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

Diff of /code/trunk/pcre_jit_compile.c

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

revision 695 by zherczeg, Sat Sep 17 06:05:38 2011 UTC revision 1452 by zherczeg, Tue Jan 28 16:07:52 2014 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2008 University of Cambridge             Copyright (c) 1997-2013 University of Cambridge
10    
11    The machine code generator part (this module) was written by Zoltan Herczeg    The machine code generator part (this module) was written by Zoltan Herczeg
12                        Copyright (c) 2010-2011                        Copyright (c) 2010-2013
13    
14  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
15  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 46  POSSIBILITY OF SUCH DAMAGE. Line 46  POSSIBILITY OF SUCH DAMAGE.
46    
47  #include "pcre_internal.h"  #include "pcre_internal.h"
48    
49  #ifdef SUPPORT_JIT  #if defined SUPPORT_JIT
50    
51  /* All-in-one: Since we use the JIT compiler only from here,  /* All-in-one: Since we use the JIT compiler only from here,
52  we just include it. This way we don't need to touch the build  we just include it. This way we don't need to touch the build
53  system files. */  system files. */
54    
55    #define SLJIT_MALLOC(size) (PUBL(malloc))(size)
56    #define SLJIT_FREE(ptr) (PUBL(free))(ptr)
57  #define SLJIT_CONFIG_AUTO 1  #define SLJIT_CONFIG_AUTO 1
58    #define SLJIT_CONFIG_STATIC 1
59  #define SLJIT_VERBOSE 0  #define SLJIT_VERBOSE 0
60  #define SLJIT_DEBUG 0  #define SLJIT_DEBUG 0
61    
62  #include "sljit/sljitLir.c"  #include "sljit/sljitLir.c"
63    
64  #if defined SLJIT_CONFIG_UNSUPPORTED && SLJIT_CONFIG_UNSUPPORTED  #if defined SLJIT_CONFIG_UNSUPPORTED && SLJIT_CONFIG_UNSUPPORTED
65  #error "Unsupported architecture"  #error Unsupported architecture
66  #endif  #endif
67    
68  /* Allocate memory on the stack. Fast, but limited size. */  /* Defines for debugging purposes. */
 #define LOCAL_SPACE_SIZE 32768  
69    
70    /* 1 - Use unoptimized capturing brackets.
71       2 - Enable capture_last_ptr (includes option 1). */
72    /* #define DEBUG_FORCE_UNOPTIMIZED_CBRAS 2 */
73    
74    /* 1 - Always have a control head. */
75    /* #define DEBUG_FORCE_CONTROL_HEAD 1 */
76    
77    /* Allocate memory for the regex stack on the real machine stack.
78    Fast, but limited size. */
79    #define MACHINE_STACK_SIZE 32768
80    
81    /* Growth rate for stack allocated by the OS. Should be the multiply
82    of page size. */
83  #define STACK_GROWTH_RATE 8192  #define STACK_GROWTH_RATE 8192
84    
85  /* Enable to check that the allocation could destroy temporaries. */  /* Enable to check that the allocation could destroy temporaries. */
# Line 79  The code generator follows the recursive Line 94  The code generator follows the recursive
94  expressions. The basic blocks of regular expressions are condition checkers  expressions. The basic blocks of regular expressions are condition checkers
95  whose execute different commands depending on the result of the condition check.  whose execute different commands depending on the result of the condition check.
96  The relationship between the operators can be horizontal (concatenation) and  The relationship between the operators can be horizontal (concatenation) and
97  vertical (sub-expression) (See struct fallback_common for more details).  vertical (sub-expression) (See struct backtrack_common for more details).
98    
99    'ab' - 'a' and 'b' regexps are concatenated    'ab' - 'a' and 'b' regexps are concatenated
100    'a+' - 'a' is the sub-expression of the '+' operator    'a+' - 'a' is the sub-expression of the '+' operator
101    
102  The condition checkers are boolean (true/false) checkers. Machine code is generated  The condition checkers are boolean (true/false) checkers. Machine code is generated
103  for the checker itself and for the actions depending on the result of the checker.  for the checker itself and for the actions depending on the result of the checker.
104  The 'true' case is called as the hot path (expected path), and the other is called as  The 'true' case is called as the matching path (expected path), and the other is called as
105  the 'fallback' path. Branch instructions are expesive for all CPUs, so we avoid taken  the 'backtrack' path. Branch instructions are expesive for all CPUs, so we avoid taken
106  branches on the hot path.  branches on the matching path.
107    
108   Greedy star operator (*) :   Greedy star operator (*) :
109     Hot path: match happens.     Matching path: match happens.
110     Fallback path: match failed.     Backtrack path: match failed.
111   Non-greedy star operator (*?) :   Non-greedy star operator (*?) :
112     Hot path: no need to perform a match.     Matching path: no need to perform a match.
113     Fallback path: match is required.     Backtrack path: match is required.
114    
115  The following example shows how the code generated for a capturing bracket  The following example shows how the code generated for a capturing bracket
116  with two alternatives. Let A, B, C, D are arbirary regular expressions, and  with two alternatives. Let A, B, C, D are arbirary regular expressions, and
# Line 105  we have the following regular expression Line 120  we have the following regular expression
120    
121  The generated code will be the following:  The generated code will be the following:
122    
123   A hot path   A matching path
124   '(' hot path (pushing arguments to the stack)   '(' matching path (pushing arguments to the stack)
125   B hot path   B matching path
126   ')' hot path (pushing arguments to the stack)   ')' matching path (pushing arguments to the stack)
127   D hot path   D matching path
128   return with successful match   return with successful match
129    
130   D fallback path   D backtrack path
131   ')' fallback path (If we arrived from "C" jump to the fallback of "C")   ')' backtrack path (If we arrived from "C" jump to the backtrack of "C")
132   B fallback path   B backtrack path
133   C expected path   C expected path
134   jump to D hot path   jump to D matching path
135   C fallback path   C backtrack path
136   A fallback path   A backtrack path
137    
138   Notice, that the order of fallback code paths are the opposite of the fast   Notice, that the order of backtrack code paths are the opposite of the fast
139   code paths. In this way the topmost value on the stack is always belong   code paths. In this way the topmost value on the stack is always belong
140   to the current fallback code path. The fallback code path must check   to the current backtrack code path. The backtrack path must check
141   whether there is a next alternative. If so, it needs to jump back to   whether there is a next alternative. If so, it needs to jump back to
142   the hot path eventually. Otherwise it needs to clear out its own stack   the matching path eventually. Otherwise it needs to clear out its own stack
143   frame and continue the execution on the fallback code paths.   frame and continue the execution on the backtrack code paths.
144  */  */
145    
146  /*  /*
147  Saved stack frames:  Saved stack frames:
148    
149  Atomic blocks and asserts require reloading the values of local variables  Atomic blocks and asserts require reloading the values of private data
150  when the fallback mechanism performed. Because of OP_RECURSE, the locals  when the backtrack mechanism performed. Because of OP_RECURSE, the data
151  are not necessarly known in compile time, thus we need a dynamic restore  are not necessarly known in compile time, thus we need a dynamic restore
152  mechanism.  mechanism.
153    
154  The stack frames are stored in a chain list, and have the following format:  The stack frames are stored in a chain list, and have the following format:
155  ([ capturing bracket offset ][ start value ][ end value ])+ ... [ 0 ] [ previous head ]  ([ capturing bracket offset ][ start value ][ end value ])+ ... [ 0 ] [ previous head ]
156    
157  Thus we can restore the locals to a particular point in the stack.  Thus we can restore the private data to a particular point in the stack.
158  */  */
159    
160  typedef struct jit_arguments {  typedef struct jit_arguments {
161    /* Pointers first. */    /* Pointers first. */
162    struct sljit_stack *stack;    struct sljit_stack *stack;
163    PCRE_SPTR str;    const pcre_uchar *str;
164    PCRE_SPTR begin;    const pcre_uchar *begin;
165    PCRE_SPTR end;    const pcre_uchar *end;
166    int *offsets;    int *offsets;
167    uschar *ptr;    pcre_uchar *uchar_ptr;
168      pcre_uchar *mark_ptr;
169      void *callout_data;
170    /* Everything else after. */    /* Everything else after. */
171    int offsetcount;    pcre_uint32 limit_match;
172    int calllimit;    int real_offset_count;
173    uschar notbol;    int offset_count;
174    uschar noteol;    pcre_uint8 notbol;
175    uschar notempty;    pcre_uint8 noteol;
176    uschar notempty_atstart;    pcre_uint8 notempty;
177      pcre_uint8 notempty_atstart;
178  } jit_arguments;  } jit_arguments;
179    
180  typedef struct executable_function {  typedef struct executable_functions {
181    void *executable_func;    void *executable_funcs[JIT_NUMBER_OF_COMPILE_MODES];
182    pcre_jit_callback callback;    sljit_uw *read_only_data[JIT_NUMBER_OF_COMPILE_MODES];
183      sljit_uw executable_sizes[JIT_NUMBER_OF_COMPILE_MODES];
184      PUBL(jit_callback) callback;
185    void *userdata;    void *userdata;
186  } executable_function;    pcre_uint32 top_bracket;
187      pcre_uint32 limit_match;
188    } executable_functions;
189    
190  typedef struct jump_list {  typedef struct jump_list {
191    struct sljit_jump *jump;    struct sljit_jump *jump;
192    struct jump_list *next;    struct jump_list *next;
193  } jump_list;  } jump_list;
194    
 enum stub_types { stack_alloc, max_index };  
   
