currententry;    recurse_entry *currententry;
402    jump_list *partialmatch;    jump_list *partialmatch;
403    jump_list *leave;    jump_list *quit;
404      jump_list *positive_assert_quit;
405      jump_list *forced_quit;
406    jump_list *accept;    jump_list *accept;
407    jump_list *calllimit;    jump_list *calllimit;
408    jump_list *stackalloc;    jump_list *stackalloc;
# Line 324  typedef struct compiler_common { Line 413  typedef struct compiler_common {
413    jump_list *vspace;    jump_list *vspace;
414    jump_list *casefulcmp;    jump_list *casefulcmp;
415    jump_list *caselesscmp;    jump_list *caselesscmp;
416      jump_list *reset_match;
417    BOOL jscript_compat;    BOOL jscript_compat;
418  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
419    BOOL utf;    BOOL utf;
420  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
421    BOOL use_ucp;    BOOL use_ucp;
422  #endif  #endif
   jump_list *utfreadchar;  
423  #ifdef COMPILE_PCRE8  #ifdef COMPILE_PCRE8
424      jump_list *utfreadchar;
425      jump_list *utfreadchar16;
426    jump_list *utfreadtype8;    jump_list *utfreadtype8;
427  #endif  #endif
428  #endif /* SUPPORT_UTF */  #endif /* SUPPORT_UTF */
# Line 348  typedef struct compare_context { Line 439  typedef struct compare_context {
439  #if defined SLJIT_UNALIGNED && SLJIT_UNALIGNED  #if defined SLJIT_UNALIGNED && SLJIT_UNALIGNED
440    int ucharptr;    int ucharptr;
441    union {    union {
442      sljit_i asint;      sljit_si asint;
443      sljit_uh asushort;      sljit_uh asushort;
444  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
445      sljit_ub asbyte;      sljit_ub asbyte;
446      sljit_ub asuchars[4];      sljit_ub asuchars[4];
447  #else  #elif defined COMPILE_PCRE16
 #ifdef COMPILE_PCRE16  
448      sljit_uh asuchars[2];      sljit_uh asuchars[2];
449  #endif  #elif defined COMPILE_PCRE32
450        sljit_ui asuchars[1];
451  #endif  #endif
452    } c;    } c;
453    union {    union {
454      sljit_i asint;      sljit_si asint;
455      sljit_uh asushort;      sljit_uh asushort;
456  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
457      sljit_ub asbyte;      sljit_ub asbyte;
458      sljit_ub asuchars[4];      sljit_ub asuchars[4];
459  #else  #elif defined COMPILE_PCRE16
 #ifdef COMPILE_PCRE16  
460      sljit_uh asuchars[2];      sljit_uh asuchars[2];
461  #endif  #elif defined COMPILE_PCRE32
462        sljit_ui asuchars[1];
463  #endif  #endif
464    } oc;    } oc;
465  #endif  #endif
466  } compare_context;  } compare_context;
467    
 enum {  
   frame_end = 0,  
   frame_setstrbegin = -1,  
   frame_setmark = -2  
 };  
   
468  /* Undefine sljit macros. */  /* Undefine sljit macros. */
469  #undef CMP  #undef CMP
470    
471  /* Used for accessing the elements of the stack. */  /* Used for accessing the elements of the stack. */
472  #define STACK(i)      ((-(i) - 1) * (int)sizeof(sljit_w))  #define STACK(i)      ((-(i) - 1) * (int)sizeof(sljit_sw))
473    
474  #define TMP1          SLJIT_TEMPORARY_REG1  #define TMP1          SLJIT_R0
475  #define TMP2          SLJIT_TEMPORARY_REG3  #define TMP2          SLJIT_R2
476  #define TMP3          SLJIT_TEMPORARY_EREG2  #define TMP3          SLJIT_R3
477  #define STR_PTR       SLJIT_SAVED_REG1  #define STR_PTR       SLJIT_S0
478  #define STR_END       SLJIT_SAVED_REG2  #define STR_END       SLJIT_S1
479  #define STACK_TOP     SLJIT_TEMPORARY_REG2  #define STACK_TOP     SLJIT_R1
480  #define STACK_LIMIT   SLJIT_SAVED_REG3  #define STACK_LIMIT   SLJIT_S2
481  #define ARGUMENTS     SLJIT_SAVED_EREG1  #define COUNT_MATCH   SLJIT_S3
482  #define CALL_COUNT    SLJIT_SAVED_EREG2  #define ARGUMENTS     SLJIT_S4
483  #define RETURN_ADDR   SLJIT_TEMPORARY_EREG1  #define RETURN_ADDR   SLJIT_R4
484    
485  /* Locals layout. */  /* Local space layout. */
486  /* These two locals can be used by the current opcode. */  /* These two locals can be used by the current opcode. */
487  #define LOCALS0          (0 * sizeof(sljit_w))  #define LOCALS0          (0 * sizeof(sljit_sw))
488  #define LOCALS1          (1 * sizeof(sljit_w))  #define LOCALS1          (1 * sizeof(sljit_sw))
489  /* Two local variables for possessive quantifiers (char1 cannot use them). */  /* Two local variables for possessive quantifiers (char1 cannot use them). */
490  #define POSSESSIVE0      (2 * sizeof(sljit_w))  #define POSSESSIVE0      (2 * sizeof(sljit_sw))
491  #define POSSESSIVE1      (3 * sizeof(sljit_w))  #define POSSESSIVE1      (3 * sizeof(sljit_sw))
492  /* Max limit of recursions. */  /* Max limit of recursions. */
493  #define CALL_LIMIT       (4 * sizeof(sljit_w))  #define LIMIT_MATCH      (4 * sizeof(sljit_sw))
494  /* The output vector is stored on the stack, and contains pointers  /* The output vector is stored on the stack, and contains pointers
495  to characters. The vector data is divided into two groups: the first  to characters. The vector data is divided into two groups: the first
496  group contains the start / end character pointers, and the second is  group contains the start / end character pointers, and the second is
497  the start pointers when the end of the capturing group has not yet reached. */  the start pointers when the end of the capturing group has not yet reached. */
498  #define OVECTOR_START    (common->ovector_start)  #define OVECTOR_START    (common->ovector_start)
499  #define OVECTOR(i)       (OVECTOR_START + (i) * sizeof(sljit_w))  #define OVECTOR(i)       (OVECTOR_START + (i) * (sljit_sw)sizeof(sljit_sw))
500  #define OVECTOR_PRIV(i)  (common->cbraptr + (i) * sizeof(sljit_w))  #define OVECTOR_PRIV(i)  (common->cbra_ptr + (i) * (sljit_sw)sizeof(sljit_sw))
501  #define PRIV_DATA(cc)    (common->localptrs[(cc) - common->start])  #define PRIVATE_DATA(cc) (common->private_data_ptrs[(cc) - common->start])
502    
503  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
504  #define MOV_UCHAR  SLJIT_MOV_UB  #define MOV_UCHAR  SLJIT_MOV_UB
505  #define MOVU_UCHAR SLJIT_MOVU_UB  #define MOVU_UCHAR SLJIT_MOVU_UB
506  #else  #elif defined COMPILE_PCRE16
 #ifdef COMPILE_PCRE16  
507  #define MOV_UCHAR  SLJIT_MOV_UH  #define MOV_UCHAR  SLJIT_MOV_UH
508  #define MOVU_UCHAR SLJIT_MOVU_UH  #define MOVU_UCHAR SLJIT_MOVU_UH
509    #elif defined COMPILE_PCRE32
510    #define MOV_UCHAR  SLJIT_MOV_UI
511    #define MOVU_UCHAR SLJIT_MOVU_UI
512  #else  #else
513  #error Unsupported compiling mode  #error Unsupported compiling mode
514  #endif  #endif
 #endif  
515    
516  /* Shortcuts. */  /* Shortcuts. */
517  #define DEFINE_COMPILER \  #define DEFINE_COMPILER \
# Line 442  the start pointers when the end of the c Line 528  the start pointers when the end of the c
528    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))
529  #define JUMPHERE(jump) \  #define JUMPHERE(jump) \
530    sljit_set_label((jump), sljit_emit_label(compiler))    sljit_set_label((jump), sljit_emit_label(compiler))
531    #define SET_LABEL(jump, label) \
532      sljit_set_label((jump), (label))
533  #define CMP(type, src1, src1w, src2, src2w) \  #define CMP(type, src1, src1w, src2, src2w) \
534    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))
535  #define CMPTO(type, src1, src1w, src2, src2w, label) \  #define CMPTO(type, src1, src1w, src2, src2w, label) \
536    sljit_set_label(sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w)), (label))    sljit_set_label(sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w)), (label))
537  #define COND_VALUE(op, dst, dstw, type) \  #define OP_FLAGS(op, dst, dstw, src, srcw, type) \
538    sljit_emit_cond_value(compiler, (op), (dst), (dstw), (type))    sljit_emit_op_flags(compiler, (op), (dst), (dstw), (src), (srcw), (type))
539    #define GET_LOCAL_BASE(dst, dstw, offset) \
540      sljit_get_local_base(compiler, (dst), (dstw), (offset))
541    
542  static pcre_uchar* bracketend(pcre_uchar* cc)  #define READ_CHAR_MAX 0x7fffffff
543    
544    static pcre_uchar *bracketend(pcre_uchar *cc)
545  {  {
546  SLJIT_ASSERT((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND));  SLJIT_ASSERT((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND));
547  do cc += GET(cc, 1); while (*cc == OP_ALT);  do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 458  cc += 1 + LINK_SIZE; Line 550  cc += 1 + LINK_SIZE;
550  return cc;  return cc;
551  }  }
552    
553    static int no_alternatives(pcre_uchar *cc)
554    {
555    int count = 0;
556    SLJIT_ASSERT((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND));
557    do
558      {
559      cc += GET(cc, 1);
560      count++;
561      }
562    while (*cc == OP_ALT);
563    SLJIT_ASSERT(*cc >= OP_KET && *cc <= OP_KETRPOS);
564    return count;
565    }
566    
567    static int ones_in_half_byte[16] = {
568      /* 0 */ 0, 1, 1, 2, /* 4 */ 1, 2, 2, 3,
569      /* 8 */ 1, 2, 2, 3, /* 12 */ 2, 3, 3, 4
570    };
571    
572  /* Functions whose might need modification for all new supported opcodes:  /* Functions whose might need modification for all new supported opcodes:
573   next_opcode   next_opcode
574   get_localspace   check_opcode_types
575   set_localptrs   set_private_data_ptrs
576   get_framesize   get_framesize
577   init_frame   init_frame
578   get_localsize   get_private_data_copy_length
579   copy_locals   copy_private_data
580   compile_hotpath   compile_matchingpath
581   compile_fallbackpath   compile_backtrackingpath
582  */  */
583    
584  static pcre_uchar *next_opcode(compiler_common *common, pcre_uchar *cc)  static pcre_uchar *next_opcode(compiler_common *common, pcre_uchar *cc)
# Line 488  switch(*cc) Line 599  switch(*cc)
599    case OP_WORDCHAR:    case OP_WORDCHAR:
600    case OP_ANY:    case OP_ANY:
601    case OP_ALLANY:    case OP_ALLANY:
602      case OP_NOTPROP:
603      case OP_PROP:
604    case OP_ANYNL:    case OP_ANYNL:
605    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
606    case OP_HSPACE:    case OP_HSPACE:
# Line 500  switch(*cc) Line 613  switch(*cc)
613    case OP_CIRCM:    case OP_CIRCM:
614    case OP_DOLL:    case OP_DOLL:
615    case OP_DOLLM:    case OP_DOLLM:
   case OP_TYPESTAR:  
   case OP_TYPEMINSTAR:  
   case OP_TYPEPLUS:  
   case OP_TYPEMINPLUS:  
   case OP_TYPEQUERY:  
   case OP_TYPEMINQUERY:  
   case OP_TYPEPOSSTAR:  
   case OP_TYPEPOSPLUS:  
   case OP_TYPEPOSQUERY:  
616    case OP_CRSTAR:    case OP_CRSTAR:
617    case OP_CRMINSTAR:    case OP_CRMINSTAR:
618    case OP_CRPLUS:    case OP_CRPLUS:
619    case OP_CRMINPLUS:    case OP_CRMINPLUS:
620    case OP_CRQUERY:    case OP_CRQUERY:
621    case OP_CRMINQUERY:    case OP_CRMINQUERY:
622      case OP_CRRANGE:
623      case OP_CRMINRANGE:
624      case OP_CRPOSSTAR:
625      case OP_CRPOSPLUS:
626      case OP_CRPOSQUERY:
627      case OP_CRPOSRANGE:
628      case OP_CLASS:
629      case OP_NCLASS:
630      case OP_REF:
631      case OP_REFI:
632      case OP_DNREF:
633      case OP_DNREFI:
634      case OP_RECURSE:
635      case OP_CALLOUT:
636      case OP_ALT:
637      case OP_KET:
638      case OP_KETRMAX:
639      case OP_KETRMIN:
640      case OP_KETRPOS:
641      case OP_REVERSE:
642      case OP_ASSERT:
643      case OP_ASSERT_NOT:
644      case OP_ASSERTBACK:
645      case OP_ASSERTBACK_NOT:
646      case OP_ONCE:
647      case OP_ONCE_NC:
648      case OP_BRA:
649      case OP_BRAPOS:
650      case OP_CBRA:
651      case OP_CBRAPOS:
652      case OP_COND:
653      case OP_SBRA:
654      case OP_SBRAPOS:
655      case OP_SCBRA:
656      case OP_SCBRAPOS:
657      case OP_SCOND:
658      case OP_CREF:
659      case OP_DNCREF:
660      case OP_RREF:
661      case OP_DNRREF:
662    case OP_DEF:    case OP_DEF:
663    case OP_BRAZERO:    case OP_BRAZERO:
664    case OP_BRAMINZERO:    case OP_BRAMINZERO:
665    case OP_BRAPOSZERO:    case OP_BRAPOSZERO:
666      case OP_PRUNE:
667      case OP_SKIP:
668      case OP_THEN:
669    case OP_COMMIT:    case OP_COMMIT:
670    case OP_FAIL:    case OP_FAIL:
671    case OP_ACCEPT:    case OP_ACCEPT:
672    case OP_ASSERT_ACCEPT:    case OP_ASSERT_ACCEPT:
673      case OP_CLOSE:
674    case OP_SKIPZERO:    case OP_SKIPZERO:
675    return cc + 1;    return cc + PRIV(OP_lengths)[*cc];
   
   case OP_ANYBYTE:  
 #ifdef SUPPORT_UTF  
   if (common->utf) return NULL;  
 #endif  
   return cc + 1;  
676    
677    case OP_CHAR:    case OP_CHAR:
678    case OP_CHARI:    case OP_CHARI:
# Line 542  switch(*cc) Line 684  switch(*cc)
684    case OP_MINPLUS:    case OP_MINPLUS:
685    case OP_QUERY:    case OP_QUERY:
686    case OP_MINQUERY:    case OP_MINQUERY:
687      case OP_UPTO:
688      case OP_MINUPTO:
689      case OP_EXACT:
690    case OP_POSSTAR:    case OP_POSSTAR:
691    case OP_POSPLUS:    case OP_POSPLUS:
692    case OP_POSQUERY:    case OP_POSQUERY:
693      case OP_POSUPTO:
694    case OP_STARI:    case OP_STARI:
695    case OP_MINSTARI:    case OP_MINSTARI:
696    case OP_PLUSI:    case OP_PLUSI:
697    case OP_MINPLUSI:    case OP_MINPLUSI:
698    case OP_QUERYI:    case OP_QUERYI:
699    case OP_MINQUERYI:    case OP_MINQUERYI:
700      case OP_UPTOI:
701      case OP_MINUPTOI:
702      case OP_EXACTI:
703    case OP_POSSTARI:    case OP_POSSTARI:
704    case OP_POSPLUSI:    case OP_POSPLUSI:
705    case OP_POSQUERYI:    case OP_POSQUERYI:
706      case OP_POSUPTOI:
707    case OP_NOTSTAR:    case OP_NOTSTAR:
708    case OP_NOTMINSTAR:    case OP_NOTMINSTAR:
709    case OP_NOTPLUS:    case OP_NOTPLUS:
710    case OP_NOTMINPLUS:    case OP_NOTMINPLUS:
711    case OP_NOTQUERY:    case OP_NOTQUERY:
712    case OP_NOTMINQUERY:    case OP_NOTMINQUERY:
713      case OP_NOTUPTO:
714      case OP_NOTMINUPTO:
715      case OP_NOTEXACT:
716    case OP_NOTPOSSTAR:    case OP_NOTPOSSTAR:
717    case OP_NOTPOSPLUS:    case OP_NOTPOSPLUS:
718    case OP_NOTPOSQUERY:    case OP_NOTPOSQUERY:
719      case OP_NOTPOSUPTO:
720    case OP_NOTSTARI:    case OP_NOTSTARI:
721    case OP_NOTMINSTARI:    case OP_NOTMINSTARI:
722    case OP_NOTPLUSI:    case OP_NOTPLUSI:
723    case OP_NOTMINPLUSI:    case OP_NOTMINPLUSI:
724    case OP_NOTQUERYI:    case OP_NOTQUERYI:
725    case OP_NOTMINQUERYI:    case OP_NOTMINQUERYI:
   case OP_NOTPOSSTARI:  
   case OP_NOTPOSPLUSI:  
   case OP_NOTPOSQUERYI:  
   cc += 2;  
 #ifdef SUPPORT_UTF  
   if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);  
 #endif  
   return cc;  
   
   case OP_UPTO:  
   case OP_MINUPTO:  
   case OP_EXACT:  
   case OP_POSUPTO:  
   case OP_UPTOI:  
   case OP_MINUPTOI:  
   case OP_EXACTI:  
   case OP_POSUPTOI:  
   case OP_NOTUPTO:  
   case OP_NOTMINUPTO:  
   case OP_NOTEXACT:  
   case OP_NOTPOSUPTO:  
726    case OP_NOTUPTOI:    case OP_NOTUPTOI:
727    case OP_NOTMINUPTOI:    case OP_NOTMINUPTOI:
728    case OP_NOTEXACTI:    case OP_NOTEXACTI:
729      case OP_NOTPOSSTARI:
730      case OP_NOTPOSPLUSI:
731      case OP_NOTPOSQUERYI:
732    case OP_NOTPOSUPTOI:    case OP_NOTPOSUPTOI:
733    cc += 2 + IMM2_SIZE;    cc += PRIV(OP_lengths)[*cc];
734  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
735    if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);    if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
736  #endif  #endif
737    return cc;    return cc;
738    
739    case OP_NOTPROP:    /* Special cases. */
740    case OP_PROP:    case OP_TYPESTAR:
741    return cc + 1 + 2;    case OP_TYPEMINSTAR:
742      case OP_TYPEPLUS:
743      case OP_TYPEMINPLUS:
744      case OP_TYPEQUERY:
745      case OP_TYPEMINQUERY:
746    case OP_TYPEUPTO:    case OP_TYPEUPTO:
747    case OP_TYPEMINUPTO:    case OP_TYPEMINUPTO:
748    case OP_TYPEEXACT:    case OP_TYPEEXACT:
749      case OP_TYPEPOSSTAR:
750      case OP_TYPEPOSPLUS:
751      case OP_TYPEPOSQUERY:
752    case OP_TYPEPOSUPTO:    case OP_TYPEPOSUPTO:
