]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | ******************************************************************************* | |
3 | * | |
4 | * Copyright (C) 2003-2004, International Business Machines | |
5 | * Corporation and others. All Rights Reserved. | |
6 | * | |
7 | ******************************************************************************* | |
8 | * file name: ucmstate.c | |
9 | * encoding: US-ASCII | |
10 | * tab size: 8 (not used) | |
11 | * indentation:4 | |
12 | * | |
13 | * created on: 2003oct09 | |
14 | * created by: Markus W. Scherer | |
15 | * | |
16 | * This file handles ICU .ucm file state information as part of the ucm module. | |
17 | * Most of this code used to be in makeconv.c. | |
18 | */ | |
19 | ||
20 | #include "unicode/utypes.h" | |
21 | #include "cstring.h" | |
22 | #include "cmemory.h" | |
23 | #include "uarrsort.h" | |
24 | #include "ucnvmbcs.h" | |
25 | #include "ucnv_ext.h" | |
26 | #include "uparse.h" | |
27 | #include "ucm.h" | |
28 | #include <stdio.h> | |
29 | ||
30 | /* MBCS state handling ------------------------------------------------------ */ | |
31 | ||
32 | /* | |
33 | * state table row grammar (ebnf-style): | |
34 | * (whitespace is allowed between all tokens) | |
35 | * | |
36 | * row=[[firstentry ','] entry (',' entry)*] | |
37 | * firstentry="initial" | "surrogates" | |
38 | * (initial state (default for state 0), output is all surrogate pairs) | |
39 | * entry=range [':' nextstate] ['.' action] | |
40 | * range=number ['-' number] | |
41 | * nextstate=number | |
42 | * (0..7f) | |
43 | * action='u' | 's' | 'p' | 'i' | |
44 | * (unassigned, state change only, surrogate pair, illegal) | |
45 | * number=(1- or 2-digit hexadecimal number) | |
46 | */ | |
47 | static const char * | |
48 | parseState(const char *s, int32_t state[256], uint32_t *pFlags) { | |
49 | const char *t; | |
50 | uint32_t start, end, i; | |
51 | int32_t entry; | |
52 | ||
53 | /* initialize the state: all illegal with U+ffff */ | |
54 | for(i=0; i<256; ++i) { | |
55 | state[i]=MBCS_ENTRY_FINAL(0, MBCS_STATE_ILLEGAL, 0xffff); | |
56 | } | |
57 | ||
58 | /* skip leading white space */ | |
59 | s=u_skipWhitespace(s); | |
60 | ||
61 | /* is there an "initial" or "surrogates" directive? */ | |
62 | if(uprv_strncmp("initial", s, 7)==0) { | |
63 | *pFlags=MBCS_STATE_FLAG_DIRECT; | |
64 | s=u_skipWhitespace(s+7); | |
65 | if(*s++!=',') { | |
66 | return s-1; | |
67 | } | |
68 | } else if(*pFlags==0 && uprv_strncmp("surrogates", s, 10)==0) { | |
69 | *pFlags=MBCS_STATE_FLAG_SURROGATES; | |
70 | s=u_skipWhitespace(s+10); | |
71 | if(*s++!=',') { | |
72 | return s-1; | |
73 | } | |
74 | } else if(*s==0) { | |
75 | /* empty state row: all-illegal */ | |
76 | return NULL; | |
77 | } | |
78 | ||
79 | for(;;) { | |
80 | /* read an entry, the start of the range first */ | |
81 | s=u_skipWhitespace(s); | |
82 | start=uprv_strtoul(s, (char **)&t, 16); | |
83 | if(s==t || 0xff<start) { | |
84 | return s; | |
85 | } | |
86 | s=u_skipWhitespace(t); | |
87 | ||
88 | /* read the end of the range if there is one */ | |
89 | if(*s=='-') { | |
90 | s=u_skipWhitespace(s+1); | |
91 | end=uprv_strtoul(s, (char **)&t, 16); | |
92 | if(s==t || end<start || 0xff<end) { | |
93 | return s; | |
94 | } | |
95 | s=u_skipWhitespace(t); | |
96 | } else { | |
97 | end=start; | |
98 | } | |
99 | ||
100 | /* determine the state entrys for this range */ | |
101 | if(*s!=':' && *s!='.') { | |
102 | /* the default is: final state with valid entries */ | |
103 | entry=MBCS_ENTRY_FINAL(0, MBCS_STATE_VALID_16, 0); | |
104 | } else { | |
105 | entry=MBCS_ENTRY_TRANSITION(0, 0); | |
106 | if(*s==':') { | |
107 | /* get the next state, default to 0 */ | |
108 | s=u_skipWhitespace(s+1); | |
109 | i=uprv_strtoul(s, (char **)&t, 16); | |