195  typedef struct stub_list {  typedef struct stub_list {
   enum stub_types type;  
   int data;  
196    struct sljit_jump *start;    struct sljit_jump *start;
197    struct sljit_label *leave;    struct sljit_label *quit;
198    struct stub_list *next;    struct stub_list *next;
199  } stub_list;  } stub_list;
200    
201    typedef struct label_addr_list {
202      struct sljit_label *label;
203      sljit_uw *addr;
204      struct label_addr_list *next;
205    } label_addr_list;
206    
207    enum frame_types {
208      no_frame = -1,
209      no_stack = -2
210    };
211    
212    enum control_types {
213      type_mark = 0,
214      type_then_trap = 1
215    };
216    
217  typedef int (SLJIT_CALL *jit_function)(jit_arguments *args);  typedef int (SLJIT_CALL *jit_function)(jit_arguments *args);
218    
219  /* The following structure is the key data type for the recursive  /* The following structure is the key data type for the recursive
220  code generator. It is allocated by compile_hotpath, and contains  code generator. It is allocated by compile_matchingpath, and contains
221  the aguments for compile_fallbackpath. Must be the first member  the arguments for compile_backtrackingpath. Must be the first member
222  of its descendants. */  of its descendants. */
223  typedef struct fallback_common {  typedef struct backtrack_common {
224    /* Concatenation stack. */    /* Concatenation stack. */
225    struct fallback_common *prev;    struct backtrack_common *prev;
226    jump_list *nextfallbacks;    jump_list *nextbacktracks;
227    /* Internal stack (for component operators). */    /* Internal stack (for component operators). */
228    struct fallback_common *top;    struct backtrack_common *top;
229    jump_list *topfallbacks;    jump_list *topbacktracks;
230    /* Opcode pointer. */    /* Opcode pointer. */
231    uschar *cc;    pcre_uchar *cc;
232  } fallback_common;  } backtrack_common;
233    
234  typedef struct assert_fallback {  typedef struct assert_backtrack {
235    fallback_common common;    backtrack_common common;
236    jump_list *condfailed;    jump_list *condfailed;
237    /* Less than 0 (-1) if a frame is not needed. */    /* Less than 0 if a frame is not needed. */
238    int framesize;    int framesize;
239    /* Points to our private memory word on the stack. */    /* Points to our private memory word on the stack. */
240    int localptr;    int private_data_ptr;
241    /* For iterators. */    /* For iterators. */
242    struct sljit_label *hotpath;    struct sljit_label *matchingpath;
243  } assert_fallback;  } assert_backtrack;
244    
245  typedef struct bracket_fallback {  typedef struct bracket_backtrack {
246    fallback_common common;    backtrack_common common;
247    /* Where to coninue if an alternative is successfully matched. */    /* Where to coninue if an alternative is successfully matched. */
248    struct sljit_label *althotpath;    struct sljit_label *alternative_matchingpath;
249    /* For rmin and rmax iterators. */    /* For rmin and rmax iterators. */
250    struct sljit_label *recursivehotpath;    struct sljit_label *recursive_matchingpath;
251    /* For greedy ? operator. */    /* For greedy ? operator. */
252    struct sljit_label *zerohotpath;    struct sljit_label *zero_matchingpath;
253    /* Contains the branches of a failed condition. */    /* Contains the branches of a failed condition. */
254    union {    union {
255      /* Both for OP_COND, OP_SCOND. */      /* Both for OP_COND, OP_SCOND. */
256      jump_list *condfailed;      jump_list *condfailed;
257      assert_fallback *assert;      assert_backtrack *assert;
258      /* For OP_ONCE. -1 if not needed. */      /* For OP_ONCE. Less than 0 if not needed. */
259      int framesize;      int framesize;
260    } u;    } u;
261    /* Points to our private memory word on the stack. */    /* Points to our private memory word on the stack. */
262    int localptr;    int private_data_ptr;
263  } bracket_fallback;  } bracket_backtrack;
264    
265  typedef struct bracketpos_fallback {  typedef struct bracketpos_backtrack {
266    fallback_common common;    backtrack_common common;
267    /* Points to our private memory word on the stack. */    /* Points to our private memory word on the stack. */
268    int localptr;    int private_data_ptr;
269    /* Reverting stack is needed. */    /* Reverting stack is needed. */
270    int framesize;    int framesize;
271    /* Allocated stack size. */    /* Allocated stack size. */
272    int stacksize;    int stacksize;
273  } bracketpos_fallback;  } bracketpos_backtrack;
274    
275  typedef struct braminzero_fallback {  typedef struct braminzero_backtrack {
276    fallback_common common;    backtrack_common common;
277    struct sljit_label *hotpath;    struct sljit_label *matchingpath;
278  } braminzero_fallback;  } braminzero_backtrack;
279    
280  typedef struct iterator_fallback {  typedef struct iterator_backtrack {
281    fallback_common common;    backtrack_common common;
282    /* Next iteration. */    /* Next iteration. */
283    struct sljit_label *hotpath;    struct sljit_label *matchingpath;
284  } iterator_fallback;  } iterator_backtrack;
285    
286  typedef struct recurse_entry {  typedef struct recurse_entry {
287    struct recurse_entry *next;    struct recurse_entry *next;
# Line 256  typedef struct recurse_entry { Line 290  typedef struct recurse_entry {
290    /* Collects the calls until the function is not created. */    /* Collects the calls until the function is not created. */
291    jump_list *calls;    jump_list *calls;
292    /* Points to the starting opcode. */    /* Points to the starting opcode. */
293    int start;    sljit_sw start;
294  } recurse_entry;  } recurse_entry;
295    
296  typedef struct recurse_fallback {  typedef struct recurse_backtrack {
297    fallback_common common;    backtrack_common common;
298  } recurse_fallback;    BOOL inlined_pattern;
299    } recurse_backtrack;
300    
301    #define OP_THEN_TRAP OP_TABLE_LENGTH
302    
303    typedef struct then_trap_backtrack {
304      backtrack_common common;
305      /* If then_trap is not NULL, this structure contains the real
306      then_trap for the backtracking path. */
307      struct then_trap_backtrack *then_trap;
308      /* Points to the starting opcode. */
309      sljit_sw start;
310      /* Exit point for the then opcodes of this alternative. */
311      jump_list *quit;
312      /* Frame size of the current alternative. */
313      int framesize;
314    } then_trap_backtrack;
315    
316    #define MAX_RANGE_SIZE 4
317    
318  typedef struct compiler_common {  typedef struct compiler_common {
319      /* The sljit ceneric compiler. */
320    struct sljit_compiler *compiler;    struct sljit_compiler *compiler;
321    uschar *start;    /* First byte code. */
322    int localsize;    pcre_uchar *start;
323    int *localptrs;    /* Maps private data offset to each opcode. */
324    const uschar *fcc;    sljit_si *private_data_ptrs;
325    sljit_w lcc;    /* This read-only data is available during runtime. */
326    int cbraptr;    sljit_uw *read_only_data;
327      /* The total size of the read-only data. */
328      sljit_uw read_only_data_size;
329      /* The next free entry of the read_only_data. */
330      sljit_uw *read_only_data_ptr;
331      /* Tells whether the capturing bracket is optimized. */
332      pcre_uint8 *optimized_cbracket;
333      /* Tells whether the starting offset is a target of then. */
334      pcre_uint8 *then_offsets;
335      /* Current position where a THEN must jump. */
336      then_trap_backtrack *then_trap;
337      /* Starting offset of private data for capturing brackets. */
338      int cbra_ptr;
339      /* Output vector starting point. Must be divisible by 2. */
340      int ovector_start;
341      /* Last known position of the requested byte. */
342      int req_char_ptr;
343      /* Head of the last recursion. */
344      int recursive_head_ptr;
345      /* First inspected character for partial matching. */
346      int start_used_ptr;
347      /* Starting pointer for partial soft matches. */
348      int hit_start;
349      /* End pointer of the first line. */
350      int first_line_end;
351      /* Points to the marked string. */
352      int mark_ptr;
353      /* Recursive control verb management chain. */
354      int control_head_ptr;
355      /* Points to the last matched capture block index. */
356      int capture_last_ptr;
357      /* Points to the starting position of the current match. */
358      int start_ptr;
359    
360      /* Flipped and lower case tables. */
361      const pcre_uint8 *fcc;
362      sljit_sw lcc;
363      /* Mode can be PCRE_STUDY_JIT_COMPILE and others. */
364      int mode;
365      /* TRUE, when minlength is greater than 0. */
366      BOOL might_be_empty;
367      /* \K is found in the pattern. */
368      BOOL has_set_som;
369      /* (*SKIP:arg) is found in the pattern. */
370      BOOL has_skip_arg;
371      /* (*THEN) is found in the pattern. */
372      BOOL has_then;
373      /* Needs to know the start position anytime. */
374      BOOL needs_start_ptr;
375      /* Currently in recurse or negative assert. */
376      BOOL local_exit;
377      /* Currently in a positive assert. */
378      BOOL positive_assert;
379      /* Newline control. */
380    int nltype;    int nltype;
381      pcre_uint32 nlmax;
382      pcre_uint32 nlmin;
383    int newline;    int newline;
384    int bsr_nltype;    int bsr_nltype;
385      pcre_uint32 bsr_nlmax;
386      pcre_uint32 bsr_nlmin;
387      /* Dollar endonly. */
388    int endonly;    int endonly;
389    sljit_w ctypes;    /* Tables. */
390    struct sljit_label *acceptlabel;    sljit_sw ctypes;
391      /* Named capturing brackets. */
392      pcre_uchar *name_table;
393      sljit_sw name_count;
394      sljit_sw name_entry_size;
395    
396      /* Labels and jump lists. */
397      struct sljit_label *partialmatchlabel;
398      struct sljit_label *quit_label;
399      struct sljit_label *forced_quit_label;
400      struct sljit_label *accept_label;
401      struct sljit_label *ff_newline_shortcut;
402    stub_list *stubs;    stub_list *stubs;
403      label_addr_list *label_addrs;
404    recurse_entry *entries;    recurse_entry *entries;
405    recurse_entry *currententry;    recurse_entry *currententry;
406      jump_list *partialmatch;
407      jump_list *quit;
408      jump_list *positive_assert_quit;
409      jump_list *forced_quit;
410    jump_list *accept;    jump_list *accept;
411    jump_list *calllimit;    jump_list *calllimit;
412    jump_list *stackalloc;    jump_list *stackalloc;
# Line 290  typedef struct compiler_common { Line 417  typedef struct compiler_common {
417    jump_list *vspace;    jump_list *vspace;
418    jump_list *casefulcmp;    jump_list *casefulcmp;
419    jump_list *caselesscmp;    jump_list *caselesscmp;
420      jump_list *reset_match;
421    BOOL jscript_compat;    BOOL jscript_compat;
422  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
423    BOOL utf8;    BOOL utf;
424  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
425    BOOL useucp;    BOOL use_ucp;
426  #endif  #endif
427    jump_list *utf8readchar;  #ifdef COMPILE_PCRE8
428    jump_list *utf8readtype8;    jump_list *utfreadchar;
429      jump_list *utfreadchar16;
430      jump_list *utfreadtype8;
431  #endif  #endif
432    #endif /* SUPPORT_UTF */
433  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
434    jump_list *getucd;    jump_list *getucd;
435  #endif  #endif
# Line 310  typedef struct compare_context { Line 441  typedef struct compare_context {
441    int length;    int length;
442    int sourcereg;    int sourcereg;
443  #if defined SLJIT_UNALIGNED && SLJIT_UNALIGNED  #if defined SLJIT_UNALIGNED && SLJIT_UNALIGNED
444    int byteptr;    int ucharptr;
445    union {    union {
446      int asint;      sljit_si asint;
447      short asshort;      sljit_uh asushort;
448    #if defined COMPILE_PCRE8
449      sljit_ub asbyte;      sljit_ub asbyte;
450      sljit_ub asbytes[4];      sljit_ub asuchars[4];
451    #elif defined COMPILE_PCRE16
452        sljit_uh asuchars[2];
453    #elif defined COMPILE_PCRE32
454        sljit_ui asuchars[1];
455    #endif
456    } c;    } c;
457    union {    union {
458      int asint;      sljit_si asint;
459      short asshort;      sljit_uh asushort;
460    #if defined COMPILE_PCRE8
461      sljit_ub asbyte;      sljit_ub asbyte;
462      sljit_ub asbytes[4];      sljit_ub asuchars[4];
463    #elif defined COMPILE_PCRE16
464        sljit_uh asuchars[2];
465    #elif defined COMPILE_PCRE32
466        sljit_ui asuchars[1];
467    #endif
468    } oc;    } oc;
469  #endif  #endif
470  } compare_context;  } compare_context;
471    
472  enum {  /* Undefine sljit macros. */
473    frame_end = 0,  #undef CMP
   frame_setmaxindex = -1,  
   frame_setstrbegin = -2  
 };  
474    
475  /* Used for accessing the elements of the stack. */  /* Used for accessing the elements of the stack. */
476  #define STACK(i)      ((-(i) - 1) * (int)sizeof(sljit_w))  #define STACK(i)      ((-(i) - 1) * (int)sizeof(sljit_sw))
477    
478  #define TMP1          SLJIT_TEMPORARY_REG1  #define TMP1          SLJIT_SCRATCH_REG1
479  #define TMP2          SLJIT_TEMPORARY_REG3  #define TMP2          SLJIT_SCRATCH_REG3
480  #define TMP3          SLJIT_TEMPORARY_EREG2  #define TMP3          SLJIT_TEMPORARY_EREG2
481  #define STR_PTR       SLJIT_GENERAL_REG1  #define STR_PTR       SLJIT_SAVED_REG1
482  #define STR_END       SLJIT_GENERAL_REG2  #define STR_END       SLJIT_SAVED_REG2
483  #define STACK_TOP     SLJIT_TEMPORARY_REG2  #define STACK_TOP     SLJIT_SCRATCH_REG2
484  #define STACK_LIMIT   SLJIT_GENERAL_REG3  #define STACK_LIMIT   SLJIT_SAVED_REG3
485  #define ARGUMENTS     SLJIT_GENERAL_EREG1  #define ARGUMENTS     SLJIT_SAVED_EREG1
486  #define CALL_COUNT    SLJIT_GENERAL_EREG2  #define COUNT_MATCH   SLJIT_SAVED_EREG2
487  #define RETURN_ADDR   SLJIT_TEMPORARY_EREG1  #define RETURN_ADDR   SLJIT_TEMPORARY_EREG1
488    
489  /* Locals layout. */  /* Local space layout. */
490  /* These two locals can be used by the current opcode. */  /* These two locals can be used by the current opcode. */
491  #define LOCALS0          (0 * sizeof(sljit_w))  #define LOCALS0          (0 * sizeof(sljit_sw))
492  #define LOCALS1          (1 * sizeof(sljit_w))  #define LOCALS1          (1 * sizeof(sljit_sw))
493  /* Two local variables for possessive quantifiers (char1 cannot use them). */  /* Two local variables for possessive quantifiers (char1 cannot use them). */
494  #define POSSESSIVE0      (2 * sizeof(sljit_w))  #define POSSESSIVE0      (2 * sizeof(sljit_sw))
495  #define POSSESSIVE1      (3 * sizeof(sljit_w))  #define POSSESSIVE1      (3 * sizeof(sljit_sw))
 /* Head of the saved local variables */  
 #define LOCALS_HEAD      (4 * sizeof(sljit_w))  
 /* Head of the last recursion. */  
 #define RECURSIVE_HEAD   (5 * sizeof(sljit_w))  
 /* Number of recursions. */  
 #define MAX_INDEX        (6 * sizeof(sljit_w))  
496  /* Max limit of recursions. */  /* Max limit of recursions. */
497  #define CALL_LIMIT       (7 * sizeof(sljit_w))  #define LIMIT_MATCH      (4 * sizeof(sljit_sw))
 /* Last known position of the requested byte. */  
 #define REQ_BYTE_PTR     (8 * sizeof(sljit_w))  
 /* End pointer of the first line. */  
 #define FIRSTLINE_END    (9 * sizeof(sljit_w))  
498  /* The output vector is stored on the stack, and contains pointers  /* The output vector is stored on the stack, and contains pointers
499  to characters. The vector data is divided into two groups: the first  to characters. The vector data is divided into two groups: the first
500  group contains the start / end character pointers, and the second is  group contains the start / end character pointers, and the second is
501  the start pointers when the end of the capturing group has not yet reached. */  the start pointers when the end of the capturing group has not yet reached. */
502  #define OVECTOR_START    (10 * sizeof(sljit_w))  #define OVECTOR_START    (common->ovector_start)
503  #define OVECTOR(i)       (OVECTOR_START + (i) * sizeof(sljit_w))  #define OVECTOR(i)       (OVECTOR_START + (i) * (sljit_sw)sizeof(sljit_sw))
504  #define OVECTOR_PRIV(i)  (common->cbraptr + (i) * sizeof(sljit_w))  #define OVECTOR_PRIV(i)  (common->cbra_ptr + (i) * (sljit_sw)sizeof(sljit_sw))
505  #define PRIV(cc)         (common->localptrs[(cc) - common->start])  #define PRIVATE_DATA(cc) (common->private_data_ptrs[(cc) - common->start])
506    
507    #if defined COMPILE_PCRE8
508    #define MOV_UCHAR  SLJIT_MOV_UB
509    #define MOVU_UCHAR SLJIT_MOVU_UB
510    #elif defined COMPILE_PCRE16
511    #define MOV_UCHAR  SLJIT_MOV_UH
512    #define MOVU_UCHAR SLJIT_MOVU_UH
513    #elif defined COMPILE_PCRE32
514    #define MOV_UCHAR  SLJIT_MOV_UI
515    #define MOVU_UCHAR SLJIT_MOVU_UI
516    #else
517    #error Unsupported compiling mode
518    #endif
519    
520  /* Shortcuts. */  /* Shortcuts. */
521  #define DEFINE_COMPILER \  #define DEFINE_COMPILER \
# Line 389  the start pointers when the end of the c Line 532  the start pointers when the end of the c
532    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))    sljit_set_label(sljit_emit_jump(compiler, (type)), (label))
533  #define JUMPHERE(jump) \  #define JUMPHERE(jump) \
534    sljit_set_label((jump), sljit_emit_label(compiler))    sljit_set_label((jump), sljit_emit_label(compiler))
535    #define SET_LABEL(jump, label) \
536      sljit_set_label((jump), (label))
537  #define CMP(type, src1, src1w, src2, src2w) \  #define CMP(type, src1, src1w, src2, src2w) \
538    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))    sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w))
539  #define CMPTO(type, src1, src1w, src2, src2w, label) \  #define CMPTO(type, src1, src1w, src2, src2w, label) \
540    sljit_set_label(sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w)), (label))    sljit_set_label(sljit_emit_cmp(compiler, (type), (src1), (src1w), (src2), (src2w)), (label))
541  #define COND_VALUE(op, dst, dstw, type) \  #define OP_FLAGS(op, dst, dstw, src, srcw, type) \
542    sljit_emit_cond_value(compiler, (op), (dst), (dstw), (type))    sljit_emit_op_flags(compiler, (op), (dst), (dstw), (src), (srcw), (type))
543    #define GET_LOCAL_BASE(dst, dstw, offset) \
544      sljit_get_local_base(compiler, (dst), (dstw), (offset))
545    
546    #define READ_CHAR_MAX 0x7fffffff
547    
548  static uschar* bracketend(uschar* cc)  static pcre_uchar* bracketend(pcre_uchar* cc)
549  {  {
550  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));
551  do cc += GET(cc, 1); while (*cc == OP_ALT);  do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 405  cc += 1 + LINK_SIZE; Line 554  cc += 1 + LINK_SIZE;
554  return cc;  return cc;
555  }  }
556    
557    static int no_alternatives(pcre_uchar* cc)
558    {
559    int count = 0;
560    SLJIT_ASSERT((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND));
561    do
562      {
563      cc += GET(cc, 1);
564      count++;
565      }
566    while (*cc == OP_ALT);
567    SLJIT_ASSERT(*cc >= OP_KET && *cc <= OP_KETRPOS);
568    return count;
569    }
570    
571    static int ones_in_half_byte[16] = {
572      /* 0 */ 0, 1, 1, 2, /* 4 */ 1, 2, 2, 3,
573      /* 8 */ 1, 2, 2, 3, /* 12 */ 2, 3, 3, 4
574    };
575    
576  /* Functions whose might need modification for all new supported opcodes:  /* Functions whose might need modification for all new supported opcodes:
577   next_opcode   next_opcode
578   get_localspace   check_opcode_types
579   set_localptrs   set_private_data_ptrs
580   get_framesize   get_framesize
581   init_frame   init_frame
582   get_localsize   get_private_data_copy_length
583   copy_locals   copy_private_data
584   compile_hotpath   compile_matchingpath
585   compile_fallbackpath   compile_backtrackingpath
586  */  */
587    
588  static uschar *next_opcode(compiler_common *common, uschar *cc)  static pcre_uchar *next_opcode(compiler_common *common, pcre_uchar *cc)
589  {  {
590  SLJIT_UNUSED_ARG(common);  SLJIT_UNUSED_ARG(common);
591  switch(*cc)  switch(*cc)
# Line 435  switch(*cc) Line 603  switch(*cc)
603    case OP_WORDCHAR:    case OP_WORDCHAR:
604    case OP_ANY:    case OP_ANY:
605    case OP_ALLANY:    case OP_ALLANY:
606      case OP_NOTPROP:
607      case OP_PROP:
608    case OP_ANYNL:    case OP_ANYNL:
609    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
610    case OP_HSPACE:    case OP_HSPACE:
# Line 447  switch(*cc) Line 617  switch(*cc)
617    case OP_CIRCM:    case OP_CIRCM:
618    case OP_DOLL:    case OP_DOLL:
619    case OP_DOLLM:    case OP_DOLLM:
   case OP_TYPESTAR:  
   case OP_TYPEMINSTAR:  
   case OP_TYPEPLUS:  
   case OP_TYPEMINPLUS:  
   case OP_TYPEQUERY:  
   case OP_TYPEMINQUERY:  
   case OP_TYPEPOSSTAR:  
   case OP_TYPEPOSPLUS:  
   case OP_TYPEPOSQUERY:  
620    case OP_CRSTAR:    case OP_CRSTAR:
621    case OP_CRMINSTAR:    case OP_CRMINSTAR:
622    case OP_CRPLUS:    case OP_CRPLUS:
623    case OP_CRMINPLUS:    case OP_CRMINPLUS:
624    case OP_CRQUERY:    case OP_CRQUERY:
625    case OP_CRMINQUERY:    case OP_CRMINQUERY:
626      case OP_CRRANGE:
627      case OP_CRMINRANGE:
628      case OP_CRPOSSTAR:
629      case OP_CRPOSPLUS:
630      case OP_CRPOSQUERY:
631      case OP_CRPOSRANGE:
632      case OP_CLASS:
633      case OP_NCLASS:
634      case OP_REF:
635      case OP_REFI:
636      case OP_DNREF:
637      case OP_DNREFI:
638      case OP_RECURSE:
639      case OP_CALLOUT:
640      case OP_ALT:
641      case OP_KET:
642      case OP_KETRMAX:
643      case OP_KETRMIN:
644      case OP_KETRPOS:
645      case OP_REVERSE:
646      case OP_ASSERT:
647      case OP_ASSERT_NOT:
648      case OP_ASSERTBACK:
649      case OP_ASSERTBACK_NOT:
650      case OP_ONCE:
651      case OP_ONCE_NC:
652      case OP_BRA:
653      case OP_BRAPOS:
654      case OP_CBRA:
655      case OP_CBRAPOS:
656      case OP_COND:
657      case OP_SBRA:
658      case OP_SBRAPOS:
659      case OP_SCBRA:
660      case OP_SCBRAPOS:
661      case OP_SCOND:
662      case OP_CREF:
663      case OP_DNCREF:
664      case OP_RREF:
665      case OP_DNRREF:
666    case OP_DEF:    case OP_DEF:
667    case OP_BRAZERO:    case OP_BRAZERO:
668    case OP_BRAMINZERO:    case OP_BRAMINZERO:
669    case OP_BRAPOSZERO:    case OP_BRAPOSZERO:
670      case OP_PRUNE:
671      case OP_SKIP:
672      case OP_THEN:
673      case OP_COMMIT:
674    case OP_FAIL:    case OP_FAIL:
675    case OP_ACCEPT:    case OP_ACCEPT:
676    case OP_ASSERT_ACCEPT:    case OP_ASSERT_ACCEPT:
677      case OP_CLOSE:
678    case OP_SKIPZERO:    case OP_SKIPZERO:
679    return cc + 1;    return cc + PRIV(OP_lengths)[*cc];
680    
681    case OP_CHAR:    case OP_CHAR:
682    case OP_CHARI:    case OP_CHARI:
683    case OP_NOT:    case OP_NOT:
684    case OP_NOTI:    case OP_NOTI:
   
685    case OP_STAR:    case OP_STAR:
686    case OP_MINSTAR:    case OP_MINSTAR:
687    case OP_PLUS:    case OP_PLUS:
688    case OP_MINPLUS:    case OP_MINPLUS:
689    case OP_QUERY:    case OP_QUERY:
690    case OP_MINQUERY:    case OP_MINQUERY:
691      case OP_UPTO:
692      case OP_MINUPTO:
693      case OP_EXACT:
694    case OP_POSSTAR:    case OP_POSSTAR:
695    case OP_POSPLUS:    case OP_POSPLUS:
696    case OP_POSQUERY:    case OP_POSQUERY:
697      case OP_POSUPTO:
698    case OP_STARI:    case OP_STARI:
699    case OP_MINSTARI:    case OP_MINSTARI:
700    case OP_PLUSI:    case OP_PLUSI:
701    case OP_MINPLUSI:    case OP_MINPLUSI:
702    case OP_QUERYI:    case OP_QUERYI:
703    case OP_MINQUERYI:    case OP_MINQUERYI:
704      case OP_UPTOI:
705      case OP_MINUPTOI:
706      case OP_EXACTI:
707    case OP_POSSTARI:    case OP_POSSTARI:
708    case OP_POSPLUSI:    case OP_POSPLUSI:
709    case OP_POSQUERYI:    case OP_POSQUERYI:
710      case OP_POSUPTOI:
711    case OP_NOTSTAR:    case OP_NOTSTAR:
712    case OP_NOTMINSTAR:    case OP_NOTMINSTAR:
713    case OP_NOTPLUS:    case OP_NOTPLUS:
714    case OP_NOTMINPLUS:    case OP_NOTMINPLUS:
715    case OP_NOTQUERY:    case OP_NOTQUERY:
716    case OP_NOTMINQUERY:    case OP_NOTMINQUERY:
717      case OP_NOTUPTO:
718      case OP_NOTMINUPTO:
719      case OP_NOTEXACT:
720    case OP_NOTPOSSTAR:    case OP_NOTPOSSTAR:
721    case OP_NOTPOSPLUS:    case OP_NOTPOSPLUS:
722    case OP_NOTPOSQUERY:    case OP_NOTPOSQUERY:
723      case OP_NOTPOSUPTO:
724    case OP_NOTSTARI:    case OP_NOTSTARI:
725    case OP_NOTMINSTARI:    case OP_NOTMINSTARI:
726    case OP_NOTPLUSI:    case OP_NOTPLUSI:
727    case OP_NOTMINPLUSI:    case OP_NOTMINPLUSI:
728    case OP_NOTQUERYI:    case OP_NOTQUERYI:
729    case OP_NOTMINQUERYI:    case OP_NOTMINQUERYI:
   case OP_NOTPOSSTARI:  
   case OP_NOTPOSPLUSI:  
   case OP_NOTPOSQUERYI:  
   cc += 2;  
 #ifdef SUPPORT_UTF8  
   if (common->utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];  
 #endif  
   return cc;  
   