753    case OP_REF:    return cc + PRIV(OP_lengths)[*cc] - 1;
   case OP_REFI:  
   case OP_CREF:  
   case OP_NCREF:  
   case OP_RREF:  
   case OP_NRREF:  
   case OP_CLOSE:  
   cc += 1 + IMM2_SIZE;  
   return cc;  
   
   case OP_CRRANGE:  
   case OP_CRMINRANGE:  
   return cc + 1 + 2 * IMM2_SIZE;  
754    
755    case OP_CLASS:    case OP_ANYBYTE:
756    case OP_NCLASS:  #ifdef SUPPORT_UTF
757    return cc + 1 + 32 / sizeof(pcre_uchar);    if (common->utf) return NULL;
758    #endif
759      return cc + 1;
760    
761  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
762    case OP_XCLASS:    case OP_XCLASS:
763    return cc + GET(cc, 1);    return cc + GET(cc, 1);
764  #endif  #endif
765    
   case OP_RECURSE:  
   case OP_ASSERT:  
   case OP_ASSERT_NOT:  
   case OP_ASSERTBACK:  
   case OP_ASSERTBACK_NOT:  
   case OP_REVERSE:  
   case OP_ONCE:  
   case OP_ONCE_NC:  
   case OP_BRA:  
   case OP_BRAPOS:  
   case OP_COND:  
   case OP_SBRA:  
   case OP_SBRAPOS:  
   case OP_SCOND:  
   case OP_ALT:  
   case OP_KET:  
   case OP_KETRMAX:  
   case OP_KETRMIN:  
   case OP_KETRPOS:  
   return cc + 1 + LINK_SIZE;  
   
   case OP_CBRA:  
   case OP_CBRAPOS:  
   case OP_SCBRA:  
   case OP_SCBRAPOS:  
   return cc + 1 + LINK_SIZE + IMM2_SIZE;  
   
766    case OP_MARK:    case OP_MARK:
767      case OP_PRUNE_ARG:
768      case OP_SKIP_ARG:
769      case OP_THEN_ARG:
770    return cc + 1 + 2 + cc[1];    return cc + 1 + 2 + cc[1];
771    
772    default:    default:
773      /* All opcodes are supported now! */
774      SLJIT_ASSERT_STOP();
775    return NULL;    return NULL;
776    }    }
777  }  }
778    
779  static int get_localspace(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend)  static BOOL check_opcode_types(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend)
780  {  {
781  int localspace = 0;  int count;
782  pcre_uchar *alternative;  pcre_uchar *slot;
783    
784  /* Calculate important variables (like stack size) and checks whether all opcodes are supported. */  /* Calculate important variables (like stack size) and checks whether all opcodes are supported. */
785  while (cc < ccend)  while (cc < ccend)
786    {    {
# Line 677  while (cc < ccend) Line 788  while (cc < ccend)
788      {      {
789      case OP_SET_SOM:      case OP_SET_SOM:
790      common->has_set_som = TRUE;      common->has_set_som = TRUE;
791        common->might_be_empty = TRUE;
792      cc += 1;      cc += 1;
793      break;      break;
794    
795      case OP_ASSERT:      case OP_REF:
796      case OP_ASSERT_NOT:      case OP_REFI:
797      case OP_ASSERTBACK:      common->optimized_cbracket[GET2(cc, 1)] = 0;
798      case OP_ASSERTBACK_NOT:      cc += 1 + IMM2_SIZE;
     case OP_ONCE:  
     case OP_ONCE_NC:  
     case OP_BRAPOS:  
     case OP_SBRA:  
     case OP_SBRAPOS:  
     case OP_SCOND:  
     localspace += sizeof(sljit_w);  
     cc += 1 + LINK_SIZE;  
799      break;      break;
800    
801      case OP_CBRAPOS:      case OP_CBRAPOS:
802      case OP_SCBRAPOS:      case OP_SCBRAPOS:
803      localspace += sizeof(sljit_w);      common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] = 0;
804      cc += 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
805      break;      break;
806    
807      case OP_COND:      case OP_COND:
808      /* Might be a hidden SCOND. */      case OP_SCOND:
809      alternative = cc + GET(cc, 1);      /* Only AUTO_CALLOUT can insert this opcode. We do
810      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)         not intend to support this case. */
811        localspace += sizeof(sljit_w);      if (cc[1 + LINK_SIZE] == OP_CALLOUT)
812          return FALSE;
813      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
814      break;      break;
815    
816        case OP_CREF:
817        common->optimized_cbracket[GET2(cc, 1)] = 0;
818        cc += 1 + IMM2_SIZE;
819        break;
820    
821        case OP_DNREF:
822        case OP_DNREFI:
823        case OP_DNCREF:
824        count = GET2(cc, 1 + IMM2_SIZE);
825        slot = common->name_table + GET2(cc, 1) * common->name_entry_size;
826        while (count-- > 0)
827          {
828          common->optimized_cbracket[GET2(slot, 0)] = 0;
829          slot += common->name_entry_size;
830          }
831        cc += 1 + 2 * IMM2_SIZE;
832        break;
833    
834      case OP_RECURSE:      case OP_RECURSE:
835      /* Set its value only once. */      /* Set its value only once. */
836      if (common->recursive_head == 0)      if (common->recursive_head_ptr == 0)
837        {        {
838        common->recursive_head = common->ovector_start;        common->recursive_head_ptr = common->ovector_start;
839        common->ovector_start += sizeof(sljit_w);        common->ovector_start += sizeof(sljit_sw);
840        }        }
841      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
842      break;      break;
843    
844        case OP_CALLOUT:
845        if (common->capture_last_ptr == 0)
846          {
847          common->capture_last_ptr = common->ovector_start;
848          common->ovector_start += sizeof(sljit_sw);
849          }
850        cc += 2 + 2 * LINK_SIZE;
851        break;
852    
853        case OP_THEN_ARG:
854        common->has_then = TRUE;
855        common->control_head_ptr = 1;
856        /* Fall through. */
857    
858        case OP_PRUNE_ARG:
859        common->needs_start_ptr = TRUE;
860        /* Fall through. */
861    
862      case OP_MARK:      case OP_MARK:
863      if (common->mark_ptr == 0)      if (common->mark_ptr == 0)
864        {        {
865        common->mark_ptr = common->ovector_start;        common->mark_ptr = common->ovector_start;
866        common->ovector_start += sizeof(sljit_w);        common->ovector_start += sizeof(sljit_sw);
867        }        }
868      cc += 1 + 2 + cc[1];      cc += 1 + 2 + cc[1];
869      break;      break;
870    
871        case OP_THEN:
872        common->has_then = TRUE;
873        common->control_head_ptr = 1;
874        /* Fall through. */
875    
876        case OP_PRUNE:
877        case OP_SKIP:
878        common->needs_start_ptr = TRUE;
879        cc += 1;
880        break;
881    
882        case OP_SKIP_ARG:
883        common->control_head_ptr = 1;
884        common->has_skip_arg = TRUE;
885        cc += 1 + 2 + cc[1];
886        break;
887    
888      default:      default:
889      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
890      if (cc == NULL)      if (cc == NULL)
891        return -1;        return FALSE;
892      break;      break;
893      }      }
894    }    }
895  return localspace;  return TRUE;
896  }  }
897    
898  static void set_localptrs(compiler_common *common, int localptr, pcre_uchar *ccend)  static int get_class_iterator_size(pcre_uchar *cc)
899  {  {
900  pcre_uchar *cc = common->start;  switch(*cc)
 pcre_uchar *alternative;  
 while (cc < ccend)  
901    {    {
902    switch(*cc)    case OP_CRSTAR:
903      {    case OP_CRPLUS:
904      case OP_ASSERT:    return 2;
     case OP_ASSERT_NOT:  
     case OP_ASSERTBACK:  
     case OP_ASSERTBACK_NOT:  
     case OP_ONCE:  
     case OP_ONCE_NC:  
     case OP_BRAPOS:  
     case OP_SBRA:  
     case OP_SBRAPOS:  
     case OP_SCOND:  
     common->localptrs[cc - common->start] = localptr;  
     localptr += sizeof(sljit_w);  
     cc += 1 + LINK_SIZE;  
     break;  
905    
906      case OP_CBRAPOS:    case OP_CRMINSTAR:
907      case OP_SCBRAPOS:    case OP_CRMINPLUS:
908      common->localptrs[cc - common->start] = localptr;    case OP_CRQUERY:
909      localptr += sizeof(sljit_w);    case OP_CRMINQUERY:
910      cc += 1 + LINK_SIZE + IMM2_SIZE;    return 1;
     break;  
911    
912      case OP_COND:    case OP_CRRANGE:
913      /* Might be a hidden SCOND. */    case OP_CRMINRANGE:
914      alternative = cc + GET(cc, 1);    if (GET2(cc, 1) == GET2(cc, 1 + IMM2_SIZE))
915      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)      return 0;
916        {    return 2;
917        common->localptrs[cc - common->start] = localptr;  
918        localptr += sizeof(sljit_w);    default:
919        }    return 0;
920      cc += 1 + LINK_SIZE;    }
921      break;  }
922    
923    static BOOL detect_repeat(compiler_common *common, pcre_uchar *begin)
924    {
925    pcre_uchar *end = bracketend(begin);
926    pcre_uchar *next;
927    pcre_uchar *next_end;
928    pcre_uchar *max_end;
929    pcre_uchar type;
930    sljit_sw length = end - begin;
931    int min, max, i;
932    
933    /* Detect fixed iterations first. */
934    if (end[-(1 + LINK_SIZE)] != OP_KET)
935      return FALSE;
936    
937    /* Already detected repeat. */
938    if (common->private_data_ptrs[end - common->start - LINK_SIZE] != 0)
939      return TRUE;
940    
941    next = end;
942    min = 1;
943    while (1)
944      {
945      if (*next != *begin)
946        break;
947      next_end = bracketend(next);
948      if (next_end - next != length || memcmp(begin, next, IN_UCHARS(length)) != 0)
949        break;
950      next = next_end;
951      min++;
952      }
953    
954    if (min == 2)
955      return FALSE;
956    
957    max = 0;
958    max_end = next;
959    if (*next == OP_BRAZERO || *next == OP_BRAMINZERO)
960      {
961      type = *next;
962      while (1)
963        {
964        if (next[0] != type || next[1] != OP_BRA || next[2 + LINK_SIZE] != *begin)
965          break;
966        next_end = bracketend(next + 2 + LINK_SIZE);
967        if (next_end - next != (length + 2 + LINK_SIZE) || memcmp(begin, next + 2 + LINK_SIZE, IN_UCHARS(length)) != 0)
968          break;
969        next = next_end;
970        max++;
971        }
972    
973      if (next[0] == type && next[1] == *begin && max >= 1)
974        {
975        next_end = bracketend(next + 1);
976        if (next_end - next == (length + 1) && memcmp(begin, next + 1, IN_UCHARS(length)) == 0)
977          {
978          for (i = 0; i < max; i++, next_end += 1 + LINK_SIZE)
979            if (*next_end != OP_KET)
980              break;
981    
982          if (i == max)
983            {
984            common->private_data_ptrs[max_end - common->start - LINK_SIZE] = next_end - max_end;
985            common->private_data_ptrs[max_end - common->start - LINK_SIZE + 1] = (type == OP_BRAZERO) ? OP_UPTO : OP_MINUPTO;
986            /* +2 the original and the last. */
987            common->private_data_ptrs[max_end - common->start - LINK_SIZE + 2] = max + 2;
988            if (min == 1)
989              return TRUE;
990            min--;
991            max_end -= (1 + LINK_SIZE) + GET(max_end, -LINK_SIZE);
992            }
993          }
994        }
995      }
996    
997    if (min >= 3)
998      {
999      common->private_data_ptrs[end - common->start - LINK_SIZE] = max_end - end;
1000      common->private_data_ptrs[end - common->start - LINK_SIZE + 1] = OP_EXACT;
1001      common->private_data_ptrs[end - common->start - LINK_SIZE + 2] = min;
1002      return TRUE;
1003      }
1004    
1005    return FALSE;
1006    }
1007    
1008    #define CASE_ITERATOR_PRIVATE_DATA_1 \
1009        case OP_MINSTAR: \
1010        case OP_MINPLUS: \
1011        case OP_QUERY: \
1012        case OP_MINQUERY: \
1013        case OP_MINSTARI: \
1014        case OP_MINPLUSI: \
1015        case OP_QUERYI: \
1016        case OP_MINQUERYI: \
1017        case OP_NOTMINSTAR: \
1018        case OP_NOTMINPLUS: \
1019        case OP_NOTQUERY: \
1020        case OP_NOTMINQUERY: \
1021        case OP_NOTMINSTARI: \
1022        case OP_NOTMINPLUSI: \
1023        case OP_NOTQUERYI: \
1024        case OP_NOTMINQUERYI:
1025    
1026    #define CASE_ITERATOR_PRIVATE_DATA_2A \
1027        case OP_STAR: \
1028        case OP_PLUS: \
1029        case OP_STARI: \
1030        case OP_PLUSI: \
1031        case OP_NOTSTAR: \
1032        case OP_NOTPLUS: \
1033        case OP_NOTSTARI: \
1034        case OP_NOTPLUSI:
1035    
1036    #define CASE_ITERATOR_PRIVATE_DATA_2B \
1037        case OP_UPTO: \
1038        case OP_MINUPTO: \
1039        case OP_UPTOI: \
1040        case OP_MINUPTOI: \
1041        case OP_NOTUPTO: \
1042        case OP_NOTMINUPTO: \
1043        case OP_NOTUPTOI: \
1044        case OP_NOTMINUPTOI:
1045    
1046    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_1 \
1047        case OP_TYPEMINSTAR: \
1048        case OP_TYPEMINPLUS: \
1049        case OP_TYPEQUERY: \
1050        case OP_TYPEMINQUERY:
1051    
1052    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2A \
1053        case OP_TYPESTAR: \
1054        case OP_TYPEPLUS:
1055    
1056    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2B \
1057        case OP_TYPEUPTO: \
1058        case OP_TYPEMINUPTO:
1059    
1060    static void set_private_data_ptrs(compiler_common *common, int *private_data_start, pcre_uchar *ccend)
1061    {
1062    pcre_uchar *cc = common->start;
1063    pcre_uchar *alternative;
1064    pcre_uchar *end = NULL;
1065    int private_data_ptr = *private_data_start;
1066    int space, size, bracketlen;
1067    
1068    while (cc < ccend)
1069      {
1070      space = 0;
1071      size = 0;
1072      bracketlen = 0;
1073      if (private_data_ptr > SLJIT_MAX_LOCAL_SIZE)
1074        return;
1075    
1076      if (*cc == OP_ONCE || *cc == OP_ONCE_NC || *cc == OP_BRA || *cc == OP_CBRA || *cc == OP_COND)
1077        if (detect_repeat(common, cc))
1078          {
1079          /* These brackets are converted to repeats, so no global
1080          based single character repeat is allowed. */
1081          if (cc >= end)
1082            end = bracketend(cc);
1083          }
1084    
1085      switch(*cc)
1086        {
1087        case OP_KET:
1088        if (common->private_data_ptrs[cc + 1 - common->start] != 0)
1089          {
1090          common->private_data_ptrs[cc - common->start] = private_data_ptr;
1091          private_data_ptr += sizeof(sljit_sw);
1092          cc += common->private_data_ptrs[cc + 1 - common->start];
1093          }
1094        cc += 1 + LINK_SIZE;
1095        break;
1096    
1097        case OP_ASSERT:
1098        case OP_ASSERT_NOT:
1099        case OP_ASSERTBACK:
1100        case OP_ASSERTBACK_NOT:
1101        case OP_ONCE:
1102        case OP_ONCE_NC:
1103        case OP_BRAPOS:
1104        case OP_SBRA:
1105        case OP_SBRAPOS:
1106        case OP_SCOND:
1107        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1108        private_data_ptr += sizeof(sljit_sw);
1109        bracketlen = 1 + LINK_SIZE;
1110        break;
1111    
1112        case OP_CBRAPOS:
1113        case OP_SCBRAPOS:
1114        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1115        private_data_ptr += sizeof(sljit_sw);
1116        bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
1117        break;
1118    
1119        case OP_COND:
1120        /* Might be a hidden SCOND. */
1121        alternative = cc + GET(cc, 1);
1122        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1123          {
1124          common->private_data_ptrs[cc - common->start] = private_data_ptr;
1125          private_data_ptr += sizeof(sljit_sw);
1126          }
1127        bracketlen = 1 + LINK_SIZE;
1128        break;
1129    
1130        case OP_BRA:
1131        bracketlen = 1 + LINK_SIZE;
1132        break;
1133    
1134        case OP_CBRA:
1135        case OP_SCBRA:
1136        bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
1137        break;
1138    
1139        CASE_ITERATOR_PRIVATE_DATA_1
1140        space = 1;
1141        size = -2;
1142        break;
1143    
1144        CASE_ITERATOR_PRIVATE_DATA_2A
1145        space = 2;
1146        size = -2;
1147        break;
1148    
1149        CASE_ITERATOR_PRIVATE_DATA_2B