110 | if(s!=t) { | |
111 | if(0x7f<i) { | |
112 | return s; | |
113 | } | |
114 | s=u_skipWhitespace(t); | |
115 | entry=MBCS_ENTRY_SET_STATE(entry, i); | |
116 | } | |
117 | } | |
118 | ||
119 | /* get the state action, default to valid */ | |
120 | if(*s=='.') { | |
121 | /* this is a final state */ | |
122 | entry=MBCS_ENTRY_SET_FINAL(entry); | |
123 | ||
124 | s=u_skipWhitespace(s+1); | |
125 | if(*s=='u') { | |
126 | /* unassigned set U+fffe */ | |
127 | entry=MBCS_ENTRY_FINAL_SET_ACTION_VALUE(entry, MBCS_STATE_UNASSIGNED, 0xfffe); | |
128 | s=u_skipWhitespace(s+1); | |
129 | } else if(*s=='p') { | |
130 | if(*pFlags!=MBCS_STATE_FLAG_DIRECT) { | |
131 | entry=MBCS_ENTRY_FINAL_SET_ACTION(entry, MBCS_STATE_VALID_16_PAIR); | |
132 | } else { | |
133 | entry=MBCS_ENTRY_FINAL_SET_ACTION(entry, MBCS_STATE_VALID_16); | |
134 | } | |
135 | s=u_skipWhitespace(s+1); | |
136 | } else if(*s=='s') { | |
137 | entry=MBCS_ENTRY_FINAL_SET_ACTION(entry, MBCS_STATE_CHANGE_ONLY); | |
138 | s=u_skipWhitespace(s+1); | |
139 | } else if(*s=='i') { | |
140 | /* illegal set U+ffff */ | |
141 | entry=MBCS_ENTRY_FINAL_SET_ACTION_VALUE(entry, MBCS_STATE_ILLEGAL, 0xffff); | |
142 | s=u_skipWhitespace(s+1); | |
143 | } else { | |
144 | /* default to valid */ | |
145 | entry=MBCS_ENTRY_FINAL_SET_ACTION(entry, MBCS_STATE_VALID_16); | |
146 | } | |
147 | } else { | |
148 | /* this is an intermediate state, nothing to do */ | |
149 | } | |
150 | } | |
151 | ||
152 | /* adjust "final valid" states according to the state flags */ | |
153 | if(MBCS_ENTRY_FINAL_ACTION(entry)==MBCS_STATE_VALID_16) { | |
154 | switch(*pFlags) { | |
155 | case 0: | |
156 | /* no adjustment */ | |
157 | break; | |
158 | case MBCS_STATE_FLAG_DIRECT: | |
159 | /* set the valid-direct code point to "unassigned"==0xfffe */ | |
160 | entry=MBCS_ENTRY_FINAL_SET_ACTION_VALUE(entry, MBCS_STATE_VALID_DIRECT_16, 0xfffe); | |
161 | break; | |
162 | case MBCS_STATE_FLAG_SURROGATES: | |
163 | entry=MBCS_ENTRY_FINAL_SET_ACTION_VALUE(entry, MBCS_STATE_VALID_16_PAIR, 0); | |
164 | break; | |
165 | default: | |
166 | break; | |
167 | } | |
168 | } | |
169 | ||
170 | /* set this entry for the range */ | |
171 | for(i=start; i<=end; ++i) { | |
172 | state[i]=entry; | |
173 | } | |
174 | ||
175 | if(*s==',') { | |
176 | ++s; | |
177 | } else { | |
178 | return *s==0 ? NULL : s; | |
179 | } | |
180 | } | |
181 | } | |
182 | ||
183 | U_CAPI void U_EXPORT2 | |
184 | ucm_addState(UCMStates *states, const char *s) { | |
185 | const char *error; | |
186 | ||
187 | if(states->countStates==MBCS_MAX_STATE_COUNT) { | |
188 | fprintf(stderr, "ucm error: too many states (maximum %u)\n", MBCS_MAX_STATE_COUNT); | |
189 | exit(U_INVALID_TABLE_FORMAT); | |
190 | } | |
191 | ||
192 | error=parseState(s, states->stateTable[states->countStates], | |
193 | &states->stateFlags[states->countStates]); | |
194 | if(error!=NULL) { | |
195 | fprintf(stderr, "ucm error: parse error in state definition at '%s'\n", error); | |
196 | exit(U_INVALID_TABLE_FORMAT); | |
197 | } | |
198 | ||
199 | ++states->countStates; | |
200 | } | |
201 | ||
202 | U_CAPI UBool U_EXPORT2 | |
203 | ucm_parseHeaderLine(UCMFile *ucm, | |
204 | char *line, char **pKey, char **pValue) { | |
205 | UCMStates *states; | |
206 | char *s, *end; | |
207 | char c; | |
208 | ||
209 | states=&ucm->states; | |
210 | ||
211 | /* remove comments and trailing CR and LF and remove whitespace from the end */ | |
212 | for(end=line; (c=*end)!=0; ++end) { | |
213 | if(c=='#' || c=='\r' || c=='\n') { | |
214 | break; | |
215 | } | |
216 | } | |
217 | while(end>line && (*(end-1)==' ' || *(end-1)=='\t')) { | |
218 | --end; | |
219 | } | |
220 | *end=0; | |
221 | ||
222 | /* skip leading white space and ignore empty lines */ | |