   case OP_UPTO:  
   case OP_MINUPTO:  
   case OP_EXACT:  
   case OP_POSUPTO:  
   case OP_UPTOI:  
   case OP_MINUPTOI:  
   case OP_EXACTI:  
   case OP_POSUPTOI:  
   case OP_NOTUPTO:  
   case OP_NOTMINUPTO:  
   case OP_NOTEXACT:  
   case OP_NOTPOSUPTO:  
730    case OP_NOTUPTOI:    case OP_NOTUPTOI:
731    case OP_NOTMINUPTOI:    case OP_NOTMINUPTOI:
732    case OP_NOTEXACTI:    case OP_NOTEXACTI:
733      case OP_NOTPOSSTARI:
734      case OP_NOTPOSPLUSI:
735      case OP_NOTPOSQUERYI:
736    case OP_NOTPOSUPTOI:    case OP_NOTPOSUPTOI:
737    cc += 4;    cc += PRIV(OP_lengths)[*cc];
738  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF
739    if (common->utf8 && cc[-1] >= 0xc0) cc += _pcre_utf8_table4[cc[-1] & 0x3f];    if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
740  #endif  #endif
741    return cc;    return cc;
742    
743    case OP_NOTPROP:    /* Special cases. */
744    case OP_PROP:    case OP_TYPESTAR:
745      case OP_TYPEMINSTAR:
746      case OP_TYPEPLUS:
747      case OP_TYPEMINPLUS:
748      case OP_TYPEQUERY:
749      case OP_TYPEMINQUERY:
750    case OP_TYPEUPTO:    case OP_TYPEUPTO:
751    case OP_TYPEMINUPTO:    case OP_TYPEMINUPTO:
752    case OP_TYPEEXACT:    case OP_TYPEEXACT:
753      case OP_TYPEPOSSTAR:
754      case OP_TYPEPOSPLUS:
755      case OP_TYPEPOSQUERY:
756    case OP_TYPEPOSUPTO:    case OP_TYPEPOSUPTO:
757    case OP_REF:    return cc + PRIV(OP_lengths)[*cc] - 1;
   case OP_REFI:  
   case OP_CREF:  
   case OP_CLOSE:  
   cc += 3;  
   return cc;  
   
   case OP_CRRANGE:  
   case OP_CRMINRANGE:  
   return cc + 5;  
758    
759    case OP_CLASS:    case OP_ANYBYTE:
760    case OP_NCLASS:  #ifdef SUPPORT_UTF
761    return cc + 33;    if (common->utf) return NULL;
762    #endif
763      return cc + 1;
764    
765  #ifdef SUPPORT_UTF8  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
766    case OP_XCLASS:    case OP_XCLASS:
767    return cc + GET(cc, 1);    return cc + GET(cc, 1);
768  #endif  #endif
769    
770    case OP_RECURSE:    case OP_MARK:
771    case OP_ASSERT:    case OP_PRUNE_ARG:
772    case OP_ASSERT_NOT:    case OP_SKIP_ARG:
773    case OP_ASSERTBACK:    case OP_THEN_ARG:
774    case OP_ASSERTBACK_NOT:    return cc + 1 + 2 + cc[1];
   case OP_REVERSE:  
   case OP_ONCE:  
   case OP_BRA:  
   case OP_BRAPOS:  
   case OP_COND:  
   case OP_SBRA:  
   case OP_SBRAPOS:  
   case OP_SCOND:  
   case OP_ALT:  
   case OP_KET:  
   case OP_KETRMAX:  
   case OP_KETRMIN:  
   case OP_KETRPOS:  
   return cc + 1 + LINK_SIZE;  
   
   case OP_CBRA:  
   case OP_CBRAPOS:  
   case OP_SCBRA:  
   case OP_SCBRAPOS:  
   return cc + 1 + LINK_SIZE + 2;  
775    
776    default:    default:
777      /* All opcodes are supported now! */
778      SLJIT_ASSERT_STOP();
779    return NULL;    return NULL;
780    }    }
781  }  }
782    
783  static int get_localspace(compiler_common *common, uschar *cc, uschar *ccend)  static BOOL check_opcode_types(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend)
784  {  {
785  int localspace = 0;  int count;
786  uschar *alternative;  pcre_uchar *slot;
787    
788  /* Calculate important variables (like stack size) and checks whether all opcodes are supported. */  /* Calculate important variables (like stack size) and checks whether all opcodes are supported. */
789  while (cc < ccend)  while (cc < ccend)
790    {    {
791    switch(*cc)    switch(*cc)
792      {      {
793      case OP_ASSERT:      case OP_SET_SOM:
794      case OP_ASSERT_NOT:      common->has_set_som = TRUE;
795      case OP_ASSERTBACK:      common->might_be_empty = TRUE;
796      case OP_ASSERTBACK_NOT:      cc += 1;
     case OP_ONCE:  
     case OP_BRAPOS:  
     case OP_SBRA:  
     case OP_SBRAPOS:  
     case OP_SCOND:  
     localspace += sizeof(sljit_w);  
     cc += 1 + LINK_SIZE;  
     break;  
   
     case OP_CBRAPOS:  
     case OP_SCBRAPOS:  
     localspace += sizeof(sljit_w);  
     cc += 1 + LINK_SIZE + 2;  
     break;  
   
     case OP_COND:  
     /* Might be a hidden SCOND. */  
     alternative = cc + GET(cc, 1);  
     if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)  
       localspace += sizeof(sljit_w);  
     cc += 1 + LINK_SIZE;  
797      break;      break;
798    
799      default:      case OP_REF:
800      cc = next_opcode(common, cc);      case OP_REFI:
801      if (cc == NULL)      common->optimized_cbracket[GET2(cc, 1)] = 0;
802        return -1;      cc += 1 + IMM2_SIZE;
803      break;      break;
     }  
   }  
 return localspace;  
 }  
804    
805  static void set_localptrs(compiler_common *common, int localptr, uschar *ccend)      case OP_BRA:
806  {      case OP_CBRA:
 uschar *cc = common->start;  
 uschar *alternative;  
 while (cc < ccend)  
   {  
   switch(*cc)  
     {  
     case OP_ASSERT:  
     case OP_ASSERT_NOT:  
     case OP_ASSERTBACK:  
     case OP_ASSERTBACK_NOT:  
     case OP_ONCE:  
     case OP_BRAPOS:  
807      case OP_SBRA:      case OP_SBRA:
808      case OP_SBRAPOS:      case OP_SCBRA:
809      case OP_SCOND:      count = no_alternatives(cc);
810      common->localptrs[cc - common->start] = localptr;      if (count > 4)
811      localptr += sizeof(sljit_w);        common->read_only_data_size += count * sizeof(sljit_uw);
812      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE + (*cc == OP_CBRA || *cc == OP_SCBRA ? IMM2_SIZE : 0);
813      break;      break;
814    
815      case OP_CBRAPOS:      case OP_CBRAPOS:
816      case OP_SCBRAPOS:      case OP_SCBRAPOS:
817      common->localptrs[cc - common->start] = localptr;      common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] = 0;
818      localptr += sizeof(sljit_w);      cc += 1 + LINK_SIZE + IMM2_SIZE;
     cc += 1 + LINK_SIZE + 2;  
819      break;      break;
820    
821      case OP_COND:      case OP_COND:
822      /* Might be a hidden SCOND. */      case OP_SCOND:
823      alternative = cc + GET(cc, 1);      /* Only AUTO_CALLOUT can insert this opcode. We do
824      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)         not intend to support this case. */
825        {      if (cc[1 + LINK_SIZE] == OP_CALLOUT)
826        common->localptrs[cc - common->start] = localptr;        return FALSE;
       localptr += sizeof(sljit_w);  
       }  
827      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
828      break;      break;
829    
830      default:      case OP_CREF:
831      cc = next_opcode(common, cc);      common->optimized_cbracket[GET2(cc, 1)] = 0;
832      SLJIT_ASSERT(cc != NULL);      cc += 1 + IMM2_SIZE;
833        break;
834    
835        case OP_DNREF:
836        case OP_DNREFI:
837        case OP_DNCREF:
838        count = GET2(cc, 1 + IMM2_SIZE);
839        slot = common->name_table + GET2(cc, 1) * common->name_entry_size;
840        while (count-- > 0)
841          {
842          common->optimized_cbracket[GET2(slot, 0)] = 0;
843          slot += common->name_entry_size;
844          }
845        cc += 1 + 2 * IMM2_SIZE;
846      break;      break;
     }  
   }  
 }  
847    
 /* Returns with -1 if no need for frame. */  
 static int get_framesize(compiler_common *common, uschar *cc, BOOL recursive)  
 {  
 uschar *ccend = bracketend(cc);  
 uschar *end;  
 int length = 0;  
 BOOL possessive = FALSE;  
 BOOL needs_frame = FALSE;  
 BOOL needs_maxindex = FALSE;  
 BOOL setsom_found = FALSE;  
   
 if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))  
   {  
   length = 3 + 2;  
   needs_maxindex = TRUE;  
   possessive = TRUE;  
   }  
   
 cc = next_opcode(common, cc);  
 SLJIT_ASSERT(cc != NULL);  
 while (cc < ccend)  
   switch(*cc)  
     {  
     case OP_SET_SOM:  
848      case OP_RECURSE:      case OP_RECURSE:
849      if (!setsom_found)      /* Set its value only once. */
850        if (common->recursive_head_ptr == 0)
851        {        {
852        length += 2;        common->recursive_head_ptr = common->ovector_start;
853        setsom_found = TRUE;        common->ovector_start += sizeof(sljit_sw);
854        }        }
855      cc += (*cc == OP_SET_SOM) ? 1 : 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
856      break;      break;
857    
858      case OP_ASSERT:      case OP_CALLOUT:
859      case OP_ASSERT_NOT:      if (common->capture_last_ptr == 0)
     case OP_ASSERTBACK:  
     case OP_ASSERTBACK_NOT:  
     case OP_ONCE:  
     if (needs_frame || length > 0)  
860        {        {
861        cc = bracketend(cc);        common->capture_last_ptr = common->ovector_start;
862        break;        common->ovector_start += sizeof(sljit_sw);
       }  
     /* Check whether a frame must be created. */  
     end = bracketend(cc);  
     while (cc < end)  
       {  
       if (*cc == OP_SET_SOM || *cc == OP_CBRA || *cc == OP_CBRAPOS  
           || *cc == OP_SCBRA || *cc == OP_SCBRAPOS || *cc == OP_RECURSE)  
         needs_frame = TRUE;  
       cc = next_opcode(common, cc);  
       SLJIT_ASSERT(cc != NULL);  
863        }        }
864        cc += 2 + 2 * LINK_SIZE;
865      break;      break;
866    
867      case OP_CBRA:      case OP_THEN_ARG:
868      case OP_CBRAPOS:      common->has_then = TRUE;
869      case OP_SCBRA:      common->control_head_ptr = 1;
870      case OP_SCBRAPOS:      /* Fall through. */
871      if (!needs_maxindex)  
872        case OP_PRUNE_ARG:
873        common->needs_start_ptr = TRUE;
874        /* Fall through. */
875    
876        case OP_MARK:
877        if (common->mark_ptr == 0)
878        {        {
879        needs_maxindex = TRUE;        common->mark_ptr = common->ovector_start;
880        length += 2;        common->ovector_start += sizeof(sljit_sw);
881        }        }
882      length += 3;      cc += 1 + 2 + cc[1];
883      cc += 1 + LINK_SIZE + 2;      break;
884    
885        case OP_THEN:
886        common->has_then = TRUE;
887        common->control_head_ptr = 1;
888        /* Fall through. */
889    
890        case OP_PRUNE:
891        case OP_SKIP:
892        common->needs_start_ptr = TRUE;
893        cc += 1;
894        break;
895    
896        case OP_SKIP_ARG:
897        common->control_head_ptr = 1;
898        common->has_skip_arg = TRUE;
899        cc += 1 + 2 + cc[1];
900      break;      break;
901    
902      default:      default:
903      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
904      SLJIT_ASSERT(cc != NULL);      if (cc == NULL)
905          return FALSE;
906      break;      break;
907      }      }
908      }
909    return TRUE;
910    }
911    
912  /* Possessive quantifiers can use a special case. */  static int get_class_iterator_size(pcre_uchar *cc)
913  if (SLJIT_UNLIKELY(possessive) && !needs_frame && length == 3 + 2)  {
914    return -1;  switch(*cc)
915      {
916      case OP_CRSTAR:
917      case OP_CRPLUS:
918      return 2;
919    
920  if (length > 0)    case OP_CRMINSTAR:
921    return length + 2;    case OP_CRMINPLUS:
922  return needs_frame ? 0 : -1;    case OP_CRQUERY:
923      case OP_CRMINQUERY:
924      return 1;
925    
926      case OP_CRRANGE:
927      case OP_CRMINRANGE:
928      if (GET2(cc, 1) == GET2(cc, 1 + IMM2_SIZE))
929        return 0;
930      return 2;
931    
932      default:
933      return 0;
934      }
935  }  }
936    
937  static void init_frame(compiler_common *common, uschar *cc, int stackpos, int stacktop, BOOL recursive)  static BOOL detect_repeat(compiler_common *common, pcre_uchar *begin)
938  {  {
939  /* TMP2 must contain STACK_TOP - (-STACK(stackpos)) */  pcre_uchar *end = bracketend(begin);
940  DEFINE_COMPILER;  pcre_uchar *next;
941  uschar *ccend = bracketend(cc);  pcre_uchar *next_end;
942  BOOL needs_maxindex = FALSE;  pcre_uchar *max_end;
943  BOOL setsom_found = FALSE;  pcre_uchar type;
944  int offset;  sljit_sw length = end - begin;
945    int min, max, i;
946    
947    /* Detect fixed iterations first. */
948    if (end[-(1 + LINK_SIZE)] != OP_KET)
949      return FALSE;
950    
951    /* Already detected repeat. */
952    if (common->private_data_ptrs[end - common->start - LINK_SIZE] != 0)
953      return TRUE;
954    
955  if (stackpos < stacktop)  next = end;
956    min = 1;
957    while (1)
958    {    {
959    SLJIT_ASSERT(stackpos + 1 == stacktop);    if (*next != *begin)
960    return;      break;
961      next_end = bracketend(next);
962      if (next_end - next != length || memcmp(begin, next, IN_UCHARS(length)) != 0)
963        break;
964      next = next_end;
965      min++;
966    }    }
967    
968  stackpos = STACK(stackpos);  if (min == 2)
969  OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS_HEAD);    return FALSE;
 OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS_HEAD, TMP2, 0);  
 OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), STACK(stacktop), TMP1, 0);  
970    
971  if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))  max = 0;
972    cc = next_opcode(common, cc);  max_end = next;
973  SLJIT_ASSERT(cc != NULL);  if (*next == OP_BRAZERO || *next == OP_BRAMINZERO)
974  while (cc < ccend)    {
975    switch(*cc)    type = *next;
976      while (1)
977      {      {
978      case OP_SET_SOM:      if (next[0] != type || next[1] != OP_BRA || next[2 + LINK_SIZE] != *begin)
979      case OP_RECURSE:        break;
980      if (!setsom_found)      next_end = bracketend(next + 2 + LINK_SIZE);
981        if (next_end - next != (length + 2 + LINK_SIZE) || memcmp(begin, next + 2 + LINK_SIZE, IN_UCHARS(length)) != 0)
982          break;
983        next = next_end;
984        max++;
985        }
986    
987      if (next[0] == type && next[1] == *begin && max >= 1)
988        {
989        next_end = bracketend(next + 1);
990        if (next_end - next == (length + 1) && memcmp(begin, next + 1, IN_UCHARS(length)) == 0)
991        {        {
992        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));        for (i = 0; i < max; i++, next_end += 1 + LINK_SIZE)
993        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setstrbegin);          if (*next_end != OP_KET)
994        stackpos += (int)sizeof(sljit_w);            break;
995        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);  
996        stackpos += (int)sizeof(sljit_w);        if (i == max)
997        setsom_found = TRUE;          {
998            common->private_data_ptrs[max_end - common->start - LINK_SIZE] = next_end - max_end;
999            common->private_data_ptrs[max_end - common->start - LINK_SIZE + 1] = (type == OP_BRAZERO) ? OP_UPTO : OP_MINUPTO;
1000            /* +2 the original and the last. */
1001            common->private_data_ptrs[max_end - common->start - LINK_SIZE + 2] = max + 2;
1002            if (min == 1)
1003              return TRUE;
1004            min--;
1005            max_end -= (1 + LINK_SIZE) + GET(max_end, -LINK_SIZE);
1006            }
1007        }        }
1008      cc += (*cc == OP_SET_SOM) ? 1 : 1 + LINK_SIZE;      }
1009      break;    }
1010    
1011      case OP_ASSERT:  if (min >= 3)
1012      case OP_ASSERT_NOT:    {
1013      case OP_ASSERTBACK:    common->private_data_ptrs[end - common->start - LINK_SIZE] = max_end - end;
1014      case OP_ASSERTBACK_NOT:    common->private_data_ptrs[end - common->start - LINK_SIZE + 1] = OP_EXACT;
1015      case OP_ONCE:    common->private_data_ptrs[end - common->start - LINK_SIZE + 2] = min;
1016      cc = bracketend(cc);    return TRUE;
1017      break;    }
1018    
1019    return FALSE;
1020    }
1021    
1022    #define CASE_ITERATOR_PRIVATE_DATA_1 \
1023        case OP_MINSTAR: \
1024        case OP_MINPLUS: \
1025        case OP_QUERY: \
1026        case OP_MINQUERY: \
1027        case OP_MINSTARI: \
1028        case OP_MINPLUSI: \
1029        case OP_QUERYI: \
1030        case OP_MINQUERYI: \
1031        case OP_NOTMINSTAR: \
1032        case OP_NOTMINPLUS: \
1033        case OP_NOTQUERY: \
1034        case OP_NOTMINQUERY: \
1035        case OP_NOTMINSTARI: \
1036        case OP_NOTMINPLUSI: \
1037        case OP_NOTQUERYI: \
1038        case OP_NOTMINQUERYI:
1039    
1040      case OP_CBRA:  #define CASE_ITERATOR_PRIVATE_DATA_2A \
1041      case OP_CBRAPOS:      case OP_STAR: \
1042      case OP_SCBRA:      case OP_PLUS: \
1043      case OP_SCBRAPOS:      case OP_STARI: \
1044      if (!needs_maxindex)      case OP_PLUSI: \
1045        {      case OP_NOTSTAR: \
1046        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), MAX_INDEX);      case OP_NOTPLUS: \
1047        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_setmaxindex);      case OP_NOTSTARI: \
1048        stackpos += (int)sizeof(sljit_w);      case OP_NOTPLUSI:
       OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);  
       stackpos += (int)sizeof(sljit_w);  
       needs_maxindex = TRUE;  
       }  
     offset = (GET2(cc, 1 + LINK_SIZE)) << 1;  
     OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));  
     stackpos += (int)sizeof(sljit_w);  
     OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset));  
     OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset + 1));  
     OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);  
     stackpos += (int)sizeof(sljit_w);  
     OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);  
     stackpos += (int)sizeof(sljit_w);  
1049    
1050      cc += 1 + LINK_SIZE + 2;  #define CASE_ITERATOR_PRIVATE_DATA_2B \
1051      break;      case OP_UPTO: \
1052        case OP_MINUPTO: \
1053        case OP_UPTOI: \
1054        case OP_MINUPTOI: \
1055        case OP_NOTUPTO: \
1056        case OP_NOTMINUPTO: \
1057        case OP_NOTUPTOI: \
1058        case OP_NOTMINUPTOI:
1059    
1060      default:  #define CASE_ITERATOR_TYPE_PRIVATE_DATA_1 \
1061      cc = next_opcode(common, cc);      case OP_TYPEMINSTAR: \
1062      SLJIT_ASSERT(cc != NULL);      case OP_TYPEMINPLUS: \
1063      break;      case OP_TYPEQUERY: \
1064      }      case OP_TYPEMINQUERY:
1065    
1066  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, frame_end);  #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2A \
1067  SLJIT_ASSERT(stackpos == STACK(stacktop + 1));      case OP_TYPESTAR: \
1068  }      case OP_TYPEPLUS:
1069    
1070    #define CASE_ITERATOR_TYPE_PRIVATE_DATA_2B \
1071        case OP_TYPEUPTO: \
1072        case OP_TYPEMINUPTO:
1073    
1074  static SLJIT_INLINE int get_localsize(compiler_common *common, uschar *cc, uschar *ccend)  static void set_private_data_ptrs(compiler_common *common, int *private_data_start, pcre_uchar *ccend)
1075  {  {
1076  int localsize = 2;  pcre_uchar *cc = common->start;
1077  uschar *alternative;  pcre_uchar *alternative;
1078  /* Calculate the sum of the local variables. */  pcre_uchar *end = NULL;
1079    int private_data_ptr = *private_data_start;
1080    int space, size, bracketlen;
1081    
1082  while (cc < ccend)  while (cc < ccend)
1083    {    {
1084      space = 0;
1085      size = 0;
1086      bracketlen = 0;
1087      if (private_data_ptr > SLJIT_MAX_LOCAL_SIZE)
1088        return;
1089    
1090      if (*cc == OP_ONCE || *cc == OP_ONCE_NC || *cc == OP_BRA || *cc == OP_CBRA || *cc == OP_COND)
1091        if (detect_repeat(common, cc))
1092          {
1093          /* These brackets are converted to repeats, so no global
1094          based single character repeat is allowed. */
1095          if (cc >= end)
1096            end = bracketend(cc);
1097          }
1098    
1099    switch(*cc)    switch(*cc)
1100      {      {
1101        case OP_KET:
1102        if (common->private_data_ptrs[cc + 1 - common->start] != 0)
1103          {
1104          common->private_data_ptrs[cc - common->start] = private_data_ptr;
1105          private_data_ptr += sizeof(sljit_sw);
1106          cc += common->private_data_ptrs[cc + 1 - common->start];
1107          }
1108        cc += 1 + LINK_SIZE;
1109        break;
1110    
1111      case OP_ASSERT:      case OP_ASSERT:
1112      case OP_ASSERT_NOT:      case OP_ASSERT_NOT:
1113      case OP_ASSERTBACK:      case OP_ASSERTBACK:
1114      case OP_ASSERTBACK_NOT:      case OP_ASSERTBACK_NOT:
1115      case OP_ONCE:      case OP_ONCE:
1116        case OP_ONCE_NC:
1117      case OP_BRAPOS:      case OP_BRAPOS:
1118      case OP_SBRA:      case OP_SBRA:
1119      case OP_SBRAPOS:      case OP_SBRAPOS:
1120      case OP_SCOND:      case OP_SCOND:
1121      localsize++;      common->private_data_ptrs[cc - common->start] = private_data_ptr;
1122      cc += 1 + LINK_SIZE;      private_data_ptr += sizeof(sljit_sw);
1123      break;      bracketlen = 1 + LINK_SIZE;
   