1150        space = 2;
1151        size = -(2 + IMM2_SIZE);
1152        break;
1153    
1154        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1155        space = 1;
1156        size = 1;
1157        break;
1158    
1159        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1160        if (cc[1] != OP_ANYNL && cc[1] != OP_EXTUNI)
1161          space = 2;
1162        size = 1;
1163        break;
1164    
1165        CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1166        if (cc[1 + IMM2_SIZE] != OP_ANYNL && cc[1 + IMM2_SIZE] != OP_EXTUNI)
1167          space = 2;
1168        size = 1 + IMM2_SIZE;
1169        break;
1170    
1171        case OP_CLASS:
1172        case OP_NCLASS:
1173        size += 1 + 32 / sizeof(pcre_uchar);
1174        space = get_class_iterator_size(cc + size);
1175        break;
1176    
1177    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1178        case OP_XCLASS:
1179        size = GET(cc, 1);
1180        space = get_class_iterator_size(cc + size);
1181        break;
1182    #endif
1183    
1184      default:      default:
1185      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
1186      SLJIT_ASSERT(cc != NULL);      SLJIT_ASSERT(cc != NULL);
1187      break;      break;
1188      }      }
1189    
1190      /* Character iterators, which are not inside a repeated bracket,
1191         gets a private slot instead of allocating it on the stack. */
1192      if (space > 0 && cc >= end)
1193        {
1194        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1195        private_data_ptr += sizeof(sljit_sw) * space;
1196        }
1197    
1198      if (size != 0)
1199        {
1200        if (size < 0)
1201          {
1202          cc += -size;
1203    #ifdef SUPPORT_UTF
1204          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1205    #endif
1206          }
1207        else
1208          cc += size;
1209        }
1210    
1211      if (bracketlen > 0)
1212        {
1213        if (cc >= end)
1214          {
1215          end = bracketend(cc);
1216          if (end[-1 - LINK_SIZE] == OP_KET)
1217            end = NULL;
1218          }
1219        cc += bracketlen;
1220        }
1221    }    }
1222    *private_data_start = private_data_ptr;
1223  }  }
1224    
1225  /* Returns with -1 if no need for frame. */  /* Returns with a frame_types (always < 0) if no need for frame. */
1226  static int get_framesize(compiler_common *common, pcre_uchar *cc, BOOL recursive)  static int get_framesize(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, BOOL recursive, BOOL *needs_control_head)
1227  {  {
 pcre_uchar *ccend = bracketend(cc);  
1228  int length = 0;  int length = 0;
1229  BOOL possessive = FALSE;  int possessive = 0;
1230    BOOL stack_restore = FALSE;
1231  BOOL setsom_found = recursive;  BOOL setsom_found = recursive;
1232  BOOL setmark_found = recursive;  BOOL setmark_found = recursive;
1233    /* The last capture is a local variable even for recursions. */
1234    BOOL capture_last_found = FALSE;
1235    
1236    #if defined DEBUG_FORCE_CONTROL_HEAD && DEBUG_FORCE_CONTROL_HEAD
1237    SLJIT_ASSERT(common->control_head_ptr != 0);
1238    *needs_control_head = TRUE;
1239    #else
1240    *needs_control_head = FALSE;
1241    #endif
1242    
1243  if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))  if (ccend == NULL)
1244    {    {
1245    length = 3;    ccend = bracketend(cc) - (1 + LINK_SIZE);
1246    possessive = TRUE;    if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))
1247        {
1248        possessive = length = (common->capture_last_ptr != 0) ? 5 : 3;
1249        /* This is correct regardless of common->capture_last_ptr. */
1250        capture_last_found = TRUE;
1251        }
1252      cc = next_opcode(common, cc);
1253    }    }
1254    
 cc = next_opcode(common, cc);  
1255  SLJIT_ASSERT(cc != NULL);  SLJIT_ASSERT(cc != NULL);
1256  while (cc < ccend)  while (cc < ccend)
1257    switch(*cc)    switch(*cc)
1258      {      {
1259      case OP_SET_SOM:      case OP_SET_SOM:
1260      SLJIT_ASSERT(common->has_set_som);      SLJIT_ASSERT(common->has_set_som);
1261        stack_restore = TRUE;
1262      if (!setsom_found)      if (!setsom_found)
1263        {        {
1264        length += 2;        length += 2;
# Line 817  while (cc < ccend) Line 1268  while (cc < ccend)
1268      break;      break;
1269    
1270      case OP_MARK:      case OP_MARK:
1271        case OP_PRUNE_ARG:
1272        case OP_THEN_ARG:
1273      SLJIT_ASSERT(common->mark_ptr != 0);      SLJIT_ASSERT(common->mark_ptr != 0);
1274        stack_restore = TRUE;
1275      if (!setmark_found)      if (!setmark_found)
1276        {        {
1277        length += 2;        length += 2;
1278        setmark_found = TRUE;        setmark_found = TRUE;
1279        }        }
1280        if (common->control_head_ptr != 0)
1281          *needs_control_head = TRUE;
1282      cc += 1 + 2 + cc[1];      cc += 1 + 2 + cc[1];
1283      break;      break;
1284    
1285      case OP_RECURSE:      case OP_RECURSE:
1286        stack_restore = TRUE;
1287      if (common->has_set_som && !setsom_found)      if (common->has_set_som && !setsom_found)
1288        {        {
1289        length += 2;        length += 2;
# Line 837  while (cc < ccend) Line 1294  while (cc < ccend)
1294        length += 2;        length += 2;
1295        setmark_found = TRUE;        setmark_found = TRUE;
1296        }        }
1297        if (common->capture_last_ptr != 0 && !capture_last_found)
1298          {
1299          length += 2;
1300          capture_last_found = TRUE;
1301          }
1302      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1303      break;      break;
1304    
# Line 844  while (cc < ccend) Line 1306  while (cc < ccend)
1306      case OP_CBRAPOS:      case OP_CBRAPOS:
1307      case OP_SCBRA:      case OP_SCBRA:
1308      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1309        stack_restore = TRUE;
1310        if (common->capture_last_ptr != 0 && !capture_last_found)
1311          {
1312          length += 2;
1313          capture_last_found = TRUE;
1314          }
1315      length += 3;      length += 3;
1316      cc += 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1317      break;      break;
1318    
1319      default:      default:
1320        stack_restore = TRUE;
1321        /* Fall through. */
1322    
1323        case OP_NOT_WORD_BOUNDARY:
1324        case OP_WORD_BOUNDARY:
1325        case OP_NOT_DIGIT:
1326        case OP_DIGIT:
1327        case OP_NOT_WHITESPACE:
1328        case OP_WHITESPACE:
1329        case OP_NOT_WORDCHAR:
1330        case OP_WORDCHAR:
1331        case OP_ANY:
1332        case OP_ALLANY:
1333        case OP_ANYBYTE:
1334        case OP_NOTPROP:
1335        case OP_PROP:
1336        case OP_ANYNL:
1337        case OP_NOT_HSPACE:
1338        case OP_HSPACE:
1339        case OP_NOT_VSPACE:
1340        case OP_VSPACE:
1341        case OP_EXTUNI:
1342        case OP_EODN:
1343        case OP_EOD:
1344        case OP_CIRC:
1345        case OP_CIRCM:
1346        case OP_DOLL:
1347        case OP_DOLLM:
1348        case OP_CHAR:
1349        case OP_CHARI:
1350        case OP_NOT:
1351        case OP_NOTI:
1352    
1353        case OP_EXACT:
1354        case OP_POSSTAR:
1355        case OP_POSPLUS:
1356        case OP_POSQUERY:
1357        case OP_POSUPTO:
1358    
1359        case OP_EXACTI:
1360        case OP_POSSTARI:
1361        case OP_POSPLUSI:
1362        case OP_POSQUERYI:
1363        case OP_POSUPTOI:
1364    
1365        case OP_NOTEXACT:
1366        case OP_NOTPOSSTAR:
1367        case OP_NOTPOSPLUS:
1368        case OP_NOTPOSQUERY:
1369        case OP_NOTPOSUPTO:
1370    
1371        case OP_NOTEXACTI:
1372        case OP_NOTPOSSTARI:
1373        case OP_NOTPOSPLUSI:
1374        case OP_NOTPOSQUERYI:
1375        case OP_NOTPOSUPTOI:
1376    
1377        case OP_TYPEEXACT:
1378        case OP_TYPEPOSSTAR:
1379        case OP_TYPEPOSPLUS:
1380        case OP_TYPEPOSQUERY:
1381        case OP_TYPEPOSUPTO:
1382    
1383        case OP_CLASS:
1384        case OP_NCLASS:
1385        case OP_XCLASS:
1386    
1387      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
1388      SLJIT_ASSERT(cc != NULL);      SLJIT_ASSERT(cc != NULL);
1389      break;      break;
1390      }      }
1391    
1392  /* Possessive quantifiers can use a special case. */  /* Possessive quantifiers can use a special case. */
1393  if (SLJIT_UNLIKELY(possessive) && length == 3)  if (SLJIT_UNLIKELY(possessive == length))
1394    return -1;    return stack_restore ? no_frame : no_stack;
1395    
1396  if (length > 0)  if (length > 0)
1397    return length + 1;    return length + 1;
1398  return -1;  return stack_restore ? no_frame : no_stack;
1399  }  }
1400    
1401  static void init_frame(compiler_common *common, pcre_uchar *cc, int stackpos, int stacktop, BOOL recursive)  static void init_frame(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, int stackpos, int stacktop, BOOL recursive)
1402  {  {
1403  DEFINE_COMPILER;  DEFINE_COMPILER;
 pcre_uchar *ccend = bracketend(cc);  
1404  BOOL setsom_found = recursive;  BOOL setsom_found = recursive;
1405  BOOL setmark_found = recursive;  BOOL setmark_found = recursive;
1406    /* The last capture is a local variable even for recursions. */
1407    BOOL capture_last_found = FALSE;
1408  int offset;  int offset;
1409    
1410  /* >= 1 + shortest item size (2) */  /* >= 1 + shortest item size (2) */
# Line 876  SLJIT_UNUSED_ARG(stacktop); Line 1412  SLJIT_UNUSED_ARG(stacktop);
1412  SLJIT_ASSERT(stackpos >= stacktop + 2);  SLJIT_ASSERT(stackpos >= stacktop + 2);
1413    
1414  stackpos = STACK(stackpos);  stackpos = STACK(stackpos);
1415  if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))  if (ccend == NULL)
1416    cc = next_opcode(common, cc);    {
1417      ccend = bracketend(cc) - (1 + LINK_SIZE);
1418      if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))
1419        cc = next_opcode(common, cc);
1420      }
1421    
1422  SLJIT_ASSERT(cc != NULL);  SLJIT_ASSERT(cc != NULL);
1423  while (cc < ccend)  while (cc < ccend)
1424    switch(*cc)    switch(*cc)
# Line 886  while (cc < ccend) Line 1427  while (cc < ccend)
1427      SLJIT_ASSERT(common->has_set_som);      SLJIT_ASSERT(common->has_set_som);
1428      if (!setsom_found)      if (!setsom_found)
1429        {        {
1430        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(0));
1431        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setstrbegin);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1432        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1433        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1434        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1435        setsom_found = TRUE;        setsom_found = TRUE;
1436        }        }
1437      cc += 1;      cc += 1;
1438      break;      break;
1439    
1440      case OP_MARK:      case OP_MARK:
1441        case OP_PRUNE_ARG:
1442        case OP_THEN_ARG:
1443      SLJIT_ASSERT(common->mark_ptr != 0);      SLJIT_ASSERT(common->mark_ptr != 0);
1444      if (!setmark_found)      if (!setmark_found)
1445        {        {
1446        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->mark_ptr);
1447        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setmark);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
1448        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1449        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1450        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1451        setmark_found = TRUE;        setmark_found = TRUE;
1452        }        }
1453      cc += 1 + 2 + cc[1];      cc += 1 + 2 + cc[1];
# Line 913  while (cc < ccend) Line 1456  while (cc < ccend)
1456      case OP_RECURSE:      case OP_RECURSE:
1457      if (common->has_set_som && !setsom_found)      if (common->has_set_som && !setsom_found)
1458        {        {
1459        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(0));
1460        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setstrbegin);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1461        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1462        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1463        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
1464        setsom_found = TRUE;        setsom_found = TRUE;
1465        }        }
1466      if (common->mark_ptr != 0 && !setmark_found)      if (common->mark_ptr != 0 && !setmark_found)
1467        {        {
1468        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->mark_ptr);
1469        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setmark);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
1470        stackpos += (int)sizeof(sljit_w);        stackpos += (int)sizeof(sljit_sw);
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        setmark_found = TRUE;        setmark_found = TRUE;
1474        }        }
1475        if (common->capture_last_ptr != 0 && !capture_last_found)
1476          {
1477          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->capture_last_ptr);
1478          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1479          stackpos += (int)sizeof(sljit_sw);
1480          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1481          stackpos += (int)sizeof(sljit_sw);
1482          capture_last_found = TRUE;
1483          }
1484      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1485      break;      break;
1486    
# Line 936  while (cc < ccend) Line 1488  while (cc < ccend)
1488      case OP_CBRAPOS:      case OP_CBRAPOS:
1489      case OP_SCBRA:      case OP_SCBRA:
1490      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1491        if (common->capture_last_ptr != 0 && !capture_last_found)
1492          {
1493          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->capture_last_ptr);
1494          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1495          stackpos += (int)sizeof(sljit_sw);
1496          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1497          stackpos += (int)sizeof(sljit_sw);
1498          capture_last_found = TRUE;
1499          }
1500      offset = (GET2(cc, 1 + LINK_SIZE)) << 1;      offset = (GET2(cc, 1 + LINK_SIZE)) << 1;
1501      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));
1502      stackpos += (int)sizeof(sljit_w);      stackpos += (int)sizeof(sljit_sw);
1503      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset));      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(offset));
1504      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset + 1));      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(offset + 1));
1505      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1506      stackpos += (int)sizeof(sljit_w);      stackpos += (int)sizeof(sljit_sw);
1507      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);
1508      stackpos += (int)sizeof(sljit_w);      stackpos += (int)sizeof(sljit_sw);
1509    
1510      cc += 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1511      break;      break;
# Line 955  while (cc < ccend) Line 1516  while (cc < ccend)
1516      break;      break;
1517      }      }
1518    
1519  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_end);  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, 0);
1520  SLJIT_ASSERT(stackpos == STACK(stacktop));  SLJIT_ASSERT(stackpos == STACK(stacktop));
1521  }  }
1522    
1523  static SLJIT_INLINE int get_localsize(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend)  static SLJIT_INLINE int get_private_data_copy_length(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, BOOL needs_control_head)