223 | s=(char *)u_skipWhitespace(line); | |
224 | if(*s==0) { | |
225 | return TRUE; | |
226 | } | |
227 | ||
228 | /* stop at the beginning of the mapping section */ | |
229 | if(uprv_memcmp(s, "CHARMAP", 7)==0) { | |
230 | return FALSE; | |
231 | } | |
232 | ||
233 | /* get the key name, bracketed in <> */ | |
234 | if(*s!='<') { | |
235 | fprintf(stderr, "ucm error: no header field <key> in line \"%s\"\n", line); | |
236 | exit(U_INVALID_TABLE_FORMAT); | |
237 | } | |
238 | *pKey=++s; | |
239 | while(*s!='>') { | |
240 | if(*s==0) { | |
241 | fprintf(stderr, "ucm error: incomplete header field <key> in line \"%s\"\n", line); | |
242 | exit(U_INVALID_TABLE_FORMAT); | |
243 | } | |
244 | ++s; | |
245 | } | |
246 | *s=0; | |
247 | ||
248 | /* get the value string, possibly quoted */ | |
249 | s=(char *)u_skipWhitespace(s+1); | |
250 | if(*s!='"') { | |
251 | *pValue=s; | |
252 | } else { | |
253 | /* remove the quotes */ | |
254 | *pValue=s+1; | |
255 | if(end>*pValue && *(end-1)=='"') { | |
256 | *--end=0; | |
257 | } | |
258 | } | |
259 | ||
260 | /* collect the information from the header field, ignore unknown keys */ | |
261 | if(uprv_strcmp(*pKey, "uconv_class")==0) { | |
262 | if(uprv_strcmp(*pValue, "DBCS")==0) { | |
263 | states->conversionType=UCNV_DBCS; | |
264 | } else if(uprv_strcmp(*pValue, "SBCS")==0) { | |
265 | states->conversionType = UCNV_SBCS; | |
266 | } else if(uprv_strcmp(*pValue, "MBCS")==0) { | |
267 | states->conversionType = UCNV_MBCS; | |
268 | } else if(uprv_strcmp(*pValue, "EBCDIC_STATEFUL")==0) { | |
269 | states->conversionType = UCNV_EBCDIC_STATEFUL; | |
270 | } else { | |
271 | fprintf(stderr, "ucm error: unknown <uconv_class> %s\n", *pValue); | |
272 | exit(U_INVALID_TABLE_FORMAT); | |
273 | } | |
274 | return TRUE; | |
275 | } else if(uprv_strcmp(*pKey, "mb_cur_max")==0) { | |
276 | c=**pValue; | |
277 | if('1'<=c && c<='4' && (*pValue)[1]==0) { | |
278 | states->maxCharLength=(int8_t)(c-'0'); | |
279 | states->outputType=(int8_t)(states->maxCharLength-1); | |
280 | } else { | |
281 | fprintf(stderr, "ucm error: illegal <mb_cur_max> %s\n", *pValue); | |
282 | exit(U_INVALID_TABLE_FORMAT); | |
283 | } | |
284 | return TRUE; | |
285 | } else if(uprv_strcmp(*pKey, "mb_cur_min")==0) { | |
286 | c=**pValue; | |
287 | if('1'<=c && c<='4' && (*pValue)[1]==0) { | |
288 | states->minCharLength=(int8_t)(c-'0'); | |
289 | } else { | |
290 | fprintf(stderr, "ucm error: illegal <mb_cur_min> %s\n", *pValue); | |
291 | exit(U_INVALID_TABLE_FORMAT); | |
292 | } | |
293 | return TRUE; | |
294 | } else if(uprv_strcmp(*pKey, "icu:state")==0) { | |
295 | /* if an SBCS/DBCS/EBCDIC_STATEFUL converter has icu:state, then turn it into MBCS */ | |
296 | switch(states->conversionType) { | |
297 | case UCNV_SBCS: | |
298 | case UCNV_DBCS: | |
299 | case UCNV_EBCDIC_STATEFUL: | |
300 | states->conversionType=UCNV_MBCS; | |
301 | break; | |
302 | case UCNV_MBCS: | |
303 | break; | |
304 | default: | |
305 | fprintf(stderr, "ucm error: <icu:state> entry for non-MBCS table or before the <uconv_class> line\n"); | |
306 | exit(U_INVALID_TABLE_FORMAT); | |
307 | } | |
308 | ||
309 | if(states->maxCharLength==0) { | |
310 | fprintf(stderr, "ucm error: <icu:state> before the <mb_cur_max> line\n"); | |
311 | exit(U_INVALID_TABLE_FORMAT); | |
312 | } | |
313 | ucm_addState(states, *pValue); | |
314 | return TRUE; | |
315 | } else if(uprv_strcmp(*pKey, "icu:base")==0) { | |
316 | if(**pValue==0) { | |
317 | fprintf(stderr, "ucm error: <icu:base> without a base table name\n"); | |
318 | exit(U_INVALID_TABLE_FORMAT); | |
319 | } | |
320 | uprv_strcpy(ucm->baseName, *pValue); | |
321 | return TRUE; | |
322 | } | |
323 | ||
324 | return FALSE; | |
325 | } | |
326 | ||
327 | /* post-processing ---------------------------------------------------------- */ | |