     case OP_CBRA:  
     case OP_SCBRA:  
     localsize++;  
     cc += 1 + LINK_SIZE + 2;  
1124      break;      break;
1125    
1126      case OP_CBRAPOS:      case OP_CBRAPOS:
1127      case OP_SCBRAPOS:      case OP_SCBRAPOS:
1128      localsize += 2;      common->private_data_ptrs[cc - common->start] = private_data_ptr;
1129      cc += 1 + LINK_SIZE + 2;      private_data_ptr += sizeof(sljit_sw);
1130        bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
1131      break;      break;
1132    
1133      case OP_COND:      case OP_COND:
1134      /* Might be a hidden SCOND. */      /* Might be a hidden SCOND. */
1135      alternative = cc + GET(cc, 1);      alternative = cc + GET(cc, 1);
1136      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)      if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1137        localsize++;        {
1138      cc += 1 + LINK_SIZE;        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1139          private_data_ptr += sizeof(sljit_sw);
1140          }
1141        bracketlen = 1 + LINK_SIZE;
1142        break;
1143    
1144        case OP_BRA:
1145        bracketlen = 1 + LINK_SIZE;
1146        break;
1147    
1148        case OP_CBRA:
1149        case OP_SCBRA:
1150        bracketlen = 1 + LINK_SIZE + IMM2_SIZE;
1151        break;
1152    
1153        CASE_ITERATOR_PRIVATE_DATA_1
1154        space = 1;
1155        size = -2;
1156        break;
1157    
1158        CASE_ITERATOR_PRIVATE_DATA_2A
1159        space = 2;
1160        size = -2;
1161        break;
1162    
1163        CASE_ITERATOR_PRIVATE_DATA_2B
1164        space = 2;
1165        size = -(2 + IMM2_SIZE);
1166        break;
1167    
1168        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1169        space = 1;
1170        size = 1;
1171        break;
1172    
1173        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1174        if (cc[1] != OP_ANYNL && cc[1] != OP_EXTUNI)
1175          space = 2;
1176        size = 1;
1177        break;
1178    
1179        CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1180        if (cc[1 + IMM2_SIZE] != OP_ANYNL && cc[1 + IMM2_SIZE] != OP_EXTUNI)
1181          space = 2;
1182        size = 1 + IMM2_SIZE;
1183        break;
1184    
1185        case OP_CLASS:
1186        case OP_NCLASS:
1187        size += 1 + 32 / sizeof(pcre_uchar);
1188        space = get_class_iterator_size(cc + size);
1189        break;
1190    
1191    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1192        case OP_XCLASS:
1193        size = GET(cc, 1);
1194        space = get_class_iterator_size(cc + size);
1195      break;      break;
1196    #endif
1197    
1198      default:      default:
1199      cc = next_opcode(common, cc);      cc = next_opcode(common, cc);
1200      SLJIT_ASSERT(cc != NULL);      SLJIT_ASSERT(cc != NULL);
1201      break;      break;
1202      }      }
   }  
 SLJIT_ASSERT(cc == ccend);  
 return localsize;  
 }  
   
 static void copy_locals(compiler_common *common, uschar *cc, uschar *ccend,  
   BOOL save, int stackptr, int stacktop)  
 {  
 DEFINE_COMPILER;  
 int srcw[2];  
 int count;  
 BOOL tmp1next = TRUE;  
 BOOL tmp1empty = TRUE;  
 BOOL tmp2empty = TRUE;  
 uschar *alternative;  
 enum {  
   start,  
   loop,  
   end  
 } status;  
1203    
1204  status = save ? start : loop;    /* Character iterators, which are not inside a repeated bracket,
1205  stackptr = STACK(stackptr - 2);       gets a private slot instead of allocating it on the stack. */
1206  stacktop = STACK(stacktop - 1);    if (space > 0 && cc >= end)
1207        {
1208        common->private_data_ptrs[cc - common->start] = private_data_ptr;
1209        private_data_ptr += sizeof(sljit_sw) * space;
1210        }
1211    
1212  if (!save)    if (size != 0)
   {  
   stackptr += sizeof(sljit_w);  
   if (stackptr < stacktop)  
1213      {      {
1214      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);      if (size < 0)
1215      stackptr += sizeof(sljit_w);        {
1216      tmp1empty = FALSE;        cc += -size;
1217    #ifdef SUPPORT_UTF
1218          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1219    #endif
1220          }
1221        else
1222          cc += size;
1223      }      }
1224    if (stackptr < stacktop)  
1225      if (bracketlen > 0)
1226      {      {
1227      OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);      if (cc >= end)
1228      stackptr += sizeof(sljit_w);        {
1229      tmp2empty = FALSE;        end = bracketend(cc);
1230          if (end[-1 - LINK_SIZE] == OP_KET)
1231            end = NULL;
1232          }
1233        cc += bracketlen;
1234      }      }
   /* The tmp1next must be TRUE in either way. */  
1235    }    }
1236    *private_data_start = private_data_ptr;
1237    }
1238    
1239    /* Returns with a frame_types (always < 0) if no need for frame. */
1240    static int get_framesize(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, BOOL recursive, BOOL* needs_control_head)
1241    {
1242    int length = 0;
1243    int possessive = 0;
1244    BOOL stack_restore = FALSE;
1245    BOOL setsom_found = recursive;
1246    BOOL setmark_found = recursive;
1247    /* The last capture is a local variable even for recursions. */
1248    BOOL capture_last_found = FALSE;
1249    
1250    #if defined DEBUG_FORCE_CONTROL_HEAD && DEBUG_FORCE_CONTROL_HEAD
1251    SLJIT_ASSERT(common->control_head_ptr != 0);
1252    *needs_control_head = TRUE;
1253    #else
1254    *needs_control_head = FALSE;
1255    #endif
1256    
1257  while (status != end)  if (ccend == NULL)
1258    {    {
1259    count = 0;    ccend = bracketend(cc) - (1 + LINK_SIZE);
1260    switch(status)    if (!recursive && (*cc == OP_CBRAPOS || *cc == OP_SCBRAPOS))
1261      {      {
1262      case start:      possessive = length = (common->capture_last_ptr != 0) ? 5 : 3;
1263      SLJIT_ASSERT(save);      /* This is correct regardless of common->capture_last_ptr. */
1264      count = 1;      capture_last_found = TRUE;
1265      srcw[0] = RECURSIVE_HEAD;      }
1266      status = loop;    cc = next_opcode(common, cc);
1267      break;    }
1268    
1269      case loop:  SLJIT_ASSERT(cc != NULL);
1270      if (cc >= ccend)  while (cc < ccend)
1271      switch(*cc)
1272        {
1273        case OP_SET_SOM:
1274        SLJIT_ASSERT(common->has_set_som);
1275        stack_restore = TRUE;
1276        if (!setsom_found)
1277        {        {
1278        status = end;        length += 2;
1279        break;        setsom_found = TRUE;
1280        }        }
1281        cc += 1;
1282        break;
1283    
1284      switch(*cc)      case OP_MARK:
1285        case OP_PRUNE_ARG:
1286        case OP_THEN_ARG:
1287        SLJIT_ASSERT(common->mark_ptr != 0);
1288        stack_restore = TRUE;
1289        if (!setmark_found)
1290        {        {
1291        case OP_ASSERT:        length += 2;
1292        case OP_ASSERT_NOT:        setmark_found = TRUE;
       case OP_ASSERTBACK:  
       case OP_ASSERTBACK_NOT:  
       case OP_ONCE:  
       case OP_BRAPOS:  
       case OP_SBRA:  
       case OP_SBRAPOS:  
       case OP_SCOND:  
       count = 1;  
       srcw[0] = PRIV(cc);  
       SLJIT_ASSERT(srcw[0] != 0);  
       cc += 1 + LINK_SIZE;  
       break;  
   
       case OP_CBRA:  
       case OP_SCBRA:  
       count = 1;  
       srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));  
       cc += 1 + LINK_SIZE + 2;  
       break;  
   
       case OP_CBRAPOS:  
       case OP_SCBRAPOS:  
       count = 2;  
       srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));  
       srcw[0] = PRIV(cc);  
       SLJIT_ASSERT(srcw[0] != 0);  
       cc += 1 + LINK_SIZE + 2;  
       break;  
   
       case OP_COND:  
       /* Might be a hidden SCOND. */  
       alternative = cc + GET(cc, 1);  
       if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)  
         {  
         count = 1;  
         srcw[0] = PRIV(cc);  
         SLJIT_ASSERT(srcw[0] != 0);  
         }  
       cc += 1 + LINK_SIZE;  
       break;  
   