1524  {  {
1525  int localsize = 2;  int private_data_length = needs_control_head ? 3 : 2;
1526    int size;
1527  pcre_uchar *alternative;  pcre_uchar *alternative;
1528  /* Calculate the sum of the local variables. */  /* Calculate the sum of the private machine words. */
1529  while (cc < ccend)  while (cc < ccend)
1530    {    {
1531      size = 0;
1532    switch(*cc)    switch(*cc)
1533      {      {
1534        case OP_KET:
1535        if (PRIVATE_DATA(cc) != 0)
1536          private_data_length++;
1537        cc += 1 + LINK_SIZE;
1538        break;
1539    
1540      case OP_ASSERT:      case OP_ASSERT:
1541      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1542      case OP_ASSERTBACK:      case OP_ASSERTBACK:
# Line 978  while (cc < ccend) Line 1547  while (cc < ccend)
1547      case OP_SBRA:      case OP_SBRA:
1548      case OP_SBRAPOS:      case OP_SBRAPOS:
1549      case OP_SCOND:      case OP_SCOND:
1550      localsize++;      private_data_length++;
1551      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1552      break;      break;
1553    
1554      case OP_CBRA:      case OP_CBRA:
1555      case OP_SCBRA:      case OP_SCBRA:
1556      localsize++;      if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1557          private_data_length++;
1558      cc += 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1559      break;      break;
1560    
1561      case OP_CBRAPOS:      case OP_CBRAPOS:
1562      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1563      localsize += 2;      private_data_length += 2;
1564      cc += 1 + LINK_SIZE + IMM2_SIZE;      cc += 1 + LINK_SIZE + IMM2_SIZE;
1565      break;      break;
1566    
# Line 998  while (cc < ccend) Line 1568  while (cc < ccend)
1568      /* Might be a hidden SCOND. */      /* Might be a hidden SCOND. */
1569      alternative = cc + GET(cc, 1);      alternative = cc + GET(cc, 1);
1570      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1571        localsize++;        private_data_length++;
1572      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1573      break;      break;
1574    
1575      default:      CASE_ITERATOR_PRIVATE_DATA_1
1576      cc = next_opcode(common, cc);      if (PRIVATE_DATA(cc))
1577      SLJIT_ASSERT(cc != NULL);        private_data_length++;
1578        cc += 2;
1579    #ifdef SUPPORT_UTF
1580        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1581    #endif
1582      break;      break;
     }  
   }  
 SLJIT_ASSERT(cc == ccend);  
 return localsize;  
 }  
   
 static void copy_locals(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend,  
   BOOL save, int stackptr, int stacktop)  
 {  
 DEFINE_COMPILER;  
 int srcw[2];  
 int count;  
 BOOL tmp1next = TRUE;  
 BOOL tmp1empty = TRUE;  
 BOOL tmp2empty = TRUE;  
 pcre_uchar *alternative;  
 enum {  
   start,  
   loop,  
   end  
 } status;  
1583    
1584  status = save ? start : loop;      CASE_ITERATOR_PRIVATE_DATA_2A
1585  stackptr = STACK(stackptr - 2);      if (PRIVATE_DATA(cc))
1586  stacktop = STACK(stacktop - 1);        private_data_length += 2;
1587        cc += 2;
1588    #ifdef SUPPORT_UTF
1589        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1590    #endif
1591        break;
1592    
1593        CASE_ITERATOR_PRIVATE_DATA_2B
1594        if (PRIVATE_DATA(cc))
1595          private_data_length += 2;
1596        cc += 2 + IMM2_SIZE;
1597    #ifdef SUPPORT_UTF
1598        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1599    #endif
1600        break;
1601    
1602        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1603        if (PRIVATE_DATA(cc))
1604          private_data_length++;
1605        cc += 1;
1606        break;
1607    
1608        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1609        if (PRIVATE_DATA(cc))
1610          private_data_length += 2;
1611        cc += 1;
1612        break;
1613    
1614        CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1615        if (PRIVATE_DATA(cc))
1616          private_data_length += 2;
1617        cc += 1 + IMM2_SIZE;
1618        break;
1619    
1620        case OP_CLASS:
1621        case OP_NCLASS:
1622    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1623        case OP_XCLASS:
1624        size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
1625    #else
1626        size = 1 + 32 / (int)sizeof(pcre_uchar);
1627    #endif
1628        if (PRIVATE_DATA(cc))
1629          private_data_length += get_class_iterator_size(cc + size);
1630        cc += size;
1631        break;
1632    
1633        default:
1634        cc = next_opcode(common, cc);
1635        SLJIT_ASSERT(cc != NULL);
1636        break;
1637        }
1638      }
1639    SLJIT_ASSERT(cc == ccend);
1640    return private_data_length;
1641    }
1642    
1643    static void copy_private_data(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend,
1644      BOOL save, int stackptr, int stacktop, BOOL needs_control_head)
1645    {
1646    DEFINE_COMPILER;
1647    int srcw[2];
1648    int count, size;
1649    BOOL tmp1next = TRUE;
1650    BOOL tmp1empty = TRUE;
1651    BOOL tmp2empty = TRUE;
1652    pcre_uchar *alternative;
1653    enum {
1654      start,
1655      loop,
1656      end
1657    } status;
1658    
1659    status = save ? start : loop;
1660    stackptr = STACK(stackptr - 2);
1661    stacktop = STACK(stacktop - 1);
1662    
1663  if (!save)  if (!save)
1664    {    {
1665    stackptr += sizeof(sljit_w);    stackptr += (needs_control_head ? 2 : 1) * sizeof(sljit_sw);
1666    if (stackptr < stacktop)    if (stackptr < stacktop)
1667      {      {
1668      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1669      stackptr += sizeof(sljit_w);      stackptr += sizeof(sljit_sw);
1670      tmp1empty = FALSE;      tmp1empty = FALSE;
1671      }      }
1672    if (stackptr < stacktop)    if (stackptr < stacktop)
1673      {      {
1674      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1675      stackptr += sizeof(sljit_w);      stackptr += sizeof(sljit_sw);
1676      tmp2empty = FALSE;      tmp2empty = FALSE;
1677      }      }
1678    /* The tmp1next must be TRUE in either way. */    /* The tmp1next must be TRUE in either way. */
1679    }    }
1680    
1681  while (status != end)  do
1682    {    {
1683    count = 0;    count = 0;
1684    switch(status)    switch(status)
1685      {      {
1686      case start:      case start:
1687      SLJIT_ASSERT(save && common->recursive_head != 0);      SLJIT_ASSERT(save && common->recursive_head_ptr != 0);
1688      count = 1;      count = 1;
1689      srcw[0] = common->recursive_head;      srcw[0] = common->recursive_head_ptr;
1690        if (needs_control_head)
1691          {
1692          SLJIT_ASSERT(common->control_head_ptr != 0);
1693          count = 2;
1694          srcw[1] = common->control_head_ptr;
1695          }
1696      status = loop;      status = loop;
1697      break;      break;
1698    
# Line 1071  while (status != end) Line 1705  while (status != end)
1705    
1706      switch(*cc)      switch(*cc)
1707        {        {
1708          case OP_KET:
1709          if (PRIVATE_DATA(cc) != 0)
1710            {
1711            count = 1;
1712            srcw[0] = PRIVATE_DATA(cc);
1713            }
1714          cc += 1 + LINK_SIZE;
1715          break;
1716    
1717        case OP_ASSERT:        case OP_ASSERT:
1718        case OP_ASSERT_NOT:        case OP_ASSERT_NOT:
1719        case OP_ASSERTBACK:        case OP_ASSERTBACK:
# Line 1082  while (status != end) Line 1725  while (status != end)
1725        case OP_SBRAPOS:        case OP_SBRAPOS:
1726        case OP_SCOND:        case OP_SCOND:
1727        count = 1;        count = 1;
1728        srcw[0] = PRIV_DATA(cc);        srcw[0] = PRIVATE_DATA(cc);
1729        SLJIT_ASSERT(srcw[0] != 0);        SLJIT_ASSERT(srcw[0] != 0);
1730        cc += 1 + LINK_SIZE;        cc += 1 + LINK_SIZE;
1731        break;        break;
1732    
1733        case OP_CBRA:        case OP_CBRA:
1734        case OP_SCBRA:        case OP_SCBRA:
1735        count = 1;        if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1736        srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));          {
1737            count = 1;
1738            srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1739            }
1740        cc += 1 + LINK_SIZE + IMM2_SIZE;        cc += 1 + LINK_SIZE + IMM2_SIZE;
1741        break;        break;
1742    
1743        case OP_CBRAPOS:        case OP_CBRAPOS:
1744        case OP_SCBRAPOS:        case OP_SCBRAPOS:
1745        count = 2;        count = 2;
1746          srcw[0] = PRIVATE_DATA(cc);
1747        srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));        srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1748        srcw[0] = PRIV_DATA(cc);        SLJIT_ASSERT(srcw[0] != 0 && srcw[1] != 0);
       SLJIT_ASSERT(srcw[0] != 0);  
1749        cc += 1 + LINK_SIZE + IMM2_SIZE;        cc += 1 + LINK_SIZE + IMM2_SIZE;
1750        break;        break;
1751    
# Line 1109  while (status != end) Line 1755  while (status != end)
1755        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1756          {          {
1757          count = 1;          count = 1;
1758          srcw[0] = PRIV_DATA(cc);          srcw[0] = PRIVATE_DATA(cc);
1759          SLJIT_ASSERT(srcw[0] != 0);          SLJIT_ASSERT(srcw[0] != 0);
1760          }          }
1761        cc += 1 + LINK_SIZE;        cc += 1 + LINK_SIZE;
1762        break;        break;
1763    
1764          CASE_ITERATOR_PRIVATE_DATA_1
1765          if (PRIVATE_DATA(cc))
1766            {
1767            count = 1;
1768            srcw[0] = PRIVATE_DATA(cc);
1769            }
1770          cc += 2;
1771    #ifdef SUPPORT_UTF
1772          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1773    #endif
1774          break;
1775    
1776          CASE_ITERATOR_PRIVATE_DATA_2A
1777          if (PRIVATE_DATA(cc))
1778            {
1779            count = 2;
1780            srcw[0] = PRIVATE_DATA(cc);
1781            srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1782            }
1783          cc += 2;
1784    #ifdef SUPPORT_UTF
1785          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1786    #endif
1787          break;
1788    
1789          CASE_ITERATOR_PRIVATE_DATA_2B
1790          if (PRIVATE_DATA(cc))
1791            {
1792            count = 2;
1793            srcw[0] = PRIVATE_DATA(cc);
1794            srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1795            }
1796          cc += 2 + IMM2_SIZE;
1797    #ifdef SUPPORT_UTF
1798          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1799    #endif
1800          break;
1801    
1802          CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1803          if (PRIVATE_DATA(cc))
1804            {
1805            count = 1;
1806            srcw[0] = PRIVATE_DATA(cc);
1807            }
1808          cc += 1;
1809          break;
1810    
1811          CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1812          if (PRIVATE_DATA(cc))
1813            {
1814            count = 2;
1815            srcw[0] = PRIVATE_DATA(cc);
1816            srcw[1] = srcw[0] + sizeof(sljit_sw);
1817            }
1818          cc += 1;
1819          break;
1820    
1821          CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1822          if (PRIVATE_DATA(cc))
1823            {
1824            count = 2;
1825            srcw[0] = PRIVATE_DATA(cc);
1826            srcw[1] = srcw[0] + sizeof(sljit_sw);
1827            }
1828          cc += 1 + IMM2_SIZE;
1829          break;
1830    
1831          case OP_CLASS:
1832          case OP_NCLASS:
1833    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1834          case OP_XCLASS:
1835          size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
1836    #else
1837          size = 1 + 32 / (int)sizeof(pcre_uchar);
1838    #endif
1839          if (PRIVATE_DATA(cc))
1840            switch(get_class_iterator_size(cc + size))
1841              {
1842              case 1:
1843              count = 1;
1844              srcw[0] = PRIVATE_DATA(cc);
1845              break;
1846    
1847              case 2:
1848              count = 2;
1849              srcw[0] = PRIVATE_DATA(cc);
1850              srcw[1] = srcw[0] + sizeof(sljit_sw);
1851              break;
1852    
1853              default:
1854              SLJIT_ASSERT_STOP();
1855              break;
1856              }
1857          cc += size;
1858          break;
1859    
1860        default:        default:
1861        cc = next_opcode(common, cc);        cc = next_opcode(common, cc);
1862        SLJIT_ASSERT(cc != NULL);        SLJIT_ASSERT(cc != NULL);
# Line 1137  while (status != end) Line 1879  while (status != end)
1879          if (!tmp1empty)          if (!tmp1empty)
1880            {            {
1881            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1882            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1883            }            }
1884          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), srcw[count]);
1885          tmp1empty = FALSE;          tmp1empty = FALSE;
1886          tmp1next = FALSE;          tmp1next = FALSE;
1887          }          }
# Line 1148  while (status != end) Line 1890  while (status != end)
1890          if (!tmp2empty)          if (!tmp2empty)
1891            {            {
1892            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);            OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1893            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1894            }            }
1895          OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);          OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_SP), srcw[count]);
1896          tmp2empty = FALSE;          tmp2empty = FALSE;
1897          tmp1next = TRUE;          tmp1next = TRUE;
1898          }          }
# Line 1160  while (status != end) Line 1902  while (status != end)
1902        if (tmp1next)        if (tmp1next)
1903          {          {
1904          SLJIT_ASSERT(!tmp1empty);          SLJIT_ASSERT(!tmp1empty);
1905          OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count], TMP1, 0);          OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), srcw[count], TMP1, 0);
1906          tmp1empty = stackptr >= stacktop;          tmp1empty = stackptr >= stacktop;
1907          if (!tmp1empty)          if (!tmp1empty)
1908            {            {
1909            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1910            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1911            }            }
1912          tmp1next = FALSE;          tmp1next = FALSE;
1913          }          }
1914        else        else
1915          {          {
1916          SLJIT_ASSERT(!tmp2empty);          SLJIT_ASSERT(!tmp2empty);
1917          OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count], TMP2, 0);          OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), srcw[count], TMP2, 0);
1918          tmp2empty = stackptr >= stacktop;          tmp2empty = stackptr >= stacktop;
1919          if (!tmp2empty)          if (!tmp2empty)
1920            {            {
1921            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1922            stackptr += sizeof(sljit_w);            stackptr += sizeof(sljit_sw);
1923            }            }
1924          tmp1next = TRUE;          tmp1next = TRUE;
1925          }          }
1926        }        }
1927      }      }
1928    }    }
1929    while (status != end);