328 | ||
329 | static int32_t | |
330 | sumUpStates(UCMStates *states) { | |
331 | int32_t entry, sum, state, cell, count; | |
332 | UBool allStatesReady; | |
333 | ||
334 | /* | |
335 | * Sum up the offsets for all states. | |
336 | * In each final state (where there are only final entries), | |
337 | * the offsets add up directly. | |
338 | * In all other state table rows, for each transition entry to another state, | |
339 | * the offsets sum of that state needs to be added. | |
340 | * This is achieved in at most countStates iterations. | |
341 | */ | |
342 | allStatesReady=FALSE; | |
343 | for(count=states->countStates; !allStatesReady && count>=0; --count) { | |
344 | allStatesReady=TRUE; | |
345 | for(state=states->countStates-1; state>=0; --state) { | |
346 | if(!(states->stateFlags[state]&MBCS_STATE_FLAG_READY)) { | |
347 | allStatesReady=FALSE; | |
348 | sum=0; | |
349 | ||
350 | /* at first, add up only the final delta offsets to keep them <512 */ | |
351 | for(cell=0; cell<256; ++cell) { | |
352 | entry=states->stateTable[state][cell]; | |
353 | if(MBCS_ENTRY_IS_FINAL(entry)) { | |
354 | switch(MBCS_ENTRY_FINAL_ACTION(entry)) { | |
355 | case MBCS_STATE_VALID_16: | |
356 | states->stateTable[state][cell]=MBCS_ENTRY_FINAL_SET_VALUE(entry, sum); | |
357 | sum+=1; | |
358 | break; | |
359 | case MBCS_STATE_VALID_16_PAIR: | |
360 | states->stateTable[state][cell]=MBCS_ENTRY_FINAL_SET_VALUE(entry, sum); | |
361 | sum+=2; | |
362 | break; | |
363 | default: | |
364 | /* no addition */ | |
365 | break; | |
366 | } | |
367 | } | |
368 | } | |
369 | ||
370 | /* now, add up the delta offsets for the transitional entries */ | |
371 | for(cell=0; cell<256; ++cell) { | |
372 | entry=states->stateTable[state][cell]; | |
373 | if(MBCS_ENTRY_IS_TRANSITION(entry)) { | |
374 | if(states->stateFlags[MBCS_ENTRY_TRANSITION_STATE(entry)]&MBCS_STATE_FLAG_READY) { | |
375 | states->stateTable[state][cell]=MBCS_ENTRY_TRANSITION_SET_OFFSET(entry, sum); | |
376 | sum+=states->stateOffsetSum[MBCS_ENTRY_TRANSITION_STATE(entry)]; | |
377 | } else { | |
378 | /* that next state does not have a sum yet, we cannot finish the one for this state */ | |
379 | sum=-1; | |
380 | break; | |
381 | } | |
382 | } | |
383 | } | |
384 | ||
385 | if(sum!=-1) { | |
386 | states->stateOffsetSum[state]=sum; | |
387 | states->stateFlags[state]|=MBCS_STATE_FLAG_READY; | |
388 | } | |
389 | } | |
390 | } | |
391 | } | |
392 | ||
393 | if(!allStatesReady) { | |
394 | fprintf(stderr, "ucm error: the state table contains loops\n"); | |
395 | exit(U_INVALID_TABLE_FORMAT); | |
396 | } | |
397 | ||
398 | /* | |
399 | * For all "direct" (i.e., initial) states>0, | |
400 | * the offsets need to be increased by the sum of | |
401 | * the previous initial states. | |
402 | */ | |
403 | sum=states->stateOffsetSum[0]; | |
404 | for(state=1; state<states->countStates; ++state) { | |
405 | if((states->stateFlags[state]&0xf)==MBCS_STATE_FLAG_DIRECT) { | |
406 | int32_t sum2=sum; | |
407 | sum+=states->stateOffsetSum[state]; | |
408 | for(cell=0; cell<256; ++cell) { | |
409 | entry=states->stateTable[state][cell]; | |
410 | if(MBCS_ENTRY_IS_TRANSITION(entry)) { | |
411 | states->stateTable[state][cell]=MBCS_ENTRY_TRANSITION_ADD_OFFSET(entry, sum2); | |
412 | } | |
413 | } | |
414 | } | |
415 | } | |
416 | ||
417 | /* round up to the next even number to have the following data 32-bit-aligned */ | |
418 | return states->countToUCodeUnits=(sum+1)&~1; | |
419 | } | |
420 | ||
421 | U_CAPI void U_EXPORT2 | |
422 | ucm_processStates(UCMStates *states) { | |
423 | int32_t entry, state, cell, count; | |
424 | ||
425 | if(states->conversionType==UCNV_UNSUPPORTED_CONVERTER) { | |
426 | fprintf(stderr, "ucm error: missing conversion type (<uconv_class>)\n"); | |