       default:  
       cc = next_opcode(common, cc);  
       SLJIT_ASSERT(cc != NULL);  
       break;  
1293        }        }
1294        if (common->control_head_ptr != 0)
1295          *needs_control_head = TRUE;
1296        cc += 1 + 2 + cc[1];
1297      break;      break;
1298    
1299      case end:      case OP_RECURSE:
1300      SLJIT_ASSERT_STOP();      stack_restore = TRUE;
1301        if (common->has_set_som && !setsom_found)
1302          {
1303          length += 2;
1304          setsom_found = TRUE;
1305          }
1306        if (common->mark_ptr != 0 && !setmark_found)
1307          {
1308          length += 2;
1309          setmark_found = TRUE;
1310          }
1311        if (common->capture_last_ptr != 0 && !capture_last_found)
1312          {
1313          length += 2;
1314          capture_last_found = TRUE;
1315          }
1316        cc += 1 + LINK_SIZE;
1317      break;      break;
     }  
1318    
1319    while (count > 0)      case OP_CBRA:
1320      {      case OP_CBRAPOS:
1321      count--;      case OP_SCBRA:
1322      if (save)      case OP_SCBRAPOS:
1323        {      stack_restore = TRUE;
1324        if (tmp1next)      if (common->capture_last_ptr != 0 && !capture_last_found)
         {  
         if (!tmp1empty)  
           {  
           OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);  
           stackptr += sizeof(sljit_w);  
           }  
         OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);  
         tmp1empty = FALSE;  
         tmp1next = FALSE;  
         }  
       else  
         {  
         if (!tmp2empty)  
           {  
           OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);  
           stackptr += sizeof(sljit_w);  
           }  
         OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);  
         tmp2empty = FALSE;  
         tmp1next = TRUE;  
         }  
       }  
     else  
1325        {        {
1326        if (tmp1next)        length += 2;
1327          {        capture_last_found = TRUE;
         SLJIT_ASSERT(!tmp1empty);  
         OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count], TMP1, 0);  
         tmp1empty = stackptr >= stacktop;  
         if (!tmp1empty)  
           {  
           OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);  
           stackptr += sizeof(sljit_w);  
           }  
         tmp1next = FALSE;  
         }  
       else  
         {  
         SLJIT_ASSERT(!tmp2empty);  
         OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count], TMP2, 0);  
         tmp2empty = stackptr >= stacktop;  
         if (!tmp2empty)  
           {  
           OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);  
           stackptr += sizeof(sljit_w);  
           }  
         tmp1next = TRUE;  
         }  
1328        }        }
1329        length += 3;
1330        cc += 1 + LINK_SIZE + IMM2_SIZE;
1331        break;
1332    
1333        default:
1334        stack_restore = TRUE;
1335        /* Fall through. */
1336    
1337        case OP_NOT_WORD_BOUNDARY:
1338        case OP_WORD_BOUNDARY:
1339        case OP_NOT_DIGIT:
1340        case OP_DIGIT:
1341        case OP_NOT_WHITESPACE:
1342        case OP_WHITESPACE:
1343        case OP_NOT_WORDCHAR:
1344        case OP_WORDCHAR:
1345        case OP_ANY:
1346        case OP_ALLANY:
1347        case OP_ANYBYTE:
1348        case OP_NOTPROP:
1349        case OP_PROP:
1350        case OP_ANYNL:
1351        case OP_NOT_HSPACE:
1352        case OP_HSPACE:
1353        case OP_NOT_VSPACE:
1354        case OP_VSPACE:
1355        case OP_EXTUNI:
1356        case OP_EODN:
1357        case OP_EOD:
1358        case OP_CIRC:
1359        case OP_CIRCM:
1360        case OP_DOLL:
1361        case OP_DOLLM:
1362        case OP_CHAR:
1363        case OP_CHARI:
1364        case OP_NOT:
1365        case OP_NOTI:
1366    
1367        case OP_EXACT:
1368        case OP_POSSTAR:
1369        case OP_POSPLUS:
1370        case OP_POSQUERY:
1371        case OP_POSUPTO:
1372    
1373        case OP_EXACTI:
1374        case OP_POSSTARI:
1375        case OP_POSPLUSI:
1376        case OP_POSQUERYI:
1377        case OP_POSUPTOI:
1378    
1379        case OP_NOTEXACT:
1380        case OP_NOTPOSSTAR:
1381        case OP_NOTPOSPLUS:
1382        case OP_NOTPOSQUERY:
1383        case OP_NOTPOSUPTO:
1384    
1385        case OP_NOTEXACTI:
1386        case OP_NOTPOSSTARI:
1387        case OP_NOTPOSPLUSI:
1388        case OP_NOTPOSQUERYI:
1389        case OP_NOTPOSUPTOI:
1390    
1391        case OP_TYPEEXACT:
1392        case OP_TYPEPOSSTAR:
1393        case OP_TYPEPOSPLUS:
1394        case OP_TYPEPOSQUERY:
1395        case OP_TYPEPOSUPTO:
1396    
1397        case OP_CLASS:
1398        case OP_NCLASS:
1399        case OP_XCLASS:
1400    
1401        cc = next_opcode(common, cc);
1402        SLJIT_ASSERT(cc != NULL);
1403        break;
1404      }      }
   }  
1405    
1406  if (save)  /* Possessive quantifiers can use a special case. */
1407    if (SLJIT_UNLIKELY(possessive == length))
1408      return stack_restore ? no_frame : no_stack;
1409    
1410    if (length > 0)
1411      return length + 1;
1412    return stack_restore ? no_frame : no_stack;
1413    }
1414    
1415    static void init_frame(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, int stackpos, int stacktop, BOOL recursive)
1416    {
1417    DEFINE_COMPILER;
1418    BOOL setsom_found = recursive;
1419    BOOL setmark_found = recursive;
1420    /* The last capture is a local variable even for recursions. */
1421    BOOL capture_last_found = FALSE;
1422    int offset;
1423    
1424    /* >= 1 + shortest item size (2) */
1425    SLJIT_UNUSED_ARG(stacktop);
1426    SLJIT_ASSERT(stackpos >= stacktop + 2);
1427    
1428    stackpos = STACK(stackpos);
1429    if (ccend == NULL)
1430    {    {
1431    if (tmp1next)    ccend = bracketend(cc) - (1 + LINK_SIZE);
1432      if (recursive || (*cc != OP_CBRAPOS && *cc != OP_SCBRAPOS))
1433        cc = next_opcode(common, cc);
1434      }
1435    
1436    SLJIT_ASSERT(cc != NULL);
1437    while (cc < ccend)
1438      switch(*cc)
1439      {      {
1440      if (!tmp1empty)      case OP_SET_SOM:
1441        SLJIT_ASSERT(common->has_set_som);
1442        if (!setsom_found)
1443        {        {
1444        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));
1445        stackptr += sizeof(sljit_w);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1446          stackpos += (int)sizeof(sljit_sw);
1447          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1448          stackpos += (int)sizeof(sljit_sw);
1449          setsom_found = TRUE;
1450        }        }
1451      if (!tmp2empty)      cc += 1;
1452        break;
1453    
1454        case OP_MARK:
1455        case OP_PRUNE_ARG:
1456        case OP_THEN_ARG:
1457        SLJIT_ASSERT(common->mark_ptr != 0);
1458        if (!setmark_found)
1459          {
1460          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);
1461          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
1462          stackpos += (int)sizeof(sljit_sw);
1463          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1464          stackpos += (int)sizeof(sljit_sw);
1465          setmark_found = TRUE;
1466          }
1467        cc += 1 + 2 + cc[1];
1468        break;
1469    
1470        case OP_RECURSE:
1471        if (common->has_set_som && !setsom_found)
1472        {        {
1473        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(0));
1474        stackptr += sizeof(sljit_w);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -OVECTOR(0));
1475          stackpos += (int)sizeof(sljit_sw);
1476          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1477          stackpos += (int)sizeof(sljit_sw);
1478          setsom_found = TRUE;
1479        }        }
1480      }      if (common->mark_ptr != 0 && !setmark_found)
   else  
     {  
     if (!tmp2empty)  
1481        {        {
1482        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);
1483        stackptr += sizeof(sljit_w);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->mark_ptr);
1484          stackpos += (int)sizeof(sljit_sw);
1485          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1486          stackpos += (int)sizeof(sljit_sw);
1487          setmark_found = TRUE;
1488        }        }
1489      if (!tmp1empty)      if (common->capture_last_ptr != 0 && !capture_last_found)
1490        {        {
1491        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->capture_last_ptr);
1492        stackptr += sizeof(sljit_w);        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1493          stackpos += (int)sizeof(sljit_sw);
1494          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1495          stackpos += (int)sizeof(sljit_sw);
1496          capture_last_found = TRUE;
1497          }
1498        cc += 1 + LINK_SIZE;
1499        break;
1500    
1501        case OP_CBRA:
1502        case OP_CBRAPOS:
1503        case OP_SCBRA:
1504        case OP_SCBRAPOS:
1505        if (common->capture_last_ptr != 0 && !capture_last_found)
1506          {
1507          OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->capture_last_ptr);
1508          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, -common->capture_last_ptr);
1509          stackpos += (int)sizeof(sljit_sw);
1510          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1511          stackpos += (int)sizeof(sljit_sw);
1512          capture_last_found = TRUE;
1513        }        }
1514        offset = (GET2(cc, 1 + LINK_SIZE)) << 1;
1515        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, OVECTOR(offset));
1516        stackpos += (int)sizeof(sljit_sw);
1517        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset));
1518        OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(offset + 1));
1519        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP1, 0);
1520        stackpos += (int)sizeof(sljit_sw);
1521        OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, TMP2, 0);
1522        stackpos += (int)sizeof(sljit_sw);
1523    
1524        cc += 1 + LINK_SIZE + IMM2_SIZE;
1525        break;
1526    
1527        default:
1528        cc = next_opcode(common, cc);
1529        SLJIT_ASSERT(cc != NULL);
1530        break;
1531      }      }
1532    }  
1533  SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));  OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackpos, SLJIT_IMM, 0);
1534    SLJIT_ASSERT(stackpos == STACK(stacktop));
1535    }
1536    
1537    static SLJIT_INLINE int get_private_data_copy_length(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend, BOOL needs_control_head)
1538    {
1539    int private_data_length = needs_control_head ? 3 : 2;
1540    int size;
1541    pcre_uchar *alternative;
1542    /* Calculate the sum of the private machine words. */
1543    while (cc < ccend)
1544      {
1545      size = 0;
1546      switch(*cc)
1547        {
1548        case OP_KET:
1549        if (PRIVATE_DATA(cc) != 0)
1550          private_data_length++;
1551        cc += 1 + LINK_SIZE;
1552        break;
1553    
1554        case OP_ASSERT:
1555        case OP_ASSERT_NOT:
1556        case OP_ASSERTBACK:
1557        case OP_ASSERTBACK_NOT:
1558        case OP_ONCE:
1559        case OP_ONCE_NC:
1560        case OP_BRAPOS:
1561        case OP_SBRA:
1562        case OP_SBRAPOS:
1563        case OP_SCOND:
1564        private_data_length++;
1565        cc += 1 + LINK_SIZE;
1566        break;
1567    
1568        case OP_CBRA:
1569        case OP_SCBRA:
1570        if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1571          private_data_length++;
1572        cc += 1 + LINK_SIZE + IMM2_SIZE;
1573        break;
1574    
1575        case OP_CBRAPOS:
1576        case OP_SCBRAPOS:
1577        private_data_length += 2;
1578        cc += 1 + LINK_SIZE + IMM2_SIZE;
1579        break;
1580    
1581        case OP_COND:
1582        /* Might be a hidden SCOND. */
1583        alternative = cc + GET(cc, 1);
1584        if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1585          private_data_length++;
1586        cc += 1 + LINK_SIZE;
1587        break;
1588    
1589        CASE_ITERATOR_PRIVATE_DATA_1
1590        if (PRIVATE_DATA(cc))
1591          private_data_length++;
1592        cc += 2;
1593    #ifdef SUPPORT_UTF
1594        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1595    #endif
1596        break;
1597    
1598        CASE_ITERATOR_PRIVATE_DATA_2A
1599        if (PRIVATE_DATA(cc))
1600          private_data_length += 2;
1601        cc += 2;
1602    #ifdef SUPPORT_UTF
1603        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1604    #endif
1605        break;
1606    
1607        CASE_ITERATOR_PRIVATE_DATA_2B
1608        if (PRIVATE_DATA(cc))
1609          private_data_length += 2;
1610        cc += 2 + IMM2_SIZE;
1611    #ifdef SUPPORT_UTF
1612        if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1613    #endif
1614        break;
1615    
1616        CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1617        if (PRIVATE_DATA(cc))
1618          private_data_length++;
1619        cc += 1;
1620        break;
1621    
1622        CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1623        if (PRIVATE_DATA(cc))
1624          private_data_length += 2;
1625        cc += 1;
1626        break;
1627    
1628        CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1629        if (PRIVATE_DATA(cc))
1630          private_data_length += 2;
1631        cc += 1 + IMM2_SIZE;
1632        break;
1633    
1634        case OP_CLASS:
1635        case OP_NCLASS:
1636    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1637        case OP_XCLASS:
1638        size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
1639    #else
1640        size = 1 + 32 / (int)sizeof(pcre_uchar);
1641    #endif
1642        if (PRIVATE_DATA(cc))
1643          private_data_length += get_class_iterator_size(cc + size);
1644        cc += size;
1645        break;
1646    
1647        default:
1648        cc = next_opcode(common, cc);
1649        SLJIT_ASSERT(cc != NULL);
1650        break;
1651        }
1652      }
1653    SLJIT_ASSERT(cc == ccend);
1654    return private_data_length;
1655    }
1656    
1657    static void copy_private_data(compiler_common *common, pcre_uchar *cc, pcre_uchar *ccend,
1658      BOOL save, int stackptr, int stacktop, BOOL needs_control_head)
1659    {
1660    DEFINE_COMPILER;
1661    int srcw[2];
1662    int count, size;
1663    BOOL tmp1next = TRUE;
1664    BOOL tmp1empty = TRUE;
1665    BOOL tmp2empty = TRUE;
1666    pcre_uchar *alternative;
1667    enum {
1668      start,
1669      loop,
1670      end
1671    } status;
1672    
1673    status = save ? start : loop;
1674    stackptr = STACK(stackptr - 2);
1675    stacktop = STACK(stacktop - 1);
1676    
1677    if (!save)
1678      {
1679      stackptr += (needs_control_head ? 2 : 1) * sizeof(sljit_sw);
1680      if (stackptr < stacktop)
1681        {
1682        OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1683        stackptr += sizeof(sljit_sw);
1684        tmp1empty = FALSE;
1685        }
1686      if (stackptr < stacktop)
1687        {
1688        OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1689        stackptr += sizeof(sljit_sw);
1690        tmp2empty = FALSE;
1691        }
1692      /* The tmp1next must be TRUE in either way. */
1693      }
1694    
1695    do
1696      {
1697      count = 0;
1698      switch(status)
1699        {
1700        case start:
1701        SLJIT_ASSERT(save && common->recursive_head_ptr != 0);
1702        count = 1;
1703        srcw[0] = common->recursive_head_ptr;
1704        if (needs_control_head)
1705          {
1706          SLJIT_ASSERT(common->control_head_ptr != 0);
1707          count = 2;
1708          srcw[1] = common->control_head_ptr;
1709          }
1710        status = loop;
1711        break;
1712    
1713        case loop:
1714        if (cc >= ccend)
1715          {
1716          status = end;
1717          break;
1718          }
1719    
1720        switch(*cc)
1721          {
1722          case OP_KET:
1723          if (PRIVATE_DATA(cc) != 0)
1724            {
1725            count = 1;
1726            srcw[0] = PRIVATE_DATA(cc);
1727            }
1728          cc += 1 + LINK_SIZE;
1729          break;
1730    
1731          case OP_ASSERT:
1732          case OP_ASSERT_NOT:
1733          case OP_ASSERTBACK:
1734          case OP_ASSERTBACK_NOT:
1735          case OP_ONCE:
1736          case OP_ONCE_NC:
1737          case OP_BRAPOS:
1738          case OP_SBRA:
1739          case OP_SBRAPOS:
1740          case OP_SCOND:
1741          count = 1;
1742          srcw[0] = PRIVATE_DATA(cc);
1743          SLJIT_ASSERT(srcw[0] != 0);
1744          cc += 1 + LINK_SIZE;
1745          break;
1746    
1747          case OP_CBRA:
1748          case OP_SCBRA:
1749          if (common->optimized_cbracket[GET2(cc, 1 + LINK_SIZE)] == 0)
1750            {
1751            count = 1;
1752            srcw[0] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1753            }
1754          cc += 1 + LINK_SIZE + IMM2_SIZE;
1755          break;
1756    
1757          case OP_CBRAPOS:
1758          case OP_SCBRAPOS:
1759          count = 2;
1760          srcw[0] = PRIVATE_DATA(cc);
1761          srcw[1] = OVECTOR_PRIV(GET2(cc, 1 + LINK_SIZE));
1762          SLJIT_ASSERT(srcw[0] != 0 && srcw[1] != 0);
1763          cc += 1 + LINK_SIZE + IMM2_SIZE;
1764          break;
1765    
1766          case OP_COND:
1767          /* Might be a hidden SCOND. */
1768          alternative = cc + GET(cc, 1);
1769          if (*alternative == OP_KETRMAX || *alternative == OP_KETRMIN)
1770            {
1771            count = 1;
1772            srcw[0] = PRIVATE_DATA(cc);
1773            SLJIT_ASSERT(srcw[0] != 0);
1774            }
1775          cc += 1 + LINK_SIZE;
1776          break;
1777    
1778          CASE_ITERATOR_PRIVATE_DATA_1
1779          if (PRIVATE_DATA(cc))
1780            {
1781            count = 1;
1782            srcw[0] = PRIVATE_DATA(cc);
1783            }
1784          cc += 2;
1785    #ifdef SUPPORT_UTF
1786          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1787    #endif
1788          break;
1789    
1790          CASE_ITERATOR_PRIVATE_DATA_2A
1791          if (PRIVATE_DATA(cc))
1792            {
1793            count = 2;
1794            srcw[0] = PRIVATE_DATA(cc);
1795            srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1796            }
1797          cc += 2;
1798    #ifdef SUPPORT_UTF
1799          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1800    #endif
1801          break;
1802    
1803          CASE_ITERATOR_PRIVATE_DATA_2B
1804          if (PRIVATE_DATA(cc))
1805            {
1806            count = 2;
1807            srcw[0] = PRIVATE_DATA(cc);
1808            srcw[1] = PRIVATE_DATA(cc) + sizeof(sljit_sw);
1809            }
1810          cc += 2 + IMM2_SIZE;
1811    #ifdef SUPPORT_UTF
1812          if (common->utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1813    #endif
1814          break;
1815    
1816          CASE_ITERATOR_TYPE_PRIVATE_DATA_1
1817          if (PRIVATE_DATA(cc))
1818            {
1819            count = 1;
1820            srcw[0] = PRIVATE_DATA(cc);
1821            }
1822          cc += 1;
1823          break;
1824    
1825          CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
1826          if (PRIVATE_DATA(cc))
1827            {
1828            count = 2;
1829            srcw[0] = PRIVATE_DATA(cc);
1830            srcw[1] = srcw[0] + sizeof(sljit_sw);
1831            }
1832          cc += 1;
1833          break;
1834    
1835          CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
1836          if (PRIVATE_DATA(cc))
1837            {
1838            count = 2;
1839            srcw[0] = PRIVATE_DATA(cc);
1840            srcw[1] = srcw[0] + sizeof(sljit_sw);
1841            }
1842          cc += 1 + IMM2_SIZE;
1843          break;
1844    
1845          case OP_CLASS:
1846          case OP_NCLASS:
1847    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1848          case OP_XCLASS:
1849          size = (*cc == OP_XCLASS) ? GET(cc, 1) : 1 + 32 / (int)sizeof(pcre_uchar);
1850    #else
1851          size = 1 + 32 / (int)sizeof(pcre_uchar);
1852    #endif
1853          if (PRIVATE_DATA(cc))
1854            switch(get_class_iterator_size(cc + size))
1855              {
1856              case 1:
1857              count = 1;
1858              srcw[0] = PRIVATE_DATA(cc);
1859              break;
1860    
1861              case 2:
1862              count = 2;
1863              srcw[0] = PRIVATE_DATA(cc);
1864              srcw[1] = srcw[0] + sizeof(sljit_sw);
1865              break;
1866    
1867              default:
1868              SLJIT_ASSERT_STOP();
1869              break;
1870              }
1871          cc += size;
1872          break;
1873    
1874          default:
1875          cc = next_opcode(common, cc);
1876          SLJIT_ASSERT(cc != NULL);
1877          break;
1878          }
1879        break;
1880    
1881        case end:
1882        SLJIT_ASSERT_STOP();
1883        break;
1884        }