1930    
1931  if (save)  if (save)
1932    {    {
# Line 1192  if (save) Line 1935  if (save)
1935      if (!tmp1empty)      if (!tmp1empty)
1936        {        {
1937        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1938        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1939        }        }
1940      if (!tmp2empty)      if (!tmp2empty)
1941        {        {
1942        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1943        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1944        }        }
1945      }      }
1946    else    else
# Line 1205  if (save) Line 1948  if (save)
1948      if (!tmp2empty)      if (!tmp2empty)
1949        {        {
1950        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1951        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1952        }        }
1953      if (!tmp1empty)      if (!tmp1empty)
1954        {        {
1955        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1956        stackptr += sizeof(sljit_w);        stackptr += sizeof(sljit_sw);
1957        }        }
1958      }      }
1959    }    }
1960  SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));  SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));
1961  }  }
1962    
1963  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)
1964    {
1965    pcre_uchar *end = bracketend(cc);
1966    BOOL has_alternatives = cc[GET(cc, 1)] == OP_ALT;
1967    
1968    /* Assert captures then. */
1969    if (*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT)
1970      current_offset = NULL;
1971    /* Conditional block does not. */
1972    if (*cc == OP_COND || *cc == OP_SCOND)
1973      has_alternatives = FALSE;
1974    
1975    cc = next_opcode(common, cc);
1976    if (has_alternatives)
1977      current_offset = common->then_offsets + (cc - common->start);
1978    
1979    while (cc < end)
1980      {
1981      if ((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND))
1982        cc = set_then_offsets(common, cc, current_offset);
1983      else
1984        {
1985        if (*cc == OP_ALT && has_alternatives)
1986          current_offset = common->then_offsets + (cc + 1 + LINK_SIZE - common->start);
1987        if (*cc >= OP_THEN && *cc <= OP_THEN_ARG && current_offset != NULL)
1988          *current_offset = 1;
1989        cc = next_opcode(common, cc);
1990        }
1991      }
1992    
1993    return end;
1994    }
1995    
1996    #undef CASE_ITERATOR_PRIVATE_DATA_1
1997    #undef CASE_ITERATOR_PRIVATE_DATA_2A
1998    #undef CASE_ITERATOR_PRIVATE_DATA_2B
1999    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_1
2000    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
2001    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
2002    
2003    static SLJIT_INLINE BOOL is_powerof2(unsigned int value)
2004  {  {
2005  return (value & (value - 1)) == 0;  return (value & (value - 1)) == 0;
2006  }  }
# Line 1227  static SLJIT_INLINE void set_jumps(jump_ Line 2010  static SLJIT_INLINE void set_jumps(jump_
2010  while (list)  while (list)
2011    {    {
2012    /* sljit_set_label is clever enough to do nothing    /* sljit_set_label is clever enough to do nothing
2013    if either the jump or the label is NULL */    if either the jump or the label is NULL. */
2014    sljit_set_label(list->jump, label);    SET_LABEL(list->jump, label);
2015    list = list->next;    list = list->next;
2016    }    }
2017  }  }
2018    
2019  static SLJIT_INLINE void add_jump(struct sljit_compiler *compiler, jump_list **list, struct sljit_jump* jump)  static SLJIT_INLINE void add_jump(struct sljit_compiler *compiler, jump_list **list, struct sljit_jump *jump)
2020  {  {
2021  jump_list *list_item = sljit_alloc_memory(compiler, sizeof(jump_list));  jump_list *list_item = sljit_alloc_memory(compiler, sizeof(jump_list));
2022  if (list_item)  if (list_item)
# Line 1244  if (list_item) Line 2027  if (list_item)
2027    }    }
2028  }  }
2029    
2030  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)
2031  {  {
2032  DEFINE_COMPILER;  DEFINE_COMPILER;
2033  stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));  stub_list *list_item = sljit_alloc_memory(compiler, sizeof(stub_list));
2034    
2035  if (list_item)  if (list_item)
2036    {    {
   list_item->type = type;  
   list_item->data = data;  
2037    list_item->start = start;    list_item->start = start;
2038    list_item->leave = LABEL();    list_item->quit = LABEL();
2039    list_item->next = common->stubs;    list_item->next = common->stubs;
2040    common->stubs = list_item;    common->stubs = list_item;
2041    }    }
# Line 1263  if (list_item) Line 2044  if (list_item)
2044  static void flush_stubs(compiler_common *common)  static void flush_stubs(compiler_common *common)
2045  {  {
2046  DEFINE_COMPILER;  DEFINE_COMPILER;
2047  stub_list* list_item = common->stubs;  stub_list *list_item = common->stubs;
2048    
2049  while (list_item)  while (list_item)
2050    {    {
2051    JUMPHERE(list_item->start);    JUMPHERE(list_item->start);
2052    switch(list_item->type)    add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));
2053      {    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);  
2054    list_item = list_item->next;    list_item = list_item->next;
2055    }    }
2056  common->stubs = NULL;  common->stubs = NULL;
2057  }  }
2058    
2059  static SLJIT_INLINE void decrease_call_count(compiler_common *common)  static void add_label_addr(compiler_common *common, sljit_uw *update_addr)
2060    {
2061    DEFINE_COMPILER;
2062    label_addr_list *label_addr;
2063    
2064    label_addr = sljit_alloc_memory(compiler, sizeof(label_addr_list));
2065    if (label_addr == NULL)
2066      return;
2067    label_addr->label = LABEL();
2068    label_addr->update_addr = update_addr;
2069    label_addr->next = common->label_addrs;
2070    common->label_addrs = label_addr;
2071    }
2072    
2073    static SLJIT_INLINE void count_match(compiler_common *common)
2074  {  {
2075  DEFINE_COMPILER;  DEFINE_COMPILER;
2076    
2077  OP2(SLJIT_SUB | SLJIT_SET_E, CALL_COUNT, 0, CALL_COUNT, 0, SLJIT_IMM, 1);  OP2(SLJIT_SUB | SLJIT_SET_E, COUNT_MATCH, 0, COUNT_MATCH, 0, SLJIT_IMM, 1);
2078  add_jump(compiler, &common->calllimit, JUMP(SLJIT_C_ZERO));  add_jump(compiler, &common->calllimit, JUMP(SLJIT_ZERO));
2079  }  }
2080    
2081  static SLJIT_INLINE void allocate_stack(compiler_common *common, int size)  static SLJIT_INLINE void allocate_stack(compiler_common *common, int size)
# Line 1293  static SLJIT_INLINE void allocate_stack( Line 2083  static SLJIT_INLINE void allocate_stack(
2083  /* May destroy all locals and registers except TMP2. */  /* May destroy all locals and registers except TMP2. */
2084  DEFINE_COMPILER;  DEFINE_COMPILER;
2085    
2086  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));
2087  #ifdef DESTROY_REGISTERS  #ifdef DESTROY_REGISTERS
2088  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);
2089  OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);  OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);
2090  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);  OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);
2091  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), LOCALS0, TMP1, 0);
2092  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), LOCALS1, TMP1, 0);
2093  #endif  #endif
2094  add_stub(common, stack_alloc, 0, CMP(SLJIT_C_GREATER, STACK_TOP, 0, STACK_LIMIT, 0));  add_stub(common, CMP(SLJIT_GREATER, STACK_TOP, 0, STACK_LIMIT, 0));
2095  }  }
2096    
2097  static SLJIT_INLINE void free_stack(compiler_common *common, int size)  static SLJIT_INLINE void free_stack(compiler_common *common, int size)
2098  {  {
2099  DEFINE_COMPILER;  DEFINE_COMPILER;
2100  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));
2101    }
2102    
2103    static sljit_uw * allocate_read_only_data(compiler_common *common, sljit_uw size)
2104    {
2105    DEFINE_COMPILER;
2106    sljit_uw *result;
2107    
2108    if (SLJIT_UNLIKELY(sljit_get_compiler_error(compiler)))
2109      return NULL;
2110    
2111    result = (sljit_uw *)SLJIT_MALLOC(size + sizeof(sljit_uw), common->allocator_data);
2112    if (SLJIT_UNLIKELY(result == NULL))
2113      {
2114      sljit_set_compiler_memory_error(compiler);
2115      return NULL;
2116      }
2117    
2118    *(void**)result = common->read_only_data_head;
2119    common->read_only_data_head = (void *)result;
2120    return result + 1;
2121    }
2122    
2123    static void free_read_only_data(void *current, void *allocator_data)
2124    {
2125    void *next;
2126    
2127    SLJIT_UNUSED_ARG(allocator_data);
2128    
2129    while (current != NULL)
2130      {
2131      next = *(void**)current;
2132      SLJIT_FREE(current, allocator_data);
2133      current = next;
2134      }
2135  }  }
2136    
2137  static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)  static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)
# Line 1315  static SLJIT_INLINE void reset_ovector(c Line 2139  static SLJIT_INLINE void reset_ovector(c
2139  DEFINE_COMPILER;  DEFINE_COMPILER;
2140  struct sljit_label *loop;  struct sljit_label *loop;
2141  int i;  int i;
2142    
2143  /* At this point we can freely use all temporary registers. */  /* At this point we can freely use all temporary registers. */
2144    SLJIT_ASSERT(length > 1);
2145  /* TMP1 returns with begin - 1. */  /* TMP1 returns with begin - 1. */
2146  OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG1, 0, SLJIT_MEM1(SLJIT_SAVED_REG1), SLJIT_OFFSETOF(jit_arguments, begin), SLJIT_IMM, IN_UCHARS(1));  OP2(SLJIT_SUB, SLJIT_R0, 0, SLJIT_MEM1(SLJIT_S0), SLJIT_OFFSETOF(jit_arguments, begin), SLJIT_IMM, IN_UCHARS(1));
2147    if (length < 8)
2148      {
2149      for (i = 1; i < length; i++)
2150        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), OVECTOR(i), SLJIT_R0, 0);
2151      }
2152    else
2153      {
2154      GET_LOCAL_BASE(SLJIT_R1, 0, OVECTOR_START);
2155      OP1(SLJIT_MOV, SLJIT_R2, 0, SLJIT_IMM, length - 1);
2156      loop = LABEL();
2157      OP1(SLJIT_MOVU, SLJIT_MEM1(SLJIT_R1), sizeof(sljit_sw), SLJIT_R0, 0);
2158      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_IMM, 1);
2159      JUMPTO(SLJIT_NOT_ZERO, loop);
2160      }
2161    }
2162    
2163    static SLJIT_INLINE void do_reset_match(compiler_common *common, int length)
2164    {
2165    DEFINE_COMPILER;
2166    struct sljit_label *loop;
2167    int i;
2168    
2169    SLJIT_ASSERT(length > 1);
2170    /* OVECTOR(1) contains the "string begin - 1" constant. */
2171    if (length > 2)
2172      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(1));
2173  if (length < 8)  if (length < 8)
2174    {    {
2175    for (i = 0; i < length; i++)    for (i = 2; i < length; i++)
2176      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), SLJIT_TEMPORARY_REG1, 0);      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), OVECTOR(i), TMP1, 0);
2177    }    }
2178  else  else
2179    {    {
2180    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));
2181    OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, length);    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_IMM, length - 2);
2182    loop = LABEL();    loop = LABEL();
2183    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);
2184    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);
2185    JUMPTO(SLJIT_C_NOT_ZERO, loop);    JUMPTO(SLJIT_NOT_ZERO, loop);
2186      }
2187    
2188    OP1(SLJIT_MOV, STACK_TOP, 0, ARGUMENTS, 0);
2189    if (common->mark_ptr != 0)
2190      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->mark_ptr, SLJIT_IMM, 0);
2191    if (common->control_head_ptr != 0)
2192      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->control_head_ptr, SLJIT_IMM, 0);
2193    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(jit_arguments, stack));
2194    OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->start_ptr);
2195    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(struct sljit_stack, base));
2196    }
2197    
2198    static sljit_sw SLJIT_CALL do_search_mark(sljit_sw *current, const pcre_uchar *skip_arg)
2199    {
2200    while (current != NULL)
2201      {
2202      switch (current[-2])
2203        {
2204        case type_then_trap:
2205        break;
2206    
2207        case type_mark:
2208        if (STRCMP_UC_UC(skip_arg, (pcre_uchar *)current[-3]) == 0)
2209          return current[-4];
2210        break;
2211    
2212        default:
2213        SLJIT_ASSERT_STOP();
2214        break;
2215        }
2216      current = (sljit_sw*)current[-1];
2217    }    }
2218    return -1;
2219  }  }
2220    
2221  static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)  static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)
2222  {  {
2223  DEFINE_COMPILER;  DEFINE_COMPILER;
2224  struct sljit_label *loop;  struct sljit_label *loop;
2225  struct sljit_jump *earlyexit;  struct sljit_jump *early_quit;
2226    
2227  /* At this point we can freely use all registers. */  /* At this point we can freely use all registers. */
2228  OP1(SLJIT_MOV, SLJIT_SAVED_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));  OP1(SLJIT_MOV, SLJIT_S2, 0, SLJIT_MEM1(SLJIT_SP), OVECTOR(1));
2229  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1), STR_PTR, 0);  OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), OVECTOR(1), STR_PTR, 0);
2230    
2231  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, ARGUMENTS, 0);  OP1(SLJIT_MOV, SLJIT_R0, 0, ARGUMENTS, 0);
2232  if (common->mark_ptr != 0)  if (common->mark_ptr != 0)
2233    OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);    OP1(SLJIT_MOV, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_SP), common->mark_ptr);
2234  OP1(SLJIT_MOV_SI, SLJIT_TEMPORARY_REG2, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, offsetcount));  OP1(SLJIT_MOV_SI, SLJIT_R1, 0, SLJIT_MEM1(SLJIT_R0), SLJIT_OFFSETOF(jit_arguments, offset_count));
2235  if (common->mark_ptr != 0)  if (common->mark_ptr != 0)
2236    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, mark_ptr), SLJIT_TEMPORARY_REG3, 0);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_R0), SLJIT_OFFSETOF(jit_arguments, mark_ptr), SLJIT_R2, 0);
2237  OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));  OP2(SLJIT_SUB, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_R0), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));
2238  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, begin));  OP1(SLJIT_MOV, SLJIT_R0, 0, SLJIT_MEM1(SLJIT_R0), SLJIT_OFFSETOF(jit_arguments, begin));
2239  OP2(SLJIT_ADD, SLJIT_SAVED_REG1, 0, SLJIT_LOCALS_REG, 0, SLJIT_IMM, OVECTOR_START);  GET_LOCAL_BASE(SLJIT_S0, 0, OVECTOR_START);
2240  /* Unlikely, but possible */  /* Unlikely, but possible */
2241  earlyexit = CMP(SLJIT_C_EQUAL, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, 0);  early_quit = CMP(SLJIT_EQUAL, SLJIT_R1, 0, SLJIT_IMM, 0);
2242  loop = LABEL();  loop = LABEL();
2243  OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, SLJIT_MEM1(SLJIT_SAVED_REG1), 0, SLJIT_TEMPORARY_REG1, 0);  OP2(SLJIT_SUB, SLJIT_S1, 0, SLJIT_MEM1(SLJIT_S0), 0, SLJIT_R0, 0);
2244  OP2(SLJIT_ADD, SLJIT_SAVED_REG1, 0, SLJIT_SAVED_REG1, 0, SLJIT_IMM, sizeof(sljit_w));  OP2(SLJIT_ADD, SLJIT_S0, 0, SLJIT_S0, 0, SLJIT_IMM, sizeof(sljit_sw));
2245  /* Copy the integer value to the output buffer */  /* Copy the integer value to the output buffer */
2246  #ifdef COMPILE_PCRE16  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2247  OP2(SLJIT_ASHR, SLJIT_SAVED_REG2, 0, SLJIT_SAVED_REG2, 0, SLJIT_IMM, 1);  OP2(SLJIT_ASHR, SLJIT_S1, 0, SLJIT_S1, 0, SLJIT_IMM, UCHAR_SHIFT);