427 | exit(U_INVALID_TABLE_FORMAT); | |
428 | } | |
429 | ||
430 | if(states->countStates==0) { | |
431 | switch(states->conversionType) { | |
432 | case UCNV_SBCS: | |
433 | /* SBCS: use MBCS data structure with a default state table */ | |
434 | if(states->maxCharLength!=1) { | |
435 | fprintf(stderr, "error: SBCS codepage with max B/char!=1\n"); | |
436 | exit(U_INVALID_TABLE_FORMAT); | |
437 | } | |
438 | states->conversionType=UCNV_MBCS; | |
439 | ucm_addState(states, "0-ff"); | |
440 | break; | |
441 | case UCNV_MBCS: | |
442 | fprintf(stderr, "ucm error: missing state table information (<icu:state>) for MBCS\n"); | |
443 | exit(U_INVALID_TABLE_FORMAT); | |
444 | break; | |
445 | case UCNV_EBCDIC_STATEFUL: | |
446 | /* EBCDIC_STATEFUL: use MBCS data structure with a default state table */ | |
447 | if(states->minCharLength!=1 || states->maxCharLength!=2) { | |
448 | fprintf(stderr, "error: DBCS codepage with min B/char!=1 or max B/char!=2\n"); | |
449 | exit(U_INVALID_TABLE_FORMAT); | |
450 | } | |
451 | states->conversionType=UCNV_MBCS; | |
452 | ucm_addState(states, "0-ff, e:1.s, f:0.s"); | |
453 | ucm_addState(states, "initial, 0-3f:4, e:1.s, f:0.s, 40:3, 41-fe:2, ff:4"); | |
454 | ucm_addState(states, "0-40:1.i, 41-fe:1., ff:1.i"); | |
455 | ucm_addState(states, "0-ff:1.i, 40:1."); | |
456 | ucm_addState(states, "0-ff:1.i"); | |
457 | break; | |
458 | case UCNV_DBCS: | |
459 | /* DBCS: use MBCS data structure with a default state table */ | |
460 | if(states->minCharLength!=2 || states->maxCharLength!=2) { | |
461 | fprintf(stderr, "error: DBCS codepage with min or max B/char!=2\n"); | |
462 | exit(U_INVALID_TABLE_FORMAT); | |
463 | } | |
464 | states->conversionType = UCNV_MBCS; | |
465 | ucm_addState(states, "0-3f:3, 40:2, 41-fe:1, ff:3"); | |
466 | ucm_addState(states, "41-fe"); | |
467 | ucm_addState(states, "40"); | |
468 | ucm_addState(states, ""); | |
469 | break; | |
470 | default: | |
471 | fprintf(stderr, "ucm error: unknown charset structure\n"); | |
472 | exit(U_INVALID_TABLE_FORMAT); | |
473 | break; | |
474 | } | |
475 | } | |
476 | ||
477 | /* | |
478 | * check that the min/max character lengths are reasonable; | |
479 | * to do this right, all paths through the state table would have to be | |
480 | * recursively walked while keeping track of the sequence lengths, | |
481 | * but these simple checks cover most state tables in practice | |
482 | */ | |
483 | if(states->maxCharLength<states->minCharLength) { | |
484 | fprintf(stderr, "ucm error: max B/char < min B/char\n"); | |
485 | exit(U_INVALID_TABLE_FORMAT); | |
486 | } | |
487 | ||
488 | /* count non-direct states and compare with max B/char */ | |
489 | count=0; | |
490 | for(state=0; state<states->countStates; ++state) { | |
491 | if((states->stateFlags[state]&0xf)!=MBCS_STATE_FLAG_DIRECT) { | |
492 | ++count; | |
493 | } | |
494 | } | |
495 | if(states->maxCharLength>count+1) { | |
496 | fprintf(stderr, "ucm error: max B/char too large\n"); | |
497 | exit(U_INVALID_TABLE_FORMAT); | |
498 | } | |
499 | ||
500 | if(states->minCharLength==1) { | |
501 | int32_t action; | |
502 | ||
503 | /* | |
504 | * if there are single-byte characters, | |
505 | * then the initial state must have direct result states | |
506 | */ | |
507 | for(cell=0; cell<256; ++cell) { | |
508 | entry=states->stateTable[0][cell]; | |
509 | if( MBCS_ENTRY_IS_FINAL(entry) && | |
510 | ((action=MBCS_ENTRY_FINAL_ACTION(entry))==MBCS_STATE_VALID_DIRECT_16 || | |
511 | action==MBCS_STATE_UNASSIGNED) | |
512 | ) { | |
513 | break; | |
514 | } | |
515 | } | |
516 | ||
517 | if(cell==256) { | |
518 | fprintf(stderr, "ucm warning: min B/char too small\n"); | |
519 | } | |
520 | } | |
521 | ||
522 | /* | |