1885    
1886      while (count > 0)
1887        {
1888        count--;
1889        if (save)
1890          {
1891          if (tmp1next)
1892            {
1893            if (!tmp1empty)
1894              {
1895              OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1896              stackptr += sizeof(sljit_sw);
1897              }
1898            OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);
1899            tmp1empty = FALSE;
1900            tmp1next = FALSE;
1901            }
1902          else
1903            {
1904            if (!tmp2empty)
1905              {
1906              OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1907              stackptr += sizeof(sljit_sw);
1908              }
1909            OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count]);
1910            tmp2empty = FALSE;
1911            tmp1next = TRUE;
1912            }
1913          }
1914        else
1915          {
1916          if (tmp1next)
1917            {
1918            SLJIT_ASSERT(!tmp1empty);
1919            OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count], TMP1, 0);
1920            tmp1empty = stackptr >= stacktop;
1921            if (!tmp1empty)
1922              {
1923              OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1924              stackptr += sizeof(sljit_sw);
1925              }
1926            tmp1next = FALSE;
1927            }
1928          else
1929            {
1930            SLJIT_ASSERT(!tmp2empty);
1931            OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), srcw[count], TMP2, 0);
1932            tmp2empty = stackptr >= stacktop;
1933            if (!tmp2empty)
1934              {
1935              OP1(SLJIT_MOV, TMP2, 0, SLJIT_MEM1(STACK_TOP), stackptr);
1936              stackptr += sizeof(sljit_sw);
1937              }
1938            tmp1next = TRUE;
1939            }
1940          }
1941        }
1942      }
1943    while (status != end);
1944    
1945    if (save)
1946      {
1947      if (tmp1next)
1948        {
1949        if (!tmp1empty)
1950          {
1951          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1952          stackptr += sizeof(sljit_sw);
1953          }
1954        if (!tmp2empty)
1955          {
1956          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1957          stackptr += sizeof(sljit_sw);
1958          }
1959        }
1960      else
1961        {
1962        if (!tmp2empty)
1963          {
1964          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP2, 0);
1965          stackptr += sizeof(sljit_sw);
1966          }
1967        if (!tmp1empty)
1968          {
1969          OP1(SLJIT_MOV, SLJIT_MEM1(STACK_TOP), stackptr, TMP1, 0);
1970          stackptr += sizeof(sljit_sw);
1971          }
1972        }
1973      }
1974    SLJIT_ASSERT(cc == ccend && stackptr == stacktop && (save || (tmp1empty && tmp2empty)));
1975    }
1976    
1977    static SLJIT_INLINE pcre_uchar *set_then_offsets(compiler_common *common, pcre_uchar *cc, pcre_uint8 *current_offset)
1978    {
1979    pcre_uchar *end = bracketend(cc);
1980    BOOL has_alternatives = cc[GET(cc, 1)] == OP_ALT;
1981    
1982    /* Assert captures then. */
1983    if (*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT)
1984      current_offset = NULL;
1985    /* Conditional block does not. */
1986    if (*cc == OP_COND || *cc == OP_SCOND)
1987      has_alternatives = FALSE;
1988    
1989    cc = next_opcode(common, cc);
1990    if (has_alternatives)
1991      current_offset = common->then_offsets + (cc - common->start);
1992    
1993    while (cc < end)
1994      {
1995      if ((*cc >= OP_ASSERT && *cc <= OP_ASSERTBACK_NOT) || (*cc >= OP_ONCE && *cc <= OP_SCOND))
1996        cc = set_then_offsets(common, cc, current_offset);
1997      else
1998        {
1999        if (*cc == OP_ALT && has_alternatives)
2000          current_offset = common->then_offsets + (cc + 1 + LINK_SIZE - common->start);
2001        if (*cc >= OP_THEN && *cc <= OP_THEN_ARG && current_offset != NULL)
2002          *current_offset = 1;
2003        cc = next_opcode(common, cc);
2004        }
2005      }
2006    
2007    return end;
2008    }
2009    
2010    #undef CASE_ITERATOR_PRIVATE_DATA_1
2011    #undef CASE_ITERATOR_PRIVATE_DATA_2A
2012    #undef CASE_ITERATOR_PRIVATE_DATA_2B
2013    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_1
2014    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2A
2015    #undef CASE_ITERATOR_TYPE_PRIVATE_DATA_2B
2016    
2017    static SLJIT_INLINE BOOL is_powerof2(unsigned int value)
2018    {
2019    return (value & (value - 1)) == 0;
2020    }
2021    
2022    static SLJIT_INLINE void set_jumps(jump_list *list, struct sljit_label *label)
2023    {
2024    while (list)
2025      {
2026      /* sljit_set_label is clever enough to do nothing
2027      if either the jump or the label is NULL. */
2028      SET_LABEL(list->jump, label);
2029      list = list->next;
2030      }
2031    }
2032    
2033    static SLJIT_INLINE void add_jump(struct sljit_compiler *compiler, jump_list **list, struct sljit_jump* jump)
2034    {
2035    jump_list *list_item = sljit_alloc_memory(compiler, sizeof(jump_list));
2036    if (list_item)
2037      {
2038      list_item->next = *list;
2039      list_item->jump = jump;
2040      *list = list_item;
2041      }
2042    }
2043    
2044    static void add_stub(compiler_common *common, struct sljit_jump *start)
2045    {
2046    DEFINE_COMPILER;
2047    stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));
2048    
2049    if (list_item)
2050      {
2051      list_item->start = start;
2052      list_item->quit = LABEL();
2053      list_item->next = common->stubs;
2054      common->stubs = list_item;
2055      }
2056    }
2057    
2058    static void flush_stubs(compiler_common *common)
2059    {
2060    DEFINE_COMPILER;
2061    stub_list* list_item = common->stubs;
2062    
2063    while (list_item)
2064      {
2065      JUMPHERE(list_item->start);
2066      add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));
2067      JUMPTO(SLJIT_JUMP, list_item->quit);
2068      list_item = list_item->next;
2069      }
2070    common->stubs = NULL;
2071    }
2072    
2073    static void add_label_addr(compiler_common *common)
2074    {
2075    DEFINE_COMPILER;
2076    label_addr_list *label_addr;
2077    
2078    label_addr = sljit_alloc_memory(compiler, sizeof(label_addr_list));
2079    if (label_addr == NULL)
2080      return;
2081    label_addr->label = LABEL();
2082    label_addr->addr = common->read_only_data_ptr;
2083    label_addr->next = common->label_addrs;
2084    common->label_addrs = label_addr;
2085    common->read_only_data_ptr++;
2086    }
2087    
2088    static SLJIT_INLINE void count_match(compiler_common *common)
2089    {
2090    DEFINE_COMPILER;
2091    
2092    OP2(SLJIT_SUB | SLJIT_SET_E, COUNT_MATCH, 0, COUNT_MATCH, 0, SLJIT_IMM, 1);
2093    add_jump(compiler, &common->calllimit, JUMP(SLJIT_C_ZERO));
2094    }
2095    
2096    static SLJIT_INLINE void allocate_stack(compiler_common *common, int size)
2097    {
2098    /* May destroy all locals and registers except TMP2. */
2099    DEFINE_COMPILER;
2100    
2101    OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_sw));
2102    #ifdef DESTROY_REGISTERS
2103    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);
2104    OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);
2105    OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);
2106    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);
2107    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);
2108    #endif
2109    add_stub(common, CMP(SLJIT_C_GREATER, STACK_TOP, 0, STACK_LIMIT, 0));
2110    }
2111    
2112    static SLJIT_INLINE void free_stack(compiler_common *common, int size)
2113    {
2114    DEFINE_COMPILER;
2115    OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_sw));
2116    }
2117    
2118    static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)
2119    {
2120    DEFINE_COMPILER;
2121    struct sljit_label *loop;
2122    int i;
2123    
2124    /* At this point we can freely use all temporary registers. */
2125    SLJIT_ASSERT(length > 1);
2126    /* TMP1 returns with begin - 1. */
2127    OP2(SLJIT_SUB, SLJIT_SCRATCH_REG1, 0, SLJIT_MEM1(SLJIT_SAVED_REG1), SLJIT_OFFSETOF(jit_arguments, begin), SLJIT_IMM, IN_UCHARS(1));
2128    if (length < 8)
2129      {
2130      for (i = 1; i < length; i++)
2131        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), SLJIT_SCRATCH_REG1, 0);
2132      }
2133    else
2134      {
2135      GET_LOCAL_BASE(SLJIT_SCRATCH_REG2, 0, OVECTOR_START);
2136      OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, length - 1);
2137      loop = LABEL();
2138      OP1(SLJIT_MOVU, SLJIT_MEM1(SLJIT_SCRATCH_REG2), sizeof(sljit_sw), SLJIT_SCRATCH_REG1, 0);
2139      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 1);
2140      JUMPTO(SLJIT_C_NOT_ZERO, loop);
2141      }
2142    }
2143    
2144    static SLJIT_INLINE void do_reset_match(compiler_common *common, int length)
2145    {
2146    DEFINE_COMPILER;
2147    struct sljit_label *loop;
2148    int i;
2149    
2150    SLJIT_ASSERT(length > 1);
2151    /* OVECTOR(1) contains the "string begin - 1" constant. */
2152    if (length > 2)
2153      OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));
2154    if (length < 8)
2155      {
2156      for (i = 2; i < length; i++)
2157        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), TMP1, 0);
2158      }
2159    else
2160      {
2161      GET_LOCAL_BASE(TMP2, 0, OVECTOR_START + sizeof(sljit_sw));
2162      OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_IMM, length - 2);
2163      loop = LABEL();
2164      OP1(SLJIT_MOVU, SLJIT_MEM1(TMP2), sizeof(sljit_sw), TMP1, 0);
2165      OP2(SLJIT_SUB | SLJIT_SET_E, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, 1);
2166      JUMPTO(SLJIT_C_NOT_ZERO, loop);
2167      }
2168    
2169    OP1(SLJIT_MOV, STACK_TOP, 0, ARGUMENTS, 0);
2170    if (common->mark_ptr != 0)
2171      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr, SLJIT_IMM, 0);
2172    if (common->control_head_ptr != 0)
2173      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->control_head_ptr, SLJIT_IMM, 0);
2174    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(jit_arguments, stack));
2175    OP1(SLJIT_MOV, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_ptr);
2176    OP1(SLJIT_MOV, STACK_TOP, 0, SLJIT_MEM1(STACK_TOP), SLJIT_OFFSETOF(struct sljit_stack, base));
2177    }
2178    
2179    static sljit_sw SLJIT_CALL do_search_mark(sljit_sw *current, const pcre_uchar *skip_arg)
2180    {
2181    while (current != NULL)
2182      {
2183      switch (current[-2])
2184        {
2185        case type_then_trap:
2186        break;
2187    
2188        case type_mark:
2189        if (STRCMP_UC_UC(skip_arg, (pcre_uchar *)current[-3]) == 0)
2190          return current[-4];
2191        break;
2192    
2193        default:
2194        SLJIT_ASSERT_STOP();
2195        break;
2196        }
2197      current = (sljit_sw*)current[-1];
2198      }
2199    return -1;
2200    }
2201    
2202    static SLJIT_INLINE void copy_ovector(compiler_common *common, int topbracket)
2203    {
2204    DEFINE_COMPILER;
2205    struct sljit_label *loop;
2206    struct sljit_jump *early_quit;
2207    
2208    /* At this point we can freely use all registers. */
2209    OP1(SLJIT_MOV, SLJIT_SAVED_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1));
2210    OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(1), STR_PTR, 0);
2211    
2212    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG1, 0, ARGUMENTS, 0);
2213    if (common->mark_ptr != 0)
2214      OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mark_ptr);
2215    OP1(SLJIT_MOV_SI, SLJIT_SCRATCH_REG2, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, offset_count));
2216    if (common->mark_ptr != 0)
2217      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, mark_ptr), SLJIT_SCRATCH_REG3, 0);
2218    OP2(SLJIT_SUB, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));
2219    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG1, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), SLJIT_OFFSETOF(jit_arguments, begin));
2220    GET_LOCAL_BASE(SLJIT_SAVED_REG1, 0, OVECTOR_START);
2221    /* Unlikely, but possible */
2222    early_quit = CMP(SLJIT_C_EQUAL, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, 0);
2223    loop = LABEL();
2224    OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, SLJIT_MEM1(SLJIT_SAVED_REG1), 0, SLJIT_SCRATCH_REG1, 0);
2225    OP2(SLJIT_ADD, SLJIT_SAVED_REG1, 0, SLJIT_SAVED_REG1, 0, SLJIT_IMM, sizeof(sljit_sw));
2226    /* Copy the integer value to the output buffer */
2227    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2228    OP2(SLJIT_ASHR, SLJIT_SAVED_REG2, 0, SLJIT_SAVED_REG2, 0, SLJIT_IMM, UCHAR_SHIFT);
2229    #endif
2230    OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG3), sizeof(int), SLJIT_SAVED_REG2, 0);
2231    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_SCRATCH_REG2, 0, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, 1);
2232    JUMPTO(SLJIT_C_NOT_ZERO, loop);
2233    JUMPHERE(early_quit);
2234    
2235    /* Calculate the return value, which is the maximum ovector value. */
2236    if (topbracket > 1)
2237      {
2238      GET_LOCAL_BASE(SLJIT_SCRATCH_REG1, 0, OVECTOR_START + topbracket * 2 * sizeof(sljit_sw));
2239      OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, topbracket + 1);
2240    
2241      /* OVECTOR(0) is never equal to SLJIT_SAVED_REG3. */
2242      loop = LABEL();
2243      OP1(SLJIT_MOVU, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG1), -(2 * (sljit_sw)sizeof(sljit_sw)));
2244      OP2(SLJIT_SUB, SLJIT_SCRATCH_REG2, 0, SLJIT_SCRATCH_REG2, 0, SLJIT_IMM, 1);
2245      CMPTO(SLJIT_C_EQUAL, SLJIT_SCRATCH_REG3, 0, SLJIT_SAVED_REG3, 0, loop);
2246      OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_SCRATCH_REG2, 0);
2247      }
2248    else
2249      OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, 1);
2250    }
2251    
2252    static SLJIT_INLINE void return_with_partial_match(compiler_common *common, struct sljit_label *quit)
2253    {
2254    DEFINE_COMPILER;
2255    struct sljit_jump *jump;
2256    
2257    SLJIT_COMPILE_ASSERT(STR_END == SLJIT_SAVED_REG2, str_end_must_be_saved_reg2);
2258    SLJIT_ASSERT(common->start_used_ptr != 0 && common->start_ptr != 0
2259      && (common->mode == JIT_PARTIAL_SOFT_COMPILE ? common->hit_start != 0 : common->hit_start == 0));
2260    
2261    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, ARGUMENTS, 0);
2262    OP1(SLJIT_MOV, SLJIT_RETURN_REG, 0, SLJIT_IMM, PCRE_ERROR_PARTIAL);
2263    OP1(SLJIT_MOV_SI, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, real_offset_count));
2264    CMPTO(SLJIT_C_SIG_LESS, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 2, quit);
2265    
2266    /* Store match begin and end. */
2267    OP1(SLJIT_MOV, SLJIT_SAVED_REG1, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, begin));
2268    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG2, 0, SLJIT_MEM1(SLJIT_SCRATCH_REG2), SLJIT_OFFSETOF(jit_arguments, offsets));
2269    
2270    jump = CMP(SLJIT_C_SIG_LESS, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, 3);
2271    OP2(SLJIT_SUB, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mode == JIT_PARTIAL_HARD_COMPILE ? common->start_ptr : (common->hit_start + (int)sizeof(sljit_sw)), SLJIT_SAVED_REG1, 0);
2272    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2273    OP2(SLJIT_ASHR, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, UCHAR_SHIFT);
2274    #endif
2275    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG2), 2 * sizeof(int), SLJIT_SCRATCH_REG3, 0);
2276    JUMPHERE(jump);
2277    
2278    OP1(SLJIT_MOV, SLJIT_SCRATCH_REG3, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->mode == JIT_PARTIAL_HARD_COMPILE ? common->start_used_ptr : common->hit_start);
2279    OP2(SLJIT_SUB, SLJIT_SAVED_REG2, 0, STR_END, 0, SLJIT_SAVED_REG1, 0);
2280    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2281    OP2(SLJIT_ASHR, SLJIT_SAVED_REG2, 0, SLJIT_SAVED_REG2, 0, SLJIT_IMM, UCHAR_SHIFT);
2282    #endif
2283    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG2), sizeof(int), SLJIT_SAVED_REG2, 0);
2284    
2285    OP2(SLJIT_SUB, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_SAVED_REG1, 0);
2286    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2287    OP2(SLJIT_ASHR, SLJIT_SCRATCH_REG3, 0, SLJIT_SCRATCH_REG3, 0, SLJIT_IMM, UCHAR_SHIFT);
2288    #endif
2289    OP1(SLJIT_MOV_SI, SLJIT_MEM1(SLJIT_SCRATCH_REG2), 0, SLJIT_SCRATCH_REG3, 0);
2290    
2291    JUMPTO(SLJIT_JUMP, quit);
2292    }
2293    
2294    static SLJIT_INLINE void check_start_used_ptr(compiler_common *common)
2295    {
2296    /* May destroy TMP1. */
2297    DEFINE_COMPILER;
2298    struct sljit_jump *jump;
2299    
2300    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2301      {
2302      /* The value of -1 must be kept for start_used_ptr! */
2303      OP2(SLJIT_ADD, TMP1, 0, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, SLJIT_IMM, 1);
2304      /* Jumps if start_used_ptr < STR_PTR, or start_used_ptr == -1. Although overwriting
2305      is not necessary if start_used_ptr == STR_PTR, it does not hurt as well. */
2306      jump = CMP(SLJIT_C_LESS_EQUAL, TMP1, 0, STR_PTR, 0);
2307      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);
2308      JUMPHERE(jump);
2309      }
2310    else if (common->mode == JIT_PARTIAL_HARD_COMPILE)
2311      {
2312      jump = CMP(SLJIT_C_LESS_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);
2313      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);
2314      JUMPHERE(jump);
2315      }
2316    }
2317    
2318    static SLJIT_INLINE BOOL char_has_othercase(compiler_common *common, pcre_uchar* cc)
2319    {
2320    /* Detects if the character has an othercase. */
2321    unsigned int c;
2322    
2323    #ifdef SUPPORT_UTF
2324    if (common->utf)
2325      {
2326      GETCHAR(c, cc);
2327      if (c > 127)
2328        {
2329    #ifdef SUPPORT_UCP
2330        return c != UCD_OTHERCASE(c);
2331    #else
2332        return FALSE;
2333    #endif
2334        }
2335    #ifndef COMPILE_PCRE8
2336      return common->fcc[c] != c;
2337    #endif
2338      }
2339    else
2340    #endif
2341      c = *cc;
2342    return MAX_255(c) ? common->fcc[c] != c : FALSE;
2343    }
2344    
2345    static SLJIT_INLINE unsigned int char_othercase(compiler_common *common, unsigned int c)
2346    {
2347    /* Returns with the othercase. */
2348    #ifdef SUPPORT_UTF
2349    if (common->utf && c > 127)
2350      {
2351    #ifdef SUPPORT_UCP
2352      return UCD_OTHERCASE(c);
2353    #else
2354      return c;
2355    #endif
2356      }
2357    #endif
2358    return TABLE_GET(c, common->fcc, c);
2359    }
2360    
2361    static unsigned int char_get_othercase_bit(compiler_common *common, pcre_uchar* cc)
2362    {
2363    /* Detects if the character and its othercase has only 1 bit difference. */
2364    unsigned int c, oc, bit;
2365    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2366    int n;
2367    #endif
2368    
2369    #ifdef SUPPORT_UTF
2370    if (common->utf)
2371      {
2372      GETCHAR(c, cc);
2373      if (c <= 127)
2374        oc = common->fcc[c];
2375      else
2376        {
2377    #ifdef SUPPORT_UCP
2378        oc = UCD_OTHERCASE(c);
2379    #else
2380        oc = c;
2381    #endif
2382        }
2383      }
2384    else
2385      {
2386      c = *cc;
2387      oc = TABLE_GET(c, common->fcc, c);
2388      }
2389    #else
2390    c = *cc;
2391    oc = TABLE_GET(c, common->fcc, c);
2392    #endif
2393    
2394    SLJIT_ASSERT(c != oc);
2395    
2396    bit = c ^ oc;
2397    /* Optimized for English alphabet. */
2398    if (c <= 127 && bit == 0x20)
2399      return (0 << 8) | 0x20;
2400    
2401    /* Since c != oc, they must have at least 1 bit difference. */
2402    if (!is_powerof2(bit))
2403      return 0;
2404    
2405    #if defined COMPILE_PCRE8
2406    
2407    #ifdef SUPPORT_UTF
2408    if (common->utf && c > 127)
2409      {
2410      n = GET_EXTRALEN(*cc);
2411      while ((bit & 0x3f) == 0)
2412        {
2413        n--;
2414        bit >>= 6;
2415        }
2416      return (n << 8) | bit;
2417      }
2418    #endif /* SUPPORT_UTF */
2419    return (0 << 8) | bit;
2420    
2421    #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
2422    
2423    #ifdef SUPPORT_UTF
2424    if (common->utf && c > 65535)
2425      {
2426      if (bit >= (1 << 10))
2427        bit >>= 10;
2428      else
2429        return (bit < 256) ? ((2 << 8) | bit) : ((3 << 8) | (bit >> 8));
2430      }
2431    #endif /* SUPPORT_UTF */
2432    return (bit < 256) ? ((0 << 8) | bit) : ((1 << 8) | (bit >> 8));
2433    
2434    #endif /* COMPILE_PCRE[8|16|32] */
2435    }
2436    
2437    static void check_partial(compiler_common *common, BOOL force)
2438    {
2439    /* Checks whether a partial matching is occurred. Does not modify registers. */
2440    DEFINE_COMPILER;
2441    struct sljit_jump *jump = NULL;
2442    
2443    SLJIT_ASSERT(!force || common->mode != JIT_COMPILE);