2248  #endif  #endif
2249  OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_TEMPORARY_REG3), sizeof(int), SLJIT_SAVED_REG2, 0);  OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_R2), sizeof(int), SLJIT_S1, 0);
2250  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_TEMPORARY_REG2, 0, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, 1);  OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_R1, 0, SLJIT_R1, 0, SLJIT_IMM, 1);
2251  JUMPTO(SLJIT_C_NOT_ZERO, loop);  JUMPTO(SLJIT_NOT_ZERO, loop);
2252  JUMPHERE(earlyexit);  JUMPHERE(early_quit);
2253    
2254  /* Calculate the return value, which is the maximum ovector value. */  /* Calculate the return value, which is the maximum ovector value. */
2255  if (topbracket > 1)  if (topbracket > 1)
2256    {    {
2257    OP2(SLJIT_ADD, SLJIT_TEMPORARY_REG1, 0, SLJIT_LOCALS_REG, 0, SLJIT_IMM, OVECTOR_START + topbracket * 2 * sizeof(sljit_w));    GET_LOCAL_BASE(SLJIT_R0, 0, OVECTOR_START + topbracket * 2 * sizeof(sljit_sw));
2258    OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, topbracket + 1);    OP1(SLJIT_MOV, SLJIT_R1, 0, SLJIT_IMM, topbracket + 1);
2259    
2260    /* OVECTOR(0) is never equal to SLJIT_SAVED_REG3. */    /* OVECTOR(0) is never equal to SLJIT_S2. */
2261    loop = LABEL();    loop = LABEL();
2262    OP1(SLJIT_MOVU, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), -(2 * (sljit_w)sizeof(sljit_w)));    OP1(SLJIT_MOVU, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_R0), -(2 * (sljit_sw)sizeof(sljit_sw)));
2263    OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG2, 0, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, 1);    OP2(SLJIT_SUB, SLJIT_R1, 0, SLJIT_R1, 0, SLJIT_IMM, 1);
2264    CMPTO(SLJIT_C_EQUAL, SLJIT_TEMPORARY_REG3, 0, SLJIT_SAVED_REG3, 0, loop);    CMPTO(SLJIT_EQUAL, SLJIT_R2, 0, SLJIT_S2, 0, loop);
2265    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_TEMPORARY_REG2, 0);    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_R1, 0);
2266    }    }
2267  else  else
2268    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, 1);    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, 1);
2269  }  }
2270    
2271  static SLJIT_INLINE void return_with_partial_match(compiler_common *common, struct sljit_label *leave)  static SLJIT_INLINE void return_with_partial_match(compiler_common *common, struct sljit_label *quit)
2272  {  {
2273  DEFINE_COMPILER;  DEFINE_COMPILER;
2274    struct sljit_jump *jump;
2275    
2276  SLJIT_COMPILE_ASSERT(STR_END == SLJIT_SAVED_REG2, str_end_must_be_saved_reg2);  SLJIT_COMPILE_ASSERT(STR_END == SLJIT_S1, str_end_must_be_saved_reg2);
2277  SLJIT_ASSERT(common->start_used_ptr != 0 && (common->mode == JIT_PARTIAL_SOFT_COMPILE ? common->hit_start != 0 : common->hit_start == 0));  SLJIT_ASSERT(common->start_used_ptr != 0 && common->start_ptr != 0
2278      && (common->mode == JIT_PARTIAL_SOFT_COMPILE ? common->hit_start != 0 : common->hit_start == 0));
2279    
2280  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG2, 0, ARGUMENTS, 0);  OP1(SLJIT_MOV, SLJIT_R1, 0, ARGUMENTS, 0);
2281  OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, PCRE_ERROR_PARTIAL);  OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, PCRE_ERROR_PARTIAL);
2282  OP1(SLJIT_MOV_SI, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), SLJIT_OFFSETOF(jit_arguments, offsetcount));  OP1(SLJIT_MOV_SI, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_R1), SLJIT_OFFSETOF(jit_arguments, real_offset_count));
2283  CMPTO(SLJIT_C_LESS, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, 2, leave);  CMPTO(SLJIT_SIG_LESS, SLJIT_R2, 0, SLJIT_IMM, 2, quit);
2284    
2285  /* Store match begin and end. */  /* Store match begin and end. */
2286  OP1(SLJIT_MOV, SLJIT_SAVED_REG1, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), SLJIT_OFFSETOF(jit_arguments, begin));  OP1(SLJIT_MOV, SLJIT_S0, 0, SLJIT_MEM1(SLJIT_R1), SLJIT_OFFSETOF(jit_arguments, begin));
2287  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG2, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), SLJIT_OFFSETOF(jit_arguments, offsets));  OP1(SLJIT_MOV, SLJIT_R1, 0, SLJIT_MEM1(SLJIT_R1), SLJIT_OFFSETOF(jit_arguments, offsets));
2288  OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mode == JIT_PARTIAL_HARD_COMPILE ? common->start_used_ptr : common->hit_start);  
2289  OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, STR_END, 0, SLJIT_SAVED_REG1, 0);  jump = CMP(SLJIT_SIG_LESS, SLJIT_R2, 0, SLJIT_IMM, 3);
2290  #ifdef COMPILE_PCRE16  OP2(SLJIT_SUB, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_SP), common->mode == JIT_PARTIAL_HARD_COMPILE ? common->start_ptr : (common->hit_start + (int)sizeof(sljit_sw)), SLJIT_S0, 0);
2291  OP2(SLJIT_ASHR, SLJIT_SAVED_REG2, 0, SLJIT_SAVED_REG2, 0, SLJIT_IMM, 1);  #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2292  #endif  OP2(SLJIT_ASHR, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_IMM, UCHAR_SHIFT);
2293  OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), sizeof(int), SLJIT_SAVED_REG2, 0);  #endif
2294    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_R1), 2 * sizeof(int), SLJIT_R2, 0);
2295  OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG3, 0, SLJIT_TEMPORARY_REG3, 0, SLJIT_SAVED_REG1, 0);  JUMPHERE(jump);
2296  #ifdef COMPILE_PCRE16  
2297  OP2(SLJIT_ASHR, SLJIT_TEMPORARY_REG3, 0, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, 1);  OP1(SLJIT_MOV, SLJIT_R2, 0, SLJIT_MEM1(SLJIT_SP), common->mode == JIT_PARTIAL_HARD_COMPILE ? common->start_used_ptr : common->hit_start);
2298    OP2(SLJIT_SUB, SLJIT_S1, 0, STR_END, 0, SLJIT_S0, 0);
2299    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2300    OP2(SLJIT_ASHR, SLJIT_S1, 0, SLJIT_S1, 0, SLJIT_IMM, UCHAR_SHIFT);
2301    #endif
2302    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_R1), sizeof(int), SLJIT_S1, 0);
2303    
2304    OP2(SLJIT_SUB, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_S0, 0);
2305    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2306    OP2(SLJIT_ASHR, SLJIT_R2, 0, SLJIT_R2, 0, SLJIT_IMM, UCHAR_SHIFT);
2307  #endif  #endif
2308  OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), 0, SLJIT_TEMPORARY_REG3, 0);  OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_R1), 0, SLJIT_R2, 0);
2309    
2310  JUMPTO(SLJIT_JUMP, leave);  JUMPTO(SLJIT_JUMP, quit);
2311  }  }
2312    
2313  static SLJIT_INLINE void check_start_used_ptr(compiler_common *common)  static SLJIT_INLINE void check_start_used_ptr(compiler_common *common)
# Line 1424  struct sljit_jump *jump; Line 2319  struct sljit_jump *jump;
2319  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2320    {    {
2321    /* The value of -1 must be kept for start_used_ptr! */    /* The value of -1 must be kept for start_used_ptr! */
2322    OP2(SLJIT_ADD, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, SLJIT_IMM, 1);    OP2(SLJIT_ADD, TMP1, 0, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, SLJIT_IMM, 1);
2323    /* Jumps if start_used_ptr < STR_PTR, or start_used_ptr == -1. Although overwriting    /* Jumps if start_used_ptr < STR_PTR, or start_used_ptr == -1. Although overwriting
2324    is not necessary if start_used_ptr == STR_PTR, it does not hurt as well. */    is not necessary if start_used_ptr == STR_PTR, it does not hurt as well. */
2325    jump = CMP(SLJIT_C_LESS_EQUAL, TMP1, 0, STR_PTR, 0);    jump = CMP(SLJIT_LESS_EQUAL, TMP1, 0, STR_PTR, 0);
2326    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0);
2327    JUMPHERE(jump);    JUMPHERE(jump);
2328    }    }
2329  else if (common->mode == JIT_PARTIAL_HARD_COMPILE)  else if (common->mode == JIT_PARTIAL_HARD_COMPILE)
2330    {    {
2331    jump = CMP(SLJIT_C_LESS_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);    jump = CMP(SLJIT_LESS_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0);
2332    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0);
2333    JUMPHERE(jump);    JUMPHERE(jump);
2334    }    }
2335  }  }
2336    
2337  static SLJIT_INLINE BOOL char_has_othercase(compiler_common *common, pcre_uchar* cc)  static SLJIT_INLINE BOOL char_has_othercase(compiler_common *common, pcre_uchar *cc)
2338  {  {
2339  /* Detects if the character has an othercase. */  /* Detects if the character has an othercase. */
2340  unsigned int c;  unsigned int c;
# Line 1482  if (common->utf && c > 127) Line 2377  if (common->utf && c > 127)
2377  return TABLE_GET(c, common->fcc, c);  return TABLE_GET(c, common->fcc, c);
2378  }  }
2379    
2380  static unsigned int char_get_othercase_bit(compiler_common *common, pcre_uchar* cc)  static unsigned int char_get_othercase_bit(compiler_common *common, pcre_uchar *cc)
2381  {  {
2382  /* Detects if the character and its othercase has only 1 bit difference. */  /* Detects if the character and its othercase has only 1 bit difference. */
2383  unsigned int c, oc, bit;  unsigned int c, oc, bit;
# Line 1523  if (c <= 127 && bit == 0x20) Line 2418  if (c <= 127 && bit == 0x20)
2418    return (0 << 8) | 0x20;    return (0 << 8) | 0x20;
2419    
2420  /* Since c != oc, they must have at least 1 bit difference. */  /* Since c != oc, they must have at least 1 bit difference. */
2421  if (!ispowerof2(bit))  if (!is_powerof2(bit))
2422    return 0;    return 0;
2423    
2424  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
2425    
2426  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
2427  if (common->utf && c > 127)  if (common->utf && c > 127)
# Line 1542  if (common->utf && c > 127) Line 2437  if (common->utf && c > 127)
2437  #endif /* SUPPORT_UTF */  #endif /* SUPPORT_UTF */
2438  return (0 << 8) | bit;  return (0 << 8) | bit;
2439    
2440  #else /* COMPILE_PCRE8 */  #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2441    
 #ifdef COMPILE_PCRE16  
2442  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
2443  if (common->utf && c > 65535)  if (common->utf && c > 65535)
2444    {    {
# Line 1555  if (common->utf && c > 65535) Line 2449  if (common->utf && c > 65535)
2449    }    }
2450  #endif /* SUPPORT_UTF */  #endif /* SUPPORT_UTF */
2451  return (bit < 256) ? ((0 << 8) | bit) : ((1 << 8) | (bit >> 8));  return (bit < 256) ? ((0 << 8) | bit) : ((1 << 8) | (bit >> 8));
 #endif /* COMPILE_PCRE16 */  
2452    
2453  #endif /* COMPILE_PCRE8 */  #endif /* COMPILE_PCRE[8|16|32] */
2454  }  }
2455    
2456  static void check_partial(compiler_common *common, BOOL force)  static void check_partial(compiler_common *common, BOOL force)
2457  {  {
2458  /* Checks whether a partial matching is occured. Does not modify registers. */  /* Checks whether a partial matching is occurred. Does not modify registers. */
2459  DEFINE_COMPILER;  DEFINE_COMPILER;
2460  struct sljit_jump *jump = NULL;  struct sljit_jump *jump = NULL;
2461    
# Line 1572  if (common->mode == JIT_COMPILE) Line 2465  if (common->mode == JIT_COMPILE)
2465    return;    return;
2466    
2467  if (!force)  if (!force)
2468    jump = CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);    jump = CMP(SLJIT_GREATER_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0);
2469  else if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  else if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2470    jump = CMP(SLJIT_C_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, SLJIT_IMM, -1);    jump = CMP(SLJIT_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, SLJIT_IMM, -1);
2471    
2472  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2473    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, -1);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->hit_start, SLJIT_IMM, 0);
2474  else  else
2475    {    {
2476    if (common->partialmatchlabel != NULL)    if (common->partialmatchlabel != NULL)
# Line 1590  if (jump != NULL) Line 2483  if (jump != NULL)
2483    JUMPHERE(jump);    JUMPHERE(jump);
2484  }  }
2485    
2486  static struct sljit_jump *check_str_end(compiler_common *common)  static void check_str_end(compiler_common *common, jump_list **end_reached)
2487  {  {
2488  /* Does not affect registers. Usually used in a tight spot. */  /* Does not affect registers. Usually used in a tight spot. */
2489  DEFINE_COMPILER;  DEFINE_COMPILER;
2490  struct sljit_jump *jump;  struct sljit_jump *jump;
 struct sljit_jump *nohit;  
 struct sljit_jump *return_value;  
2491    
2492  if (common->mode == JIT_COMPILE)  if (common->mode == JIT_COMPILE)
2493    return CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);    {
2494      add_jump(compiler, end_reached, CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2495      return;
2496      }
2497    
2498  jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);  jump = CMP(SLJIT_LESS, STR_PTR, 0, STR_END, 0);
2499  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2500    {    {
2501    nohit = CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);    add_jump(compiler, end_reached, CMP(SLJIT_GREATER_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0));
2502    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, -1);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->hit_start, SLJIT_IMM, 0);
2503    JUMPHERE(nohit);    add_jump(compiler, end_reached, JUMP(SLJIT_JUMP));
   return_value = JUMP(SLJIT_JUMP);  
2504    }    }
2505  else  else
2506    {    {
2507    return_value = CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);    add_jump(compiler, end_reached, CMP(SLJIT_GREATER_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0));
2508    if (common->partialmatchlabel != NULL)    if (common->partialmatchlabel != NULL)
2509      JUMPTO(SLJIT_JUMP, common->partialmatchlabel);      JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2510    else    else
2511      add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));      add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2512    }    }
2513  JUMPHERE(jump);  JUMPHERE(jump);
 return return_value;  
2514  }  }
2515    
2516  static void fallback_at_str_end(compiler_common *common, jump_list **fallbacks)  static void detect_partial_match(compiler_common *common, jump_list **backtracks)
2517  {  {
2518  DEFINE_COMPILER;  DEFINE_COMPILER;
2519  struct sljit_jump *jump;  struct sljit_jump *jump;
2520    
2521  if (common->mode == JIT_COMPILE)  if (common->mode == JIT_COMPILE)
2522    {    {
2523    add_jump(compiler, fallbacks, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));    add_jump(compiler, backtracks, CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2524    return;    return;
2525    }    }
2526    
2527  /* Partial matching mode. */  /* Partial matching mode. */
2528  jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);  jump = CMP(SLJIT_LESS, STR_PTR, 0, STR_END, 0);
2529  add_jump(compiler, fallbacks, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));  add_jump(compiler, backtracks, CMP(SLJIT_GREATER_EQUAL, SLJIT_MEM1(SLJIT_SP), common->start_used_ptr, STR_PTR, 0));
2530  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)  if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2531    {    {
2532    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, -1);    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->hit_start, SLJIT_IMM, 0);
2533    add_jump(compiler, fallbacks, JUMP(SLJIT_JUMP));    add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));
2534    }    }
2535  else  else
2536    {    {
# Line 1650  else Line 2542  else
2542  JUMPHERE(jump);  JUMPHERE(jump);
2543  }  }
2544    
2545  static void read_char(compiler_common *common)  static void peek_char(compiler_common *common, pcre_uint32 max)
2546  {  {
2547  /* Reads the character into TMP1, updates STR_PTR.  /* Reads the character into TMP1, keeps STR_PTR.