523 | * make sure that all "next state" values are within limits | |
524 | * and that all next states after final ones have the "direct" | |
525 | * flag of initial states | |
526 | */ | |
527 | for(state=states->countStates-1; state>=0; --state) { | |
528 | for(cell=0; cell<256; ++cell) { | |
529 | entry=states->stateTable[state][cell]; | |
530 | if((uint8_t)MBCS_ENTRY_STATE(entry)>=states->countStates) { | |
531 | fprintf(stderr, "ucm error: state table entry [%x][%x] has a next state of %x that is too high\n", | |
532 | (int)state, (int)cell, (int)MBCS_ENTRY_STATE(entry)); | |
533 | exit(U_INVALID_TABLE_FORMAT); | |
534 | } | |
535 | if(MBCS_ENTRY_IS_FINAL(entry) && (states->stateFlags[MBCS_ENTRY_STATE(entry)]&0xf)!=MBCS_STATE_FLAG_DIRECT) { | |
536 | fprintf(stderr, "ucm error: state table entry [%x][%x] is final but has a non-initial next state of %x\n", | |
537 | (int)state, (int)cell, (int)MBCS_ENTRY_STATE(entry)); | |
538 | exit(U_INVALID_TABLE_FORMAT); | |
539 | } else if(MBCS_ENTRY_IS_TRANSITION(entry) && (states->stateFlags[MBCS_ENTRY_STATE(entry)]&0xf)==MBCS_STATE_FLAG_DIRECT) { | |
540 | fprintf(stderr, "ucm error: state table entry [%x][%x] is not final but has an initial next state of %x\n", | |
541 | (int)state, (int)cell, (int)MBCS_ENTRY_STATE(entry)); | |
542 | exit(U_INVALID_TABLE_FORMAT); | |
543 | } | |
544 | } | |
545 | } | |
546 | ||
547 | /* is this an SI/SO (like EBCDIC-stateful) state table? */ | |
548 | if(states->countStates>=2 && (states->stateFlags[1]&0xf)==MBCS_STATE_FLAG_DIRECT) { | |
549 | if(states->maxCharLength!=2) { | |
550 | fprintf(stderr, "ucm error: SI/SO codepages must have max 2 bytes/char (not %x)\n", (int)states->maxCharLength); | |
551 | exit(U_INVALID_TABLE_FORMAT); | |
552 | } | |
553 | if(states->countStates<3) { | |
554 | fprintf(stderr, "ucm error: SI/SO codepages must have at least 3 states (not %x)\n", (int)states->countStates); | |
555 | exit(U_INVALID_TABLE_FORMAT); | |
556 | } | |
557 | /* are the SI/SO all in the right places? */ | |
558 | if( states->stateTable[0][0xe]==MBCS_ENTRY_FINAL(1, MBCS_STATE_CHANGE_ONLY, 0) && | |
559 | states->stateTable[0][0xf]==MBCS_ENTRY_FINAL(0, MBCS_STATE_CHANGE_ONLY, 0) && | |
560 | states->stateTable[1][0xe]==MBCS_ENTRY_FINAL(1, MBCS_STATE_CHANGE_ONLY, 0) && | |
561 | states->stateTable[1][0xf]==MBCS_ENTRY_FINAL(0, MBCS_STATE_CHANGE_ONLY, 0) | |
562 | ) { | |
563 | states->outputType=MBCS_OUTPUT_2_SISO; | |
564 | } else { | |
565 | fprintf(stderr, "ucm error: SI/SO codepages must have in states 0 and 1 transitions e:1.s, f:0.s\n"); | |
566 | exit(U_INVALID_TABLE_FORMAT); | |
567 | } | |
568 | state=2; | |
569 | } else { | |
570 | state=1; | |
571 | } | |
572 | ||
573 | /* check that no unexpected state is a "direct" one */ | |
574 | while(state<states->countStates) { | |
575 | if((states->stateFlags[state]&0xf)==MBCS_STATE_FLAG_DIRECT) { | |
576 | fprintf(stderr, "ucm error: state %d is 'initial' - not supported except for SI/SO codepages\n", (int)state); | |
577 | exit(U_INVALID_TABLE_FORMAT); | |
578 | } | |
579 | ++state; | |
580 | } | |
581 | ||
582 | sumUpStates(states); | |
583 | } | |
584 | ||
585 | /* find a fallback for this offset; return the index or -1 if not found */ | |
586 | U_CAPI int32_t U_EXPORT2 | |
587 | ucm_findFallback(_MBCSToUFallback *toUFallbacks, int32_t countToUFallbacks, | |
588 | uint32_t offset) { | |
589 | int32_t i; | |
590 | ||
591 | if(countToUFallbacks==0) { | |
592 | /* shortcut: most codepages do not have fallbacks from codepage to Unicode */ | |
593 | return -1; | |
594 | } | |
595 | ||
596 | /* do a linear search for the fallback mapping (the table is not yet sorted) */ | |
597 | for(i=0; i<countToUFallbacks; ++i) { | |
598 | if(offset==toUFallbacks[i].offset) { | |
599 | return i; | |
600 | } | |