2444    
2445    if (common->mode == JIT_COMPILE)
2446      return;
2447    
2448    if (!force)
2449      jump = CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0);
2450    else if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2451      jump = CMP(SLJIT_C_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, SLJIT_IMM, -1);
2452    
2453    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2454      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2455    else
2456      {
2457      if (common->partialmatchlabel != NULL)
2458        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2459      else
2460        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2461      }
2462    
2463    if (jump != NULL)
2464      JUMPHERE(jump);
2465    }
2466    
2467    static void check_str_end(compiler_common *common, jump_list **end_reached)
2468    {
2469    /* Does not affect registers. Usually used in a tight spot. */
2470    DEFINE_COMPILER;
2471    struct sljit_jump *jump;
2472    
2473    if (common->mode == JIT_COMPILE)
2474      {
2475      add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2476      return;
2477      }
2478    
2479    jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);
2480    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2481      {
2482      add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));
2483      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2484      add_jump(compiler, end_reached, JUMP(SLJIT_JUMP));
2485      }
2486    else
2487      {
2488      add_jump(compiler, end_reached, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));
2489      if (common->partialmatchlabel != NULL)
2490        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2491      else
2492        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2493      }
2494    JUMPHERE(jump);
2495    }
2496    
2497    static void detect_partial_match(compiler_common *common, jump_list **backtracks)
2498    {
2499    DEFINE_COMPILER;
2500    struct sljit_jump *jump;
2501    
2502    if (common->mode == JIT_COMPILE)
2503      {
2504      add_jump(compiler, backtracks, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));
2505      return;
2506      }
2507    
2508    /* Partial matching mode. */
2509    jump = CMP(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0);
2510    add_jump(compiler, backtracks, CMP(SLJIT_C_GREATER_EQUAL, SLJIT_MEM1(SLJIT_LOCALS_REG), common->start_used_ptr, STR_PTR, 0));
2511    if (common->mode == JIT_PARTIAL_SOFT_COMPILE)
2512      {
2513      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->hit_start, SLJIT_IMM, 0);
2514      add_jump(compiler, backtracks, JUMP(SLJIT_JUMP));
2515      }
2516    else
2517      {
2518      if (common->partialmatchlabel != NULL)
2519        JUMPTO(SLJIT_JUMP, common->partialmatchlabel);
2520      else
2521        add_jump(compiler, &common->partialmatch, JUMP(SLJIT_JUMP));
2522      }
2523    JUMPHERE(jump);
2524    }
2525    
2526    static void peek_char(compiler_common *common, pcre_uint32 max)
2527    {
2528    /* Reads the character into TMP1, keeps STR_PTR.
2529    Does not check STR_END. TMP2 Destroyed. */
2530    DEFINE_COMPILER;
2531    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2532    struct sljit_jump *jump;
2533    #endif
2534    
2535    SLJIT_UNUSED_ARG(max);
2536    
2537    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
2538    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2539    if (common->utf)
2540      {
2541      if (max < 128) return;
2542    
2543      jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2544      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2545      add_jump(compiler, &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2546      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2547      JUMPHERE(jump);
2548      }
2549    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2550    
2551    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2552    if (common->utf)
2553      {
2554      if (max < 0xd800) return;
2555    
2556      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2557      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2558      /* TMP2 contains the high surrogate. */
2559      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2560      OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2561      OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2562      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2563      OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2564      JUMPHERE(jump);
2565      }
2566    #endif
2567    }
2568    
2569    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2570    
2571    static BOOL is_char7_bitset(const pcre_uint8 *bitset, BOOL nclass)
2572    {
2573    /* Tells whether the character codes below 128 are enough
2574    to determine a match. */
2575    const pcre_uint8 value = nclass ? 0xff : 0;
2576    const pcre_uint8* end = bitset + 32;
2577    
2578    bitset += 16;
2579    do
2580      {
2581      if (*bitset++ != value)
2582        return FALSE;
2583      }
2584    while (bitset < end);
2585    return TRUE;
2586    }
2587    
2588    static void read_char7_type(compiler_common *common, BOOL full_read)
2589    {
2590    /* Reads the precise character type of a character into TMP1, if the character
2591    is less than 128. Otherwise it returns with zero. Does not check STR_END. The
2592    full_read argument tells whether characters above max are accepted or not. */
2593    DEFINE_COMPILER;
2594    struct sljit_jump *jump;
2595    
2596    SLJIT_ASSERT(common->utf);
2597    
2598    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2599    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2600    
2601    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2602    
2603    if (full_read)
2604      {
2605      jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2606      OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2607      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2608      JUMPHERE(jump);
2609      }
2610    }
2611    
2612    #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2613    
2614    static void read_char_range(compiler_common *common, pcre_uint32 min, pcre_uint32 max, BOOL update_str_ptr)
2615    {
2616    /* Reads the precise value of a character into TMP1, if the character is
2617    between min and max (c >= min && c <= max). Otherwise it returns with a value
2618    outside the range. Does not check STR_END. */
2619    DEFINE_COMPILER;
2620    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2621    struct sljit_jump *jump;
2622    #endif
2623    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2624    struct sljit_jump *jump2;
2625    #endif
2626    
2627    SLJIT_UNUSED_ARG(update_str_ptr);
2628    SLJIT_UNUSED_ARG(min);
2629    SLJIT_UNUSED_ARG(max);
2630    SLJIT_ASSERT(min <= max);
2631    
2632    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2633    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2634    
2635    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2636    if (common->utf)
2637      {
2638      if (max < 128 && !update_str_ptr) return;
2639    
2640      jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
2641      if (min >= 0x10000)
2642        {
2643        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xf0);
2644        if (update_str_ptr)
2645          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2646        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2647        jump2 = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0x7);
2648        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2649        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2650        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2651        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2652        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2653        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2654        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2655        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2656        if (!update_str_ptr)
2657          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2658        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2659        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2660        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2661        JUMPHERE(jump2);
2662        if (update_str_ptr)
2663          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2664        }
2665      else if (min >= 0x800 && max <= 0xffff)
2666        {
2667        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xe0);
2668        if (update_str_ptr)
2669          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2670        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2671        jump2 = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xf);
2672        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2673        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2674        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2675        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2676        if (!update_str_ptr)
2677          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2678        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2679        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2680        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2681        JUMPHERE(jump2);
2682        if (update_str_ptr)
2683          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2684        }
2685      else if (max >= 0x800)
2686        add_jump(compiler, (max < 0x10000) ? &common->utfreadchar16 : &common->utfreadchar, JUMP(SLJIT_FAST_CALL));
2687      else if (max < 128)
2688        {
2689        OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2690        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2691        }
2692      else
2693        {
2694        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2695        if (!update_str_ptr)
2696          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2697        else
2698          OP1(SLJIT_MOV_UB, RETURN_ADDR, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2699        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2700        OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2701        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2702        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2703        if (update_str_ptr)
2704          OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, RETURN_ADDR, 0);
2705        }
2706      JUMPHERE(jump);
2707      }
2708    #endif
2709    
2710    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2711    if (common->utf)
2712      {
2713      if (max >= 0x10000)
2714        {
2715        OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2716        jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2717        /* TMP2 contains the high surrogate. */
2718        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2719        OP2(SLJIT_ADD, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x40);
2720        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 10);
2721        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2722        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3ff);
2723        OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2724        JUMPHERE(jump);
2725        return;
2726        }
2727    
2728      if (max < 0xd800 && !update_str_ptr) return;
2729    
2730      /* Skip low surrogate if necessary. */
2731      OP2(SLJIT_SUB, TMP2, 0, TMP1, 0, SLJIT_IMM, 0xd800);
2732      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2733      if (update_str_ptr)
2734        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2735      if (max >= 0xd800)
2736        OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0x10000);
2737      JUMPHERE(jump);
2738      }
2739    #endif
2740    }
2741    
2742    static SLJIT_INLINE void read_char(compiler_common *common)
2743    {
2744    read_char_range(common, 0, READ_CHAR_MAX, TRUE);
2745    }
2746    
2747    static void read_char8_type(compiler_common *common, BOOL update_str_ptr)
2748    {
2749    /* Reads the character type into TMP1, updates STR_PTR. Does not check STR_END. */
2750    DEFINE_COMPILER;
2751    #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2752    struct sljit_jump *jump;
2753    #endif
2754    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2755    struct sljit_jump *jump2;
2756    #endif
2757    
2758    SLJIT_UNUSED_ARG(update_str_ptr);
2759    
2760    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), 0);
2761    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2762    
2763    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
2764    if (common->utf)
2765      {
2766      /* This can be an extra read in some situations, but hopefully
2767      it is needed in most cases. */
2768      OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2769      jump = CMP(SLJIT_C_LESS, TMP2, 0, SLJIT_IMM, 0xc0);
2770      if (!update_str_ptr)
2771        {
2772        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2773        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2774        OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2775        OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2776        OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2777        OP2(SLJIT_OR, TMP2, 0, TMP2, 0, TMP1, 0);
2778        OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2779        jump2 = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);
2780        OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2781        JUMPHERE(jump2);
2782        }
2783      else
2784        add_jump(compiler, &common->utfreadtype8, JUMP(SLJIT_FAST_CALL));
2785      JUMPHERE(jump);
2786      return;
2787      }
2788    #endif /* SUPPORT_UTF && COMPILE_PCRE8 */
2789    
2790    #if !defined COMPILE_PCRE8
2791    /* The ctypes array contains only 256 values. */
2792    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2793    jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 255);
2794    #endif
2795    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2796    #if !defined COMPILE_PCRE8
2797    JUMPHERE(jump);
2798    #endif
2799    
2800    #if defined SUPPORT_UTF && defined COMPILE_PCRE16
2801    if (common->utf && update_str_ptr)
2802      {
2803      /* Skip low surrogate if necessary. */
2804      OP2(SLJIT_SUB, TMP2, 0, TMP2, 0, SLJIT_IMM, 0xd800);
2805      jump = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0xdc00 - 0xd800 - 1);
2806      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2807      JUMPHERE(jump);
2808      }
2809    #endif /* SUPPORT_UTF && COMPILE_PCRE16 */
2810    }
2811    
2812    static void skip_char_back(compiler_common *common)
2813    {
2814    /* Goes one character back. Affects STR_PTR and TMP1. Does not check begin. */
2815    DEFINE_COMPILER;
2816    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2817    #if defined COMPILE_PCRE8
2818    struct sljit_label *label;
2819    
2820    if (common->utf)
2821      {
2822      label = LABEL();
2823      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
2824      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2825      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xc0);
2826      CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, 0x80, label);
2827      return;
2828      }
2829    #elif defined COMPILE_PCRE16
2830    if (common->utf)
2831      {
2832      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), -IN_UCHARS(1));
2833      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2834      /* Skip low surrogate if necessary. */
2835      OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
2836      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xdc00);
2837      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
2838      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
2839      OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
2840      return;
2841      }
2842    #endif /* COMPILE_PCRE[8|16] */
2843    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
2844    OP2(SLJIT_SUB, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2845    }
2846    
2847    static void check_newlinechar(compiler_common *common, int nltype, jump_list **backtracks, BOOL jumpifmatch)
2848    {
2849    /* Character comes in TMP1. Checks if it is a newline. TMP2 may be destroyed. */
2850    DEFINE_COMPILER;
2851    struct sljit_jump *jump;
2852    
2853    if (nltype == NLTYPE_ANY)
2854      {
2855      add_jump(compiler, &common->anynewline, JUMP(SLJIT_FAST_CALL));
2856      add_jump(compiler, backtracks, JUMP(jumpifmatch ? SLJIT_C_NOT_ZERO : SLJIT_C_ZERO));
2857      }
2858    else if (nltype == NLTYPE_ANYCRLF)
2859      {
2860      if (jumpifmatch)
2861        {
2862        add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR));
2863        add_jump(compiler, backtracks, CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL));
2864        }
2865      else
2866        {
2867        jump = CMP(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_CR);
2868        add_jump(compiler, backtracks, CMP(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, CHAR_NL));
2869        JUMPHERE(jump);
2870        }
2871      }
2872    else
2873      {
2874      SLJIT_ASSERT(nltype == NLTYPE_FIXED && common->newline < 256);
2875      add_jump(compiler, backtracks, CMP(jumpifmatch ? SLJIT_C_EQUAL : SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, common->newline));
2876      }
2877    }
2878    
2879    #ifdef SUPPORT_UTF
2880    
2881    #if defined COMPILE_PCRE8
2882    static void do_utfreadchar(compiler_common *common)
2883    {
2884    /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2885    of the character (>= 0xc0). Return char value in TMP1, length in TMP2. */
2886    DEFINE_COMPILER;
2887    struct sljit_jump *jump;
2888    
2889    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2890    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2891    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2892    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2893    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2894    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2895    
2896    /* Searching for the first zero. */
2897    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
2898    jump = JUMP(SLJIT_C_NOT_ZERO);
2899    /* Two byte sequence. */
2900    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2901    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(2));
2902    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2903    
2904    JUMPHERE(jump);
2905    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2906    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
2907    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2908    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2909    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2910    
2911    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2912    jump = JUMP(SLJIT_C_NOT_ZERO);
2913    /* Three byte sequence. */
2914    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2915    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(3));
2916    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2917    
2918    /* Four byte sequence. */
2919    JUMPHERE(jump);
2920    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(2));
2921    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x10000);
2922    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2923    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(3));
2924    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2925    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2926    OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, IN_UCHARS(4));
2927    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2928    }
2929    
2930    static void do_utfreadchar16(compiler_common *common)
2931    {
2932    /* Fast decoding a UTF-8 character. TMP1 contains the first byte
2933    of the character (>= 0xc0). Return value in TMP1. */
2934    DEFINE_COMPILER;
2935    struct sljit_jump *jump;
2936    
2937    sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
2938    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2939    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2940    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2941    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2942    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2943    
2944    /* Searching for the first zero. */
2945    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x800);
2946    jump = JUMP(SLJIT_C_NOT_ZERO);
2947    /* Two byte sequence. */
2948    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2949    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2950    
2951    JUMPHERE(jump);
2952    OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0x400);