2548  Does not check STR_END. TMP2 Destroyed. */  Does not check STR_END. TMP2 Destroyed. */
2549  DEFINE_COMPILER;  DEFINE_COMPILER;
2550  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2551  struct sljit_jump *jump;  struct sljit_jump *jump;
2552  #endif  #endif
2553    
2554    SLJIT_UNUSED_ARG(max);
2555    
2556  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2557  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2558  if (common->utf)  if (common->utf)
2559    {    {
2560  #ifdef COMPILE_PCRE8    if (max < 128) return;
2561    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);  
2562  #else    jump = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2563  #ifdef COMPILE_PCRE16    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
   jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);  
 #endif  
 #endif /* COMPILE_PCRE8 */  
2564    add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2565      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2566      JUMPHERE(jump);
2567      }
2568    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2569    
2570    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2571    if (common->utf)
2572      {
2573      if (max < 0xd800) return;
2574    
2575      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2576      jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2577      /* TMP2 contains the high surrogate. */
2578      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2579      OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2580      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2581      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2582      OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2583    JUMPHERE(jump);    JUMPHERE(jump);
2584    }    }
2585  #endif  #endif
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  
2586  }  }
2587    
2588  static void peek_char(compiler_common *common)  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2589    
2590    static BOOL is_char7_bitset(const pcre_uint8 *bitset, BOOL nclass)
2591  {  {
2592  /* Reads the character into TMP1, keeps STR_PTR.  /* Tells whether the character codes below 128 are enough
2593  Does not check STR_END. TMP2 Destroyed. */  to determine a match. */
2594    const pcre_uint8 value = nclass ? 0xff : 0;
2595    const pcre_uint8 *end = bitset + 32;
2596    
2597    bitset += 16;
2598    do
2599      {
2600      if (*bitset++ != value)
2601        return FALSE;
2602      }
2603    while (bitset < end);
2604    return TRUE;
2605    }
2606    
2607    static void read_char7_type(compiler_common *common, BOOL full_read)
2608    {
2609    /* Reads the precise character type of a character into TMP1, if the character
2610    is less than 128. Otherwise it returns with zero. Does not check STR_END. The
2611    full_read argument tells whether characters above max are accepted or not. */
2612  DEFINE_COMPILER;  DEFINE_COMPILER;
 #ifdef SUPPORT_UTF  
2613  struct sljit_jump *jump;  struct sljit_jump *jump;
 #endif  
2614    
2615  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  SLJIT_ASSERT(common->utf);
2616  #ifdef SUPPORT_UTF  
2617  if (common->utf)  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2618    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2619    
2620    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2621    
2622    if (full_read)
2623    {    {
2624  #ifdef COMPILE_PCRE8    jump = CMP(SLJIT_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2625    jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);    OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2626  #else    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
 #ifdef COMPILE_PCRE16  
   jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);  
 #endif  
 #endif /* COMPILE_PCRE8 */  
   add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));  
   OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);  
2627    JUMPHERE(jump);    JUMPHERE(jump);
2628    }    }
 #endif  
2629  }  }
2630    
2631  static void read_char8_type(compiler_common *common)  #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2632    
2633    static void read_char_range(compiler_common *common, pcre_uint32 min, pcre_uint32 max, BOOL update_str_ptr)
2634  {  {
2635  /* Reads the character type into TMP1, updates STR_PTR. Does not check STR_END. */  /* Reads the precise value of a character into TMP1, if the character is
2636    between min and max (c >= min && c <= max). Otherwise it returns with a value
2637    outside the range. Does not check STR_END. */
2638  DEFINE_COMPILER;  DEFINE_COMPILER;
2639  #if defined SUPPORT_UTF || defined COMPILE_PCRE16  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2640  struct sljit_jump *jump;  struct sljit_jump *jump;
2641  #endif  #endif
2642    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2643    struct sljit_jump *jump2;
2644    #endif
2645    
2646  #ifdef SUPPORT_UTF  SLJIT_UNUSED_ARG(update_str_ptr);
2647  if (common->utf)  SLJIT_UNUSED_ARG(min);
2648    {  SLJIT_UNUSED_ARG(max);
2649    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);  SLJIT_ASSERT(min <= max);
2650    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  
2651  #ifdef COMPILE_PCRE8  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2652    /* This can be an extra read in some situations, but hopefully  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2653    it is needed in most cases. */  
2654    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2655    jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);  if (common->utf)
2656    add_jump(compiler, &common->utfreadtype8, JUMP(SLJIT_FAST_CALL));    {
2657    JUMPHERE(jump);    if (max < 128 && !update_str_ptr) return;
2658  #else  
2659  #ifdef COMPILE_PCRE16    jump = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2660    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);    if (min >= 0x10000)
2661    jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);      {
2662    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xf0);
2663        if (update_str_ptr)
2664          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2665        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2666        jump2 = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0x7);
2667        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2668        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2669        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2670        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2671        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2672        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2673        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2674        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2675        if (!update_str_ptr)
2676          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2677        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2678        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2679        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2680        JUMPHERE(jump2);
2681        if (update_str_ptr)
2682          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2683        }
2684      else if (min >= 0x800 && max <= 0xffff)
2685        {
2686        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xe0);
2687        if (update_str_ptr)
2688          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2689        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2690        jump2 = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xf);
2691        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2692        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2693        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2694        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2695        if (!update_str_ptr)
2696          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2697        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2698        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2699        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2700        JUMPHERE(jump2);
2701        if (update_str_ptr)
2702          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2703        }
2704      else if (max >= 0x800)
2705        add_jump(compiler, (max < 0x10000) ? &common->utfreadchar16 : &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2706      else if (max < 128)
2707        {
2708        OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2709        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2710        }
2711      else
2712        {
2713        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2714        if (!update_str_ptr)
2715          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2716        else
2717          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2718        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2719        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2720        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2721        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2722        if (update_str_ptr)
2723          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2724        }
2725    JUMPHERE(jump);    JUMPHERE(jump);
2726    /* Skip low surrogate if necessary. */    }
   OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0xfc00);  
   OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0xd800);  
   COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_EQUAL);  
   OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 1);  
   OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);  
2727  #endif  #endif
2728  #endif /* COMPILE_PCRE8 */  
2729    return;  #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2730    if (common->utf)
2731      {
2732      if (max >= 0x10000)
2733        {
2734        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2735        jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2736        /* TMP2 contains the high surrogate. */
2737        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2738        OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2739        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2740        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2741        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2742        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2743        JUMPHERE(jump);
2744        return;
2745        }
2746    
2747      if (max < 0xd800 && !update_str_ptr) return;
2748    
2749      /* Skip low surrogate if necessary. */
2750      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2751      jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2752      if (update_str_ptr)
2753        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2754      if (max >= 0xd800)
2755        OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0x10000);
2756      JUMPHERE(jump);
2757    }    }
2758  #endif  #endif
2759    }
2760    
2761    static SLJIT_INLINE void read_char(compiler_common *common)
2762    {
2763    read_char_range(common, 0, READ_CHAR_MAX, TRUE);
2764    }
2765    
2766    static void read_char8_type(compiler_common *common, BOOL update_str_ptr)
2767    {
2768    /* Reads the character type into TMP1, updates STR_PTR. Does not check STR_END. */
2769    DEFINE_COMPILER;
2770    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2771    struct sljit_jump *jump;
2772    #endif
2773    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2774    struct sljit_jump *jump2;
2775    #endif
2776    
2777    SLJIT_UNUSED_ARG(update_str_ptr);
2778    
2779  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2780  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2781  #ifdef COMPILE_PCRE16  
2782    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2783    if (common->utf)
2784      {
2785      /* This can be an extra read in some situations, but hopefully
2786      it is needed in most cases. */
2787      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2788      jump = CMP(SLJIT_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2789      if (!update_str_ptr)
2790        {
2791        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2792        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2793        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2794        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2795        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2796        OP2(SLJIT_OR, TMP2, 0, TMP2, 0, TMP1, 0);
2797        OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2798        jump2 = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 255);
2799        OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2800        JUMPHERE(jump2);
2801        }
2802      else
2803        add_jump(compiler, &common->utfreadtype8, JUMP(SLJIT_FAST_CALL));
2804      JUMPHERE(jump);
2805      return;
2806      }
2807    #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2808    
2809    #if !defined COMPILE_PCRE8
2810  /* The ctypes array contains only 256 values. */  /* The ctypes array contains only 256 values. */
2811  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2812  jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);  jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 255);
2813  #endif  #endif
2814  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2815  #ifdef COMPILE_PCRE16  #if !defined COMPILE_PCRE8
2816  JUMPHERE(jump);  JUMPHERE(jump);
2817  #endif  #endif
2818    
2819    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2820    if (common->utf && update_str_ptr)
2821      {
2822      /* Skip low surrogate if necessary. */
2823      OP2(SLJIT_SUB, TMP2, 0, TMP2, 0, SLJIT_IMM, 0xd800);
2824      jump = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2825      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2826      JUMPHERE(jump);
2827      }
2828    #endif /* SUPPORT_UTF && COMPILE_PCRE16 */
2829  }  }
2830    
2831  static void skip_char_back(compiler_common *common)  static void skip_char_back(compiler_common *common)
2832  {  {
2833  /* Goes one character back. Affects STR_PTR and TMP1. Does not check begin. */  /* Goes one character back. Affects STR_PTR and TMP1. Does not check begin. */
2834  DEFINE_COMPILER;  DEFINE_COMPILER;
2835  #if defined SUPPORT_UTF && defined COMPILE_PCRE8  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2836    #if defined COMPILE_PCRE8
2837  struct sljit_label *label;  struct sljit_label *label;
2838    
2839  if (common->utf)  if (common->utf)
# Line 1767  if (common->utf) Line 2842  if (common->utf)
2842    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
2843    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2844    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xc0);    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xc0);
2845    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, 0x80, label);    CMPTO(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, 0x80, label);
2846    return;    return;
2847    }    }
2848  #endif  #elif defined COMPILE_PCRE16
 #if defined SUPPORT_UTF && defined COMPILE_PCRE16  
2849  if (common->utf)  if (common->utf)
2850    {    {
2851    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
# Line 1779  if (common->utf) Line 2853  if (common->utf)
2853    /* Skip low surrogate if necessary. */    /* Skip low surrogate if necessary. */
2854    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
2855    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xdc00);    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xdc00);
2856    COND_VALUE(SLJIT_MOV, TMP1, 0, SLJIT_C_EQUAL);    OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
2857    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
2858    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP1, 0);    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2859    return;    return;
2860    }    }
2861  #endif  #endif /* COMPILE_PCRE[8|16] */
2862    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2863  OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2864  }  }
2865    
2866  static void check_newlinechar(compiler_common *common, int nltype, jump_list **fallbacks, BOOL jumpiftrue)  static void check_newlinechar(compiler_common *common, int nltype, jump_list **backtracks, BOOL jumpifmatch)
2867  {  {
2868  /* 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. */
2869  DEFINE_COMPILER;  DEFINE_COMPILER;
2870    struct sljit_jump *jump;
2871    
2872  if (nltype == NLTYPE_ANY)  if (nltype == NLTYPE_ANY)
2873    {    {
2874    add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));    add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));
2875    add_jump(compiler, fallbacks, JUMP(jumpiftrue ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));    add_jump(compiler, backtracks, JUMP(jumpifmatch ? SLJIT_NOT_ZERO : SLJIT_ZERO));
2876    }    }
2877  else if (nltype == NLTYPE_ANYCRLF)  else if (nltype == NLTYPE_ANYCRLF)
2878    {    {
2879    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_CR);    if (jumpifmatch)
2880    COND_VALUE(SLJIT_MOV, TMP2, 0, SLJIT_C_EQUAL);      {
2881    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, CHAR_NL);      add_jump(compiler, backtracks, CMP(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR));
2882    COND_VALUE(SLJIT_OR | SLJIT_SET_E, TMP2, 0, SLJIT_C_EQUAL);      add_jump(compiler, backtracks, CMP(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL));
2883    add_jump(compiler, fallbacks, JUMP(jumpiftrue ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));      }
2884      else
2885        {
2886        jump = CMP(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR);
2887        add_jump(compiler, backtracks, CMP(SLJIT_NOT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL));
2888        JUMPHERE(jump);
2889        }
2890    }    }
2891  else  else
2892    {    {
2893    SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline < 256);    SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline < 256);
2894    add_jump(compiler, fallbacks, CMP(jumpiftrue ? SLJIT_C_EQUAL : SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, common->newline));    add_jump(compiler, backtracks, CMP(jumpifmatch ? SLJIT_EQUAL : SLJIT_NOT_EQUAL, TMP1, 0, SLJIT_IMM, common->newline));
2895    }    }
2896  }  }
2897    
2898  #ifdef SUPPORT_UTF  #ifdef SUPPORT_UTF
2899    
2900  #ifdef COMPILE_PCRE8  #if defined COMPILE_PCRE8
2901  static void do_utfreadchar(compiler_common *common)  static void do_utfreadchar(compiler_common *common)
2902  {  {
2903  /* Fast decoding a UTF-8 character. TMP1 contains the first byte  /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2904  of the character (>= 0xc0). Return char value in TMP1, length - 1 in TMP2. */  of the character (>= 0xc0). Return char value in TMP1, length in TMP2. */
2905  DEFINE_COMPILER;  DEFINE_COMPILER;
2906  struct sljit_jump *jump;  struct sljit_jump *jump;
2907    
2908  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2909  /* Searching for the first zero. */  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2910  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x20);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
 jump = JUMP(SLJIT_C_NOT_ZERO);  
 /* Two byte sequence. */  
 OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  
 OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x1f);  
2911  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2912  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2913  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2914  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(1));  
2915    /* Searching for the first zero. */
2916    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
2917    jump = JUMP(SLJIT_NOT_ZERO);
2918    /* Two byte sequence. */
2919    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2920    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(2));
2921  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
 JUMPHERE(jump);  
2922    
2923  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x10);  JUMPHERE(jump);
 jump = JUMP(SLJIT_C_NOT_ZERO);  
 /* Three byte sequence. */  
2924  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2925  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x0f);  OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
2926  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 12);  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2927  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
 OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  
2928  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2929  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));  
2930    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2931    jump = JUMP(SLJIT_NOT_ZERO);
2932    /* Three byte sequence. */
2933  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2934  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(3));
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
 OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(2));  
2935  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
 JUMPHERE(jump);  
2936    
2937  /* Four byte sequence. */  /* Four byte sequence. */
2938  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  JUMPHERE(jump);
2939  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x07);  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2940  OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 18);  OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2941    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2942    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2943  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
 OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 12);  