601 | } | |
602 | return -1; | |
603 | } | |
604 | ||
605 | /* | |
606 | * This function tries to compact toUnicode tables for 2-byte codepages | |
607 | * by finding lead bytes with all-unassigned trail bytes and adding another state | |
608 | * for them. | |
609 | */ | |
610 | static void | |
611 | compactToUnicode2(UCMStates *states, | |
612 | uint16_t **pUnicodeCodeUnits, | |
613 | _MBCSToUFallback *toUFallbacks, int32_t countToUFallbacks, | |
614 | UBool verbose) { | |
615 | int32_t (*oldStateTable)[256]; | |
616 | uint16_t count[256]; | |
617 | uint16_t *oldUnicodeCodeUnits; | |
618 | int32_t entry, offset, oldOffset, trailOffset, oldTrailOffset, savings, sum; | |
619 | int32_t i, j, leadState, trailState, newState, fallback; | |
620 | uint16_t unit; | |
621 | ||
622 | /* find the lead state */ | |
623 | if(states->outputType==MBCS_OUTPUT_2_SISO) { | |
624 | /* use the DBCS lead state for SI/SO codepages */ | |
625 | leadState=1; | |
626 | } else { | |
627 | leadState=0; | |
628 | } | |
629 | ||
630 | /* find the main trail state: the most used target state */ | |
631 | uprv_memset(count, 0, sizeof(count)); | |
632 | for(i=0; i<256; ++i) { | |
633 | entry=states->stateTable[leadState][i]; | |
634 | if(MBCS_ENTRY_IS_TRANSITION(entry)) { | |
635 | ++count[MBCS_ENTRY_TRANSITION_STATE(entry)]; | |
636 | } | |
637 | } | |
638 | trailState=0; | |
639 | for(i=1; i<states->countStates; ++i) { | |
640 | if(count[i]>count[trailState]) { | |
641 | trailState=i; | |
642 | } | |
643 | } | |
644 | ||
645 | /* count possible savings from lead bytes with all-unassigned results in all trail bytes */ | |
646 | uprv_memset(count, 0, sizeof(count)); | |
647 | savings=0; | |
648 | /* for each lead byte */ | |
649 | for(i=0; i<256; ++i) { | |
650 | entry=states->stateTable[leadState][i]; | |
651 | if(MBCS_ENTRY_IS_TRANSITION(entry) && (MBCS_ENTRY_TRANSITION_STATE(entry))==trailState) { | |
652 | /* the offset is different for each lead byte */ | |
653 | offset=MBCS_ENTRY_TRANSITION_OFFSET(entry); | |
654 | /* for each trail byte for this lead byte */ | |
655 | for(j=0; j<256; ++j) { | |
656 | entry=states->stateTable[trailState][j]; | |
657 | switch(MBCS_ENTRY_FINAL_ACTION(entry)) { | |
658 | case MBCS_STATE_VALID_16: | |
659 | entry=offset+MBCS_ENTRY_FINAL_VALUE_16(entry); | |
660 | if((*pUnicodeCodeUnits)[entry]==0xfffe && ucm_findFallback(toUFallbacks, countToUFallbacks, entry)<0) { | |
661 | ++count[i]; | |
662 | } else { | |
663 | j=999; /* do not count for this lead byte because there are assignments */ | |
664 | } | |
665 | break; | |
666 | case MBCS_STATE_VALID_16_PAIR: | |
667 | entry=offset+MBCS_ENTRY_FINAL_VALUE_16(entry); | |
668 | if((*pUnicodeCodeUnits)[entry]==0xfffe) { | |
669 | count[i]+=2; | |
670 | } else { | |
671 | j=999; /* do not count for this lead byte because there are assignments */ | |
672 | } | |
673 | break; | |
674 | default: | |
675 | break; | |
676 | } | |
677 | } | |
678 | if(j==256) { | |
679 | /* all trail bytes for this lead byte are unassigned */ | |
680 | savings+=count[i]; | |
681 | } else { | |
682 | count[i]=0; | |
683 | } | |
684 | } | |
685 | } | |
686 | /* subtract from the possible savings the cost of an additional state */ | |
687 | savings=savings*2-1024; /* count bytes, not 16-bit words */ | |
688 | if(savings<=0) { | |
689 | return; | |
690 | } | |
691 | if(verbose) { | |
692 | printf("compacting toUnicode data saves %ld bytes\n", (long)savings); | |
693 | } | |
694 | if(states->countStates>=MBCS_MAX_STATE_COUNT) { | |
695 | fprintf(stderr, "cannot compact toUnicode because the maximum number of states is reached\n"); | |
696 | return; | |
697 | } | |
698 | ||
699 | /* make a copy of the state table */ | |
700 | oldStateTable=(int32_t (*)[256])uprv_malloc(states->countStates*1024); | |
701 | if(oldStateTable==NULL) { | |
702 | fprintf(stderr, "cannot compact toUnicode: out of memory\n"); | |
703 | return; | |
704 | } | |
705 | uprv_memcpy(oldStateTable, states->stateTable, states->countStates*1024); | |
706 | ||
707 | /* add the new state */ | |
708 | /* | |
709 | * this function does not catch the degenerate case where all lead bytes | |
710 | * have all-unassigned trail bytes and the lead state could be removed | |
711 | */ | |
712 | newState=states->countStates++; | |
713 | states->stateFlags[newState]=0; | |
714 | /* copy the old trail state, turning all assigned states into unassigned ones */ | |
715 | for(i=0; i<256; ++i) { | |
716 | entry=states->stateTable[trailState][i]; | |
717 | switch(MBCS_ENTRY_FINAL_ACTION(entry)) { | |
718 | case MBCS_STATE_VALID_16: | |
719 | case MBCS_STATE_VALID_16_PAIR: | |
720 | states->stateTable[newState][i]=MBCS_ENTRY_FINAL_SET_ACTION_VALUE(entry, MBCS_STATE_UNASSIGNED, 0xfffe); | |
721 | break; | |
722 | default: | |
723 | states->stateTable[newState][i]=entry; | |
724 | break; | |
725 | } | |
726 | } | |
727 | ||
728 | /* in the lead state, redirect all lead bytes with all-unassigned trail bytes to the new state */ | |
729 | for(i=0; i<256; ++i) { | |
730 | if(count[i]>0) { | |
731 | states->stateTable[leadState][i]=MBCS_ENTRY_SET_STATE(states->stateTable[leadState][i], newState); | |
732 | } | |
733 | } | |
734 | ||
735 | /* sum up the new state table */ | |
736 | for(i=0; i<states->countStates; ++i) { | |
737 | states->stateFlags[i]&=~MBCS_STATE_FLAG_READY; | |
738 | } | |
739 | sum=sumUpStates(states); | |
740 | ||
741 | /* allocate a new, smaller code units array */ | |
742 | oldUnicodeCodeUnits=*pUnicodeCodeUnits; | |
743 | if(sum==0) { | |
744 | *pUnicodeCodeUnits=NULL; | |
745 | if(oldUnicodeCodeUnits!=NULL) { | |
746 | uprv_free(oldUnicodeCodeUnits); | |
747 | } | |
748 | uprv_free(oldStateTable); | |
749 | return; | |
750 | } | |
751 | *pUnicodeCodeUnits=(uint16_t *)uprv_malloc(sum*sizeof(uint16_t)); | |
752 | if(*pUnicodeCodeUnits==NULL) { | |
753 | fprintf(stderr, "cannot compact toUnicode: out of memory allocating %ld 16-bit code units\n", | |
754 | (long)sum); | |
755 | /* revert to the old state table */ | |
756 | *pUnicodeCodeUnits=oldUnicodeCodeUnits; | |
757 | --states->countStates; | |
758 | uprv_memcpy(states->stateTable, oldStateTable, states->countStates*1024); | |
759 | uprv_free(oldStateTable); | |
760 | return; | |
761 | } | |
762 | for(i=0; i<sum; ++i) { | |
763 | (*pUnicodeCodeUnits)[i]=0xfffe; | |
764 | } | |
765 | ||
766 | /* copy the code units for all assigned characters */ | |
767 | /* | |
768 | * The old state table has the same lead _and_ trail states for assigned characters! | |
769 | * The differences are in the offsets, and in the trail states for some unassigned characters. | |
770 | * For each character with an assigned state in the new table, it was assigned in the old one. | |
771 | * Only still-assigned characters are copied. | |
772 | * Note that fallback mappings need to get their offset values adjusted. | |
773 | */ | |
774 | ||
775 | /* for each initial state */ | |
776 | for(leadState=0; leadState<states->countStates; ++leadState) { | |
777 | if((states->stateFlags[leadState]&0xf)==MBCS_STATE_FLAG_DIRECT) { | |
778 | /* for each lead byte from there */ | |
779 | for(i=0; i<256; ++i) { | |
780 | entry=states->stateTable[leadState][i]; | |
781 | if(MBCS_ENTRY_IS_TRANSITION(entry)) { | |
782 | trailState=(uint8_t)MBCS_ENTRY_TRANSITION_STATE(entry); | |
783 | /* the new state does not have assigned states */ | |
784 | if(trailState!=newState) { | |
785 | trailOffset=MBCS_ENTRY_TRANSITION_OFFSET(entry); | |
786 | oldTrailOffset=MBCS_ENTRY_TRANSITION_OFFSET(oldStateTable[leadState][i]); | |