2953    OP_FLAGS(SLJIT_MOV, TMP2, 0, SLJIT_UNUSED, 0, SLJIT_C_NOT_ZERO);
2954    /* This code runs only in 8 bit mode. No need to shift the value. */
2955    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2956    OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(1));
2957    OP2(SLJIT_XOR, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x800);
2958    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 6);
2959    OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x3f);
2960    OP2(SLJIT_OR, TMP1, 0, TMP1, 0, TMP2, 0);
2961    /* Three byte sequence. */
2962    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(2));
2963    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2964  }  }
2965    
2966  static SLJIT_INLINE BOOL ispowerof2(unsigned int value)  static void do_utfreadtype8(compiler_common *common)
2967  {  {
2968  return (value & (value - 1)) == 0;  /* Fast decoding a UTF-8 character type. TMP2 contains the first byte
2969  }  of the character (>= 0xc0). Return value in TMP1. */
2970    DEFINE_COMPILER;
2971    struct sljit_jump *jump;
2972    struct sljit_jump *compare;
2973    
2974  static SLJIT_INLINE void set_jumps(jump_list *list, struct sljit_label *label)  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
 {  
 while (list)  
   {  
   /* sljit_set_label is clever enough to do nothing  
   if either the jump or the label is NULL */  
   sljit_set_label(list->jump, label);  
   list = list->next;  
   }  
 }  
2975    
2976  static SLJIT_INLINE void add_jump(struct sljit_compiler *compiler, jump_list **list, struct sljit_jump* jump)  OP2(SLJIT_AND | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP2, 0, SLJIT_IMM, 0x20);
2977  {  jump = JUMP(SLJIT_C_NOT_ZERO);
2978  jump_list *list_item = sljit_alloc_memory(compiler, sizeof(jump_list));  /* Two byte sequence. */
2979  if (list_item)  OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
2980    {  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
2981    list_item->next = *list;  OP2(SLJIT_AND, TMP2, 0, TMP2, 0, SLJIT_IMM, 0x1f);
2982    list_item->jump = jump;  /* The upper 5 bits are known at this point. */
2983    *list = list_item;  compare = CMP(SLJIT_C_GREATER, TMP2, 0, SLJIT_IMM, 0x3);
2984    }  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, 6);
2985    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0x3f);
2986    OP2(SLJIT_OR, TMP2, 0, TMP2, 0, TMP1, 0);
2987    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP2), common->ctypes);
2988    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2989    
2990    JUMPHERE(compare);
2991    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2992    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
2993    
2994    /* We only have types for characters less than 256. */
2995    JUMPHERE(jump);
2996    OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(utf8_table4) - 0xc0);
2997    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 0);
2998    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP2, 0);
2999    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3000  }  }
3001    
3002  static void add_stub(compiler_common *common, enum stub_types type, int data, struct sljit_jump *start)  #endif /* COMPILE_PCRE8 */
3003    
3004    #endif /* SUPPORT_UTF */
3005    
3006    #ifdef SUPPORT_UCP
3007    
3008    /* UCD_BLOCK_SIZE must be 128 (see the assert below). */
3009    #define UCD_BLOCK_MASK 127
3010    #define UCD_BLOCK_SHIFT 7
3011    
3012    static void do_getucd(compiler_common *common)
3013  {  {
3014    /* Search the UCD record for the character comes in TMP1.
3015    Returns chartype in TMP1 and UCD offset in TMP2. */
3016  DEFINE_COMPILER;  DEFINE_COMPILER;
 stub_list* list_item = sljit_alloc_memory(compiler, sizeof(stub_list));  
3017    
3018  if (list_item)  SLJIT_ASSERT(UCD_BLOCK_SIZE == 128 && sizeof(ucd_record) == 8);
3019    {  
3020    list_item->type = type;  sljit_emit_fast_enter(compiler, RETURN_ADDR, 0);
3021    list_item->data = data;  OP2(SLJIT_LSHR, TMP2, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
3022    list_item->start = start;  OP1(SLJIT_MOV_UB, TMP2, 0, SLJIT_MEM1(TMP2), (sljit_sw)PRIV(ucd_stage1));
3023    list_item->leave = LABEL();  OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, UCD_BLOCK_MASK);
3024    list_item->next = common->stubs;  OP2(SLJIT_SHL, TMP2, 0, TMP2, 0, SLJIT_IMM, UCD_BLOCK_SHIFT);
3025    common->stubs = list_item;  OP2(SLJIT_ADD, TMP1, 0, TMP1, 0, TMP2, 0);
3026    }  OP1(SLJIT_MOV, TMP2, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_stage2));
3027    OP1(SLJIT_MOV_UH, TMP2, 0, SLJIT_MEM2(TMP2, TMP1), 1);
3028    OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, (sljit_sw)PRIV(ucd_records) + SLJIT_OFFSETOF(ucd_record, chartype));
3029    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM2(TMP1, TMP2), 3);
3030    sljit_emit_fast_return(compiler, RETURN_ADDR, 0);
3031  }  }
3032    #endif
3033    
3034  static void flush_stubs(compiler_common *common)  static SLJIT_INLINE struct sljit_label *mainloop_entry(compiler_common *common, BOOL hascrorlf, BOOL firstline)
3035  {  {
3036  DEFINE_COMPILER;  DEFINE_COMPILER;
3037  stub_list* list_item = common->stubs;  struct sljit_label *mainloop;
3038    struct sljit_label *newlinelabel = NULL;
3039    struct sljit_jump *start;
3040    struct sljit_jump *end = NULL;
3041    struct sljit_jump *nl = NULL;
3042    #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3043    struct sljit_jump *singlechar;
3044    #endif
3045    jump_list *newline = NULL;
3046    BOOL newlinecheck = FALSE;
3047    BOOL readuchar = FALSE;
3048    
3049  while (list_item)  if (!(hascrorlf || firstline) && (common->nltype == NLTYPE_ANY ||
3050        common->nltype == NLTYPE_ANYCRLF || common->newline > 255))
3051      newlinecheck = TRUE;
3052    
3053    if (firstline)
3054    {    {
3055    JUMPHERE(list_item->start);    /* Search for the end of the first line. */
3056    switch(list_item->type)    SLJIT_ASSERT(common->first_line_end != 0);
3057      {    OP1(SLJIT_MOV, TMP3, 0, STR_PTR, 0);
     case stack_alloc:  
     add_jump(compiler, &common->stackalloc, JUMP(SLJIT_FAST_CALL));  
     break;  
3058    
3059      case max_index:    if (common->nltype == NLTYPE_FIXED && common->newline > 255)
3060      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), MAX_INDEX, SLJIT_IMM, list_item->data);      {
3061      break;      mainloop = LABEL();
3062        OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3063        end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3064        OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(-1));
3065        OP1(MOV_UCHAR, TMP2, 0, SLJIT_MEM1(STR_PTR), IN_UCHARS(0));
3066        CMPTO(SLJIT_C_NOT_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, mainloop);
3067        CMPTO(SLJIT_C_NOT_EQUAL, TMP2, 0, SLJIT_IMM, common->newline & 0xff, mainloop);
3068        JUMPHERE(end);
3069        OP2(SLJIT_SUB, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3070      }      }
3071    JUMPTO(SLJIT_JUMP, list_item->leave);    else
3072    list_item = list_item->next;      {
3073        end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3074        mainloop = LABEL();
3075        /* Continual stores does not cause data dependency. */
3076        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);
3077        read_char_range(common, common->nlmin, common->nlmax, TRUE);
3078        check_newlinechar(common, common->nltype, &newline, TRUE);
3079        CMPTO(SLJIT_C_LESS, STR_PTR, 0, STR_END, 0, mainloop);
3080        JUMPHERE(end);
3081        OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), common->first_line_end, STR_PTR, 0);
3082        set_jumps(newline, LABEL());
3083        }
3084    
3085      OP1(SLJIT_MOV, STR_PTR, 0, TMP3, 0);
3086    }    }
 common->stubs = NULL;  
 }  
3087    
3088  static SLJIT_INLINE void decrease_call_count(compiler_common *common)  start = JUMP(SLJIT_JUMP);
 {  
 DEFINE_COMPILER;  
3089    
3090  OP2(SLJIT_SUB | SLJIT_SET_E, CALL_COUNT, 0, CALL_COUNT, 0, SLJIT_IMM, 1);  if (newlinecheck)
3091  add_jump(compiler, &common->calllimit, JUMP(SLJIT_C_ZERO));    {
3092  }    newlinelabel = LABEL();
3093      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3094      end = CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0);
3095      OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
3096      OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, common->newline & 0xff);
3097      OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
3098    #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
3099      OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, UCHAR_SHIFT);
3100    #endif
3101      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3102      nl = JUMP(SLJIT_JUMP);
3103      }
3104    
3105  static SLJIT_INLINE void allocate_stack(compiler_common *common, int size)  mainloop = LABEL();
 {  
 /* May destroy all locals and registers except TMP2. */  
 DEFINE_COMPILER;  
3106    
3107  OP2(SLJIT_ADD, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_w));  /* Increasing the STR_PTR here requires one less jump in the most common case. */
3108  #ifdef DESTROY_REGISTERS  #ifdef SUPPORT_UTF
3109  OP1(SLJIT_MOV, TMP1, 0, SLJIT_IMM, 12345);  if (common->utf) readuchar = TRUE;
 OP1(SLJIT_MOV, TMP3, 0, TMP1, 0);  
 OP1(SLJIT_MOV, RETURN_ADDR, 0, TMP1, 0);  
 OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS0, TMP1, 0);  
 OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), LOCALS1, TMP1, 0);  
3110  #endif  #endif
3111  add_stub(common, stack_alloc, 0, CMP(SLJIT_C_GREATER, STACK_TOP, 0, STACK_LIMIT, 0));  if (newlinecheck) readuchar = TRUE;
 }  
3112    
3113  static SLJIT_INLINE void free_stack(compiler_common *common, int size)  if (readuchar)
3114  {    OP1(MOV_UCHAR, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);
 DEFINE_COMPILER;  
 OP2(SLJIT_SUB, STACK_TOP, 0, STACK_TOP, 0, SLJIT_IMM, size * sizeof(sljit_w));  
 }  
3115    
3116  static SLJIT_INLINE void reset_ovector(compiler_common *common, int length)  if (newlinecheck)
3117  {    CMPTO(SLJIT_C_EQUAL, TMP1, 0, SLJIT_IMM, (common->newline >> 8) & 0xff, newlinelabel);
3118  DEFINE_COMPILER;  
3119  struct sljit_label *loop;  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, IN_UCHARS(1));
3120  int i;  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3121  /* At this point we can freely use all temporary registers. */  #if defined COMPILE_PCRE8
3122  /* TMP1 returns with begin - 1. */  if (common->utf)
 OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), MAX_INDEX, SLJIT_IMM, 1);  
 OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG1, 0, SLJIT_MEM1(SLJIT_GENERAL_REG1), SLJIT_OFFSETOF(jit_arguments, begin), SLJIT_IMM, 1);  
 if (length < 8)  
3123    {    {
3124    for (i = 0; i < length; i++)    singlechar = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xc0);
3125      OP1(SLJIT_MOV, SLJIT_MEM1(SLJIT_LOCALS_REG), OVECTOR(i), SLJIT_TEMPORARY_REG1, 0);    OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(TMP1), (sljit_sw)PRIV(utf8_table4) - 0xc0);
3126      OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3127      JUMPHERE(singlechar);
3128    }    }
3129  else  #elif defined COMPILE_PCRE16
3130    if (common->utf)
3131    {    {
3132    OP2(SLJIT_ADD, SLJIT_TEMPORARY_REG2, 0, SLJIT_LOCALS_REG, 0, SLJIT_IMM, OVECTOR_START - sizeof(sljit_w));    singlechar = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 0xd800);
3133    OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, length);    OP2(SLJIT_AND, TMP1, 0, TMP1, 0, SLJIT_IMM, 0xfc00);
3134    loop = LABEL();    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_UNUSED, 0, TMP1, 0, SLJIT_IMM, 0xd800);
3135    OP1(SLJIT_MOVU, SLJIT_MEM1(SLJIT_TEMPORARY_REG2), sizeof(sljit_w), SLJIT_TEMPORARY_REG1, 0);    OP_FLAGS(SLJIT_MOV, TMP1, 0, SLJIT_UNUSED, 0, SLJIT_C_EQUAL);
3136    OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_TEMPORARY_REG3, 0, SLJIT_TEMPORARY_REG3, 0, SLJIT_IMM, 1);    OP2(SLJIT_SHL, TMP1, 0, TMP1, 0, SLJIT_IMM, 1);
3137    JUMPTO(SLJIT_C_NOT_ZERO, loop);    OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, TMP1, 0);
3138      JUMPHERE(singlechar);
3139    }    }
3140  }  #endif /* COMPILE_PCRE[8|16] */
3141    #endif /* SUPPORT_UTF && !COMPILE_PCRE32 */
3142    JUMPHERE(start);
3143    
3144  static SLJIT_INLINE void copy_ovector(compiler_common *common)  if (newlinecheck)
3145  {    {
3146  DEFINE_COMPILER;    JUMPHERE(end);
3147  struct sljit_label *loop;    JUMPHERE(nl);
3148  struct sljit_jump *earlyexit;    }
3149    
3150  /* At this point we can freely use all registers. */  return mainloop;
 OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, ARGUMENTS, 0);  
 OP1(SLJIT_MOV_SI, SLJIT_TEMPORARY_REG2, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, offsetcount));  
 OP2(SLJIT_SUB, SLJIT_TEMPORARY_REG3, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, offsets), SLJIT_IMM, sizeof(int));  
 OP1(SLJIT_MOV, SLJIT_TEMPORARY_REG1, 0, SLJIT_MEM1(SLJIT_TEMPORARY_REG1), SLJIT_OFFSETOF(jit_arguments, begin));  
 OP2(SLJIT_ADD, SLJIT_GENERAL_REG1, 0, SLJIT_LOCALS_REG, 0, SLJIT_IMM, OVECTOR_START);  
 /* Unlikely, but possible */  
 earlyexit = CMP(SLJIT_C_EQUAL, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, 0);  
 loop = LABEL();  
 OP2(SLJIT_SUB, SLJIT_GENERAL_REG2, 0, SLJIT_MEM1(SLJIT_GENERAL_REG1), 0, SLJIT_TEMPORARY_REG1, 0);  
 OP2(SLJIT_ADD, SLJIT_GENERAL_REG1, 0, SLJIT_GENERAL_REG1, 0, SLJIT_IMM, sizeof(sljit_w));  
 /* Copy the integer value to the output buffer */  
 OP1(SLJIT_MOVU_SI, SLJIT_MEM1(SLJIT_TEMPORARY_REG3), sizeof(int), SLJIT_GENERAL_REG2, 0);  
 OP2(SLJIT_SUB | SLJIT_SET_E, SLJIT_TEMPORARY_REG2, 0, SLJIT_TEMPORARY_REG2, 0, SLJIT_IMM, 1);  
 JUMPTO(SLJIT_C_NOT_ZERO, loop);  
 JUMPHERE(earlyexit);  
3151  }  }
3152    
3153  static SLJIT_INLINE BOOL char_has_othercase(compiler_common *common, uschar* cc)  #define MAX_N_CHARS 16
3154    #define MAX_N_BYTES 8
3155    
3156    static SLJIT_INLINE void add_prefix_byte(pcre_uint8 byte, pcre_uint8 *bytes)
3157  {  {
3158  /* Detects if the character has an othercase. */  pcre_uint8 len = bytes[0];
3159  unsigned int c;  int i;
3160    
3161    if (len == 255)
3162      return;
3163    
3164  #ifdef SUPPORT_UTF8  if (len == 0)
 if (common->utf8)  
3165    {    {
3166    GETCHAR(c, cc);    bytes[0] = 1;
3167    if (c > 127)    bytes[1] = byte;
3168      {    return;
 #ifdef SUPPORT_UCP  
     return c != UCD_OTHERCASE(c);  
 #else  
     return FALSE;  
 #endif  
     }  
3169    }    }
 else  
 #endif  
   c = *cc;  
 return common->fcc[c] != c;  
 }  
3170    
3171  static SLJIT_INLINE unsigned int char_othercase(compiler_common *common, unsigned int c)  for (i = len; i > 0; i--)
3172  {    if (bytes[i] == byte)
3173  /* Returns with the othercase. */      return;
3174  #ifdef SUPPORT_UTF8  
3175  if (common->utf8 && c > 127)  if (len >= MAX_N_BYTES - 1)
3176    {    {
3177  #ifdef SUPPORT_UCP    bytes[0] = 255;
3178    return UCD_OTHERCASE(c);    return;
 #else  
   return c;  
 #endif  
3179    }    }
3180    
3181    len++;
3182    bytes[len] = byte;
3183    bytes[0] = len;
3184    }
3185    
3186    static int scan_prefix(compiler_common *common, pcre_uchar *cc, pcre_uint32 *chars, pcre_uint8 *bytes, int max_chars)
3187    {
3188    /* Recursive function, which scans prefix literals. */
3189    BOOL last, any, caseless;
3190    int len, repeat, len_save, consumed = 0;
3191    pcre_uint32 chr, mask;
3192    pcre_uchar *alternative, *cc_save, *oc;
3193    #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3194    pcre_uchar othercase[8];
3195    #elif defined SUPPORT_UTF && defined COMPILE_PCRE16
3196    pcre_uchar othercase[2];
3197    #else
3198    pcre_uchar othercase[1];
3199  #endif  #endif
 return common->fcc[c];  
 }  
3200    
3201  static unsigned int char_get_othercase_bit(compiler_common *common, uschar* cc)  repeat = 1;
3202  {  while (TRUE)
3203  /* Detects if the character and its othercase has only 1 bit difference. */    {
3204  unsigned int c, oc, bit;    last = TRUE;
3205  #ifdef SUPPORT_UTF8    any = FALSE;
3206  int n;    caseless = FALSE;
3207      switch (*cc)
3208        {
3209        case OP_CHARI:
3210        caseless = TRUE;
3211        case OP_CHAR:
3212        last = FALSE;
3213        cc++;
3214        break;
3215    
3216        case OP_SOD:
3217        case OP_SOM:
3218        case OP_SET_SOM:
3219        case OP_NOT_WORD_BOUNDARY:
3220        case OP_WORD_BOUNDARY:
3221        case OP_EODN:
3222        case OP_EOD:
3223        case OP_CIRC:
3224        case OP_CIRCM:
3225        case OP_DOLL:
3226        case OP_DOLLM:
3227        /* Zero width assertions. */
3228        cc++;
3229        continue;
3230    
3231        case OP_ASSERT:
3232        case OP_ASSERT_NOT:
3233        case OP_ASSERTBACK:
3234        case OP_ASSERTBACK_NOT:
3235        cc = bracketend(cc);
3236        continue;
3237    
3238        case OP_PLUSI:
3239        case OP_MINPLUSI:
3240        case OP_POSPLUSI:
3241        caseless = TRUE;
3242        case OP_PLUS:
3243        case OP_MINPLUS:
3244        case OP_POSPLUS:
3245        cc++;
3246        break;
3247    
3248        case OP_EXACTI:
3249        caseless = TRUE;
3250        case OP_EXACT:
3251        repeat = GET2(cc, 1);
3252        last = FALSE;
3253        cc += 1 + IMM2_SIZE;
3254        break;
3255    
3256        case OP_QUERYI:
3257        case OP_MINQUERYI:
3258        case OP_POSQUERYI:
3259        caseless = TRUE;
3260        case OP_QUERY:
3261        case OP_MINQUERY:
3262        case OP_POSQUERY:
3263        len = 1;
3264        cc++;
3265    #ifdef SUPPORT_UTF
3266        if (common->utf && HAS_EXTRALEN(*cc)) len += GET_EXTRALEN(*cc);
3267  #endif  #endif
3268        max_chars = scan_prefix(common, cc + len, chars, bytes, max_chars);
3269        if (max_chars == 0)
3270          return consumed;
3271        last = FALSE;
3272        break;
3273    
3274  #ifdef SUPPORT_UTF8      case OP_KET:
3275  if (common->utf8)      cc += 1 + LINK_SIZE;
3276    {      continue;
   GETCHAR(c, cc);  
   if (c <= 127)  
     oc = common->fcc[c];  
   else  
     {  
 #ifdef SUPPORT_UCP  
     oc = UCD_OTHERCASE(c);  
 #else  
     oc = c;  
 #endif  
     }  
   }  
 else  
   {  
   c = *cc;  
   oc = common->fcc[c];  
   }  
 #else  
 c = *cc;  
 oc = common->fcc[c];  
 #endif  
3277    
3278  SLJIT_ASSERT(c != oc);      case OP_ALT:
3279        cc += GET(cc, 1);
3280        continue;
3281    
3282  bit = c ^ oc;      case OP_ONCE:
3283  /* Optimized for English alphabet. */      case OP_ONCE_NC:
3284  if (c <= 127 && bit == 0x20)      case OP_BRA:
3285    return (0 << 8) | 0x20;      case OP_BRAPOS:
3286        case OP_CBRA:
3287        case OP_CBRAPOS:
3288        alternative = cc + GET(cc, 1);
3289        while (*alternative == OP_ALT)
3290          {
3291          max_chars = scan_prefix(common, alternative + 1 + LINK_SIZE, chars, bytes, max_chars);
3292          if (max_chars == 0)
3293            return consumed;
3294          alternative += GET(alternative, 1);
3295          }
3296    
3297  /* Since c != oc, they must have at least 1 bit difference. */      if (*cc == OP_CBRA || *cc == OP_CBRAPOS)
3298  if (!ispowerof2(bit))        cc += IMM2_SIZE;
3299    return 0;      cc += 1 + LINK_SIZE;
3300        continue;
3301    
3302  #ifdef SUPPORT_UTF8      case OP_CLASS:
3303  if (common->utf8 && c > 127)  #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3304    {      if (common->utf && !is_char7_bitset((const pcre_uint8 *)(cc + 1), FALSE)) return consumed;
   n = _pcre_utf8_table4[*cc & 0x3f];  
   while ((bit & 0x3f) == 0)  
     {  
     n--;  
     bit >>= 6;  
     }  
   return (n << 8) | bit;  
   }  
3305  #endif  #endif
3306  return (0 << 8) | bit;      any = TRUE;
3307  }      cc += 1 + 32 / sizeof(pcre_uchar);
3308        break;
 static SLJIT_INLINE void check_input_end(compiler_common *common, jump_list **fallbacks)  
 {  
 DEFINE_COMPILER;  
 add_jump(compiler, fallbacks, CMP(SLJIT_C_GREATER_EQUAL, STR_PTR, 0, STR_END, 0));  
 }  
3309    
3310  static void read_char(compiler_common *common)      case OP_NCLASS:
3311  {  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3312  /* Reads the character into TMP1, updates STR_PTR.      if (common->utf) return consumed;
 Does not check STR_END. TMP2 Destroyed. */  
 DEFINE_COMPILER;  
 #ifdef SUPPORT_UTF8  
 struct sljit_jump *jump;  
3313  #endif  #endif
3314        any = TRUE;
3315        cc += 1 + 32 / sizeof(pcre_uchar);
3316        break;
3317    
3318  OP1(SLJIT_MOV_UB, TMP1, 0, SLJIT_MEM1(STR_PTR), 0);  #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3319  #ifdef SUPPORT_UTF8      case OP_XCLASS:
3320  if (common->utf8)  #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3321    {      if (common->utf) return consumed;
   /* Should not found a value between 128 and 192 here. */  
   jump = CMP(SLJIT_C_LESS, TMP1, 0, SLJIT_IMM, 192);  
   add_jump(compiler, &common->utf8readchar, JUMP(SLJIT_FAST_CALL));  
   JUMPHERE(jump);  
   }  
3322  #endif  #endif
3323  OP2(SLJIT_ADD, STR_PTR, 0, STR_PTR, 0, SLJIT_IMM, 1);      any = TRUE;
3324