2944  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2945  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(4));
2946    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2947    }
2948    
2949    static void do_utfreadchar16(compiler_common *common)
2950    {
2951    /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2952    of the character (>= 0xc0). Return value in TMP1. */
2953    DEFINE_COMPILER;
2954    struct sljit_jump *jump;
2955    
2956    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2957    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2958    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2959    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2960  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
 OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  
2961  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2962  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(3));  
2963  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));  /* Searching for the first zero. */
2964    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
2965    jump = JUMP(SLJIT_NOT_ZERO);
2966    /* Two byte sequence. */
2967    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2968    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2969    
2970    JUMPHERE(jump);
2971    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x400);
2972    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_NOT_ZERO);
2973    /* This code runs only in 8 bit mode. No need to shift the value. */
2974    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2975    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2976    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
2977    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2978  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2979  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2980  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(3));  /* Three byte sequence. */
2981    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2982  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2983  }  }
2984    
# Line 1882  DEFINE_COMPILER; Line 2990  DEFINE_COMPILER;
2990  struct sljit_jump *jump;  struct sljit_jump *jump;
2991  struct sljit_jump *compare;  struct sljit_jump *compare;
2992    
2993  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2994    
2995  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);
2996  jump = JUMP(SLJIT_C_NOT_ZERO);  jump = JUMP(SLJIT_NOT_ZERO);
2997  /* Two byte sequence. */  /* Two byte sequence. */
2998  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2999  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3000  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);
3001    /* The upper 5 bits are known at this point. */
3002    compare = CMP(SLJIT_GREATER, TMP2, 0, SLJIT_IMM, 0x3);
3003  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
3004  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
3005  OP2(SLJIT_OR, TMP2, 0, TMP2, 0, TMP1, 0);  OP2(SLJIT_OR, TMP2, 0, TMP2, 0, TMP1, 0);
 compare = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);  
3006  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
3007  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3008    
3009  JUMPHERE(compare);  JUMPHERE(compare);
3010  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
3011  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
 JUMPHERE(jump);  
3012    
3013  /* We only have types for characters less than 256. */  /* We only have types for characters less than 256. */
 OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), (sljit_w)PRIV(utf8_table4) - 0xc0);  
 OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);  
 OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);  
 sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  
 }  
   
 #else /* COMPILE_PCRE8 */  
   
 #ifdef COMPILE_PCRE16  
 static void do_utfreadchar(compiler_common *common)  
 {  
 /* Fast decoding a UTF-16 character. TMP1 contains the first 16 bit char  
 of the character (>= 0xd800). Return char value in TMP1, length - 1 in TMP2. */  
 DEFINE_COMPILER;  
 struct sljit_jump *jump;  
   
 sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  
 jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xdc00);  
 /* Do nothing, only return. */  
 sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  
   
3014  JUMPHERE(jump);  JUMPHERE(jump);
3015  /* Combine two 16 bit characters. */  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
3016  OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
3017  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
 OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);  
 OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 10);  
 OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3ff);  
 OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);  
 OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(1));  
 OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x10000);  
3018  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3019  }  }
 #endif /* COMPILE_PCRE16 */  
3020    
3021  #endif /* COMPILE_PCRE8 */  #endif /* COMPILE_PCRE8 */
3022    
# Line 1956  DEFINE_COMPILER; Line 3036  DEFINE_COMPILER;
3036    
3037  SLJIT_ASSERT(UCD_BLOCK_SIZE == 128 && sizeof(ucd_record) == 8);  SLJIT_ASSERT(UCD_BLOCK_SIZE == 128 && sizeof(ucd_record) == 8);
3038    
3039  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0, 1, 5, 5, common->localsize);  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
3040  OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);  OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
3041  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_w)PRIV(ucd_stage1));  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(ucd_stage1));
3042  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);
3043  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
3044  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, TMP2, 0);  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, TMP2, 0);
3045  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, (sljit_w)PRIV(ucd_stage2));  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_stage2));
3046  OP1(SLJIT_MOV_UH, TMP2, 0, SLJIT_MEM2(TMP2, TMP1), 1);  OP1(SLJIT_MOV_UH, TMP2, 0, SLJIT_MEM2(TMP2, TMP1), 1);
3047  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, (sljit_w)PRIV(ucd_records) + SLJIT_OFFSETOF(ucd_record, chartype));  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_records) + SLJIT_OFFSETOF(ucd_record, chartype));
3048  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM2(TMP1, TMP2), 3);  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM2(TMP1, TMP2), 3);
3049  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);  sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3050  }  }
# Line 1978  struct sljit_label *newlinelabel = NULL; Line 3058  struct sljit_label *newlinelabel = NULL;
3058  struct sljit_jump *start;  struct sljit_jump *start;
3059  struct sljit_jump *end = NULL;  struct sljit_jump *end = NULL;
3060  struct sljit_jump *nl = NULL;  struct sljit_jump *nl = NULL;
3061  #ifdef SUPPORT_UTF  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3062  struct sljit_jump *singlechar;  struct sljit_jump *singlechar;
3063  #endif  #endif
3064  jump_list *newline = NULL;  jump_list *newline = NULL;
# Line 1989  if (!(hascrorlf || firstline) && (common Line 3069  if (!(hascrorlf || firstline) && (common
3069      common->nltype == NLTYPE_ANYCRLF || common->newline > 255))      common->nltype == NLTYPE_ANYCRLF || common->newline > 255))
3070    newlinecheck = TRUE;    newlinecheck = TRUE;
3071    
3072  if (firstline)  if (firstline)
3073    {    {
3074    /* Search for the end of the first line. */    /* Search for the end of the first line. */
3075    SLJIT_ASSERT(common->first_line_end != 0);    SLJIT_ASSERT(common->first_line_end != 0);
3076    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, STR_PTR, 0);    OP1(SLJIT_MOV, TMP3, 0, STR_PTR, 0);
3077    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_END, 0);  
3078      if (common->nltype == NLTYPE_FIXED && common->newline > 255)
3079        {
3080        mainloop = LABEL();
3081        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3082        end = CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3083        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(-1));
3084        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
3085        CMPTO(SLJIT_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, mainloop);
3086        CMPTO(SLJIT_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, mainloop);
3087        JUMPHERE(end);
3088        OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_SP), common->first_line_end, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3089        }
3090      else
3091        {
3092        end = CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3093        mainloop = LABEL();
3094        /* Continual stores does not cause data dependency. */
3095        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->first_line_end, STR_PTR, 0);
3096        read_char_range(common, common->nlmin, common->nlmax, TRUE);
3097        check_newlinechar(common, common->nltype, &newline, TRUE);
3098        CMPTO(SLJIT_LESS, STR_PTR, 0, STR_END, 0, mainloop);
3099        JUMPHERE(end);
3100        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SP), common->first_line_end, STR_PTR, 0);
3101        set_jumps(newline, LABEL());
3102        }
3103    
3104      OP1(SLJIT_MOV, STR_PTR, 0, TMP3, 0);
3105      }
3106    
3107    start = JUMP(SLJIT_JUMP);
3108    
3109    if (newlinecheck)
3110      {
3111      newlinelabel = LABEL();
3112      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3113      end = CMP(SLJIT_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3114      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3115      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, common->newline & 0xff);
3116      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
3117    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
3118      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, UCHAR_SHIFT);
3119    #endif
3120      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3121      nl = JUMP(SLJIT_JUMP);
3122      }
3123    
3124    mainloop = LABEL();
3125    
3126    /* Increasing the STR_PTR here requires one less jump in the most common case. */
3127    #ifdef SUPPORT_UTF
3128    if (common->utf) readuchar = TRUE;
3129    #endif
3130    if (newlinecheck) readuchar = TRUE;
3131    
3132    if (readuchar)
3133      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3134    
3135    if (newlinecheck)
3136      CMPTO(SLJIT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, newlinelabel);
3137    
3138    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3139    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3140    #if defined COMPILE_PCRE8
3141    if (common->utf)
3142      {
3143      singlechar = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
3144      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
3145      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3146      JUMPHERE(singlechar);
3147      }
3148    #elif defined COMPILE_PCRE16
3149    if (common->utf)
3150      {
3151      singlechar = CMP(SLJIT_LESS, TMP1, 0, SLJIT_IMM, 0xd800);
3152      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
3153      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);
3154      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_EQUAL);
3155      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
3156      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3157      JUMPHERE(singlechar);
3158      }
3159    #endif /* COMPILE_PCRE[8|16] */
3160    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
3161    JUMPHERE(start);
3162    
3163    if (newlinecheck)
3164      {
3165      JUMPHERE(end);
3166      JUMPHERE(nl);
3167      }
3168    
3169    return mainloop;
3170    }
3171    
3172    #define MAX_N_CHARS 16
3173    #define MAX_N_BYTES 8
3174    
3175    static SLJIT_INLINE void add_prefix_byte(pcre_uint8 byte, pcre_uint8 *bytes)
3176    {
3177    pcre_uint8 len = bytes[0];
3178    int i;
3179    
3180    if (len == 255)
3181      return;
3182    
3183    if (len == 0)
3184      {
3185      bytes[0] = 1;
3186      bytes[1] = byte;
3187      return;
3188      }
3189    
3190    for (i = len; i > 0; i--)
3191      if (bytes[i] == byte)
3192        return;
3193    
3194    if (len >= MAX_N_BYTES - 1)
3195      {
3196      bytes[0] = 255;
3197      return;
3198      }
3199    
3200    len++;
3201    bytes[len] = byte;
3202    bytes[0] = len;
3203    }
3204    
3205    static int scan_prefix(compiler_common *common, pcre_uchar *cc, pcre_uint32 *chars, pcre_uint8 *bytes, int max_chars)
3206    {
3207    /* Recursive function, which scans prefix literals. */
3208    BOOL last, any, caseless;
3209    int len, repeat, len_save, consumed = 0;
3210    pcre_uint32 chr, mask;
3211    pcre_uchar *alternative, *cc_save, *oc;
3212    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3213    pcre_uchar othercase[8];
3214    #elif defined SUPPORT_UTF && defined COMPILE_PCRE16
3215    pcre_uchar othercase[2];
3216    #else
3217    pcre_uchar othercase[1];
3218    #endif
3219    
3220    repeat = 1;
3221    while (TRUE)
3222      {
3223      last = TRUE;
3224      any = FALSE;
3225      caseless = FALSE;
3226      switch (*cc)
3227        {
3228        case OP_CHARI:
3229        caseless = TRUE;
3230        case OP_CHAR:
3231        last = FALSE;
3232        cc++;
3233        break;
3234    
3235        case OP_SOD:
3236        case OP_SOM:
3237        case OP_SET_SOM:
3238        case OP_NOT_WORD_BOUNDARY:
3239        case OP_WORD_BOUNDARY:
3240        case OP_EODN:
3241        case OP_EOD:
3242        case OP_CIRC:
3243        case OP_CIRCM:
3244        case OP_DOLL:
3245        case OP_DOLLM:
3246        /* Zero width assertions. */
3247        cc++;
3248        continue;
3249    
3250        case OP_ASSERT:
3251        case OP_ASSERT_NOT:
3252        case OP_ASSERTBACK:
3253        case OP_ASSERTBACK_NOT:
3254        cc = bracketend(cc);
3255        continue;
3256    
3257        case OP_PLUSI:
3258        case OP_MINPLUSI:
3259        case OP_POSPLUSI:
3260        caseless = TRUE;
3261        case OP_PLUS:
3262        case OP_MINPLUS:
3263        case OP_POSPLUS:
3264        cc++;
3265        break;
3266    
3267        case OP_EXACTI:
3268        caseless = TRUE;
3269        case OP_EXACT:
3270        repeat = GET2(cc, 1);
3271        last = FALSE;
3272        cc += 1 + IMM2_SIZE;
3273        break;
3274    
3275        case OP_QUERYI:
3276        case OP_MINQUERYI:
3277        case OP_POSQUERYI:
3278        caseless = TRUE;
3279        case OP_QUERY:
3280        case OP_MINQUERY:
3281        case OP_POSQUERY:
3282        len = 1;
3283        cc++;
3284    #ifdef SUPPORT_UTF
3285        if (common->utf && HAS_EXTRALEN(*cc)) len += GET_EXTRALEN(*cc);
3286    #endif
3287        max_chars = scan_prefix(common, cc + len, chars, bytes, max_chars);
3288        if (max_chars == 0)
3289          return consumed;
3290        last = FALSE;
3291        break;
3292    
3293        case OP_KET:
3294        cc += 1 + LINK_SIZE;
3295        continue;
3296    
3297        case OP_ALT:
3298        cc += GET(cc, 1);
3299        continue;
3300    
3301        case OP_ONCE:
3302        case OP_ONCE_NC:
3303        case OP_BRA:
3304        case OP_BRAPOS:
3305        case OP_CBRA:
3306        case OP_CBRAPOS:
3307        alternative = cc + GET(cc, 1);
3308        while (*alternative == OP_ALT)
3309          {
3310          max_chars = scan_prefix(common, alternative + 1 + LINK_SIZE, chars, bytes, max_chars);
3311          if (max_chars == 0)
3312            return consumed;
3313          alternative += GET(alternative, 1);
3314          }
3315    
3316        if (*cc == OP_CBRA || *cc == OP_CBRAPOS)
3317          cc += IMM2_SIZE;
3318        cc += 1 + LINK_SIZE;
3319        continue;
3320    
3321        case OP_CLASS:
3322    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3323        if (common->utf && !is_char7_bitset((const pcre_uint8 *)(cc + 1), FALSE)) return consumed;
3324    #endif
3325        any = TRUE;
3326        cc += 1 + 32 / sizeof(pcre_uchar);
3327        break;
3328    
3329        case OP_NCLASS:
3330    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3331        if (common->utf) return consumed;
3332    #endif
3333        any = TRUE;
3334        cc += 1 + 32 / sizeof(pcre_uchar);
3335        break;
3336    
3337    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3338        case OP_XCLASS:
3339    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3340        if (common->utf) return consumed;
3341    #endif
3342        any = TRUE;
3343        cc += GET(cc, 1);
3344        break;
3345    #endif
3346    
3347        case OP_DIGIT:
3348    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3349        if (common->utf && !is_char7_bitset((const pcre_uint8 *)common->ctypes - cbit_length + cbit_digit, FALSE))
3350          return consumed;
3351    #endif
3352        any = TRUE;
3353        cc++;
3354        break;
3355    
3356        case OP_WHITESPACE:
3357    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3358        if (common->utf && !is_char7_bitset((const pcre_uint8 *)common->ctypes - cbit_length + cbit_space, FALSE))
3359          return consumed;
3360    #endif
3361        any = TRUE;
3362        cc++;
3363        break;
3364    
3365        case OP_WORDCHAR:
3366    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3367        if (common->utf && !is_char7_bitset((const pcre_uint8 *)common->ctypes - cbit_length + cbit_word, FALSE))
3368          return consumed;
3369    #endif
3370        any = TRUE;
3371        cc++;
3372        break;
3373    
3374        case OP_NOT:
3375        case OP_NOTI:
3376        cc++;
3377        /* Fall through. */
3378        case OP_NOT_DIGIT:
3379        case OP_NOT_WHITESPACE:
3380        case OP_NOT_WORDCHAR:
3381        case OP_ANY:
3382        case OP_ALLANY:
3383    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3384        if (common->utf) return consumed;
3385    #endif
3386        any = TRUE;
3387        cc++;
3388        break;
3389    
3390    #ifdef SUPPORT_UCP
3391        case OP_NOTPROP:
3392        case OP_PROP:
3393    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3394        if (common->utf) return consumed;
3395    #endif
3396        any = TRUE;
3397        cc += 1 + 2;
3398        break;
3399    #endif
3400    
3401        case OP_TYPEEXACT:
3402        repeat = GET2(cc, 1);
3403        cc += 1 + IMM2_SIZE;
3404        continue;
3405    
3406        case OP_NOTEXACT:
3407        case OP_NOTEXACTI:
3408    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3409        if (common->utf) return consumed;
3410    #endif
3411        any = TRUE;
3412        repeat = GET2(cc, 1);
3413        cc += 1 + IMM2_SIZE + 1;
3414        break;
3415    
3416        default:
3417        return consumed;
3418        }
3419    
3420      if (any)
3421        {
3422    #if defined COMPILE_PCRE8
3423        mask = 0xff;
3424    #elif defined COMPILE_PCRE16
3425        mask = 0xffff;
3426    #elif defined COMPILE_PCRE32
3427        mask = 0xffffffff;
3428    #else
3429        SLJIT_ASSERT_STOP();
3430    #endif
3431    
3432        do
3433          {
3434          chars[0] = mask;
3435          chars[1] = mask;
3436          bytes[0] = 255;
3437    
3438          consumed++;
3439          if (--max_chars == 0)
3440            return consumed;
3441          chars += 2;
3442          bytes += MAX_N_BYTES;
3443          }
3444        while (--repeat > 0);
3445    
3446        repeat = 1;
3447        continue;
3448        }
3449    
3450      len = 1;
3451    #ifdef SUPPORT_UTF
3452      if (common->utf && HAS_EXTRALEN(*cc)) len += GET_EXTRALEN(*cc);
3453    #endif
3454    
3455      if (caseless && char_has_othercase(common, cc))
3456        {
3457    #ifdef SUPPORT_UTF
3458        if (common->utf)
3459          {
3460          GETCHAR(chr, cc);
3461          if ((int)PRIV(ord2utf)(char_othercase(common, chr), othercase) != len)
3462            return consumed;
3463          }
3464        else
3465    #endif
3466          {
3467          chr = *cc;
3468          othercase[0] = TABLE_GET(chr, common->fcc, chr);
3469          }
3470        }
3471      else
3472        caseless = FALSE;
3473    
3474      len_save = len;
3475      cc_save = cc;
3476      while (TRUE)
3477        {
3478        oc = othercase;
3479        do
3480          {
3481          chr = *cc;
3482    #ifdef COMPILE_PCRE32
3483          if (SLJIT_UNLIKELY(chr == NOTACHAR))
3484            return consumed;
3485    #endif
3486          add_prefix_byte((pcre_uint8)chr, bytes);
3487    
3488          mask = 0;
3489          if (caseless)
3490            {
3491            add_prefix_byte((pcre_uint8)*oc, bytes);
3492            mask = *cc ^ *oc;
3493            chr |= mask;
3494            }
3495    
3496    #ifdef COMPILE_PCRE32
3497          if (chars[0] == NOTACHAR && chars[1] == 0)
3498    #else
3499          if (chars[0] == NOTACHAR)
3500    #endif
3501            {
3502            chars[0] = chr;
3503            chars[1] = mask;
3504            }
3505          else
3506            {
3507            mask |= chars[0] ^ chr;
3508            chr |= mask;
3509            chars[0] = chr;
3510            chars[1] |= mask;
3511            }
3512    
3513          len--;
3514          consumed++;
3515          if (--max_chars == 0)
3516            return consumed;
3517          chars += 2;
3518          bytes += MAX_N_BYTES;
3519          cc++;
3520          oc++;
3521          }
3522        while (len > 0);
3523    
3524        if (--repeat == 0)
3525          break;
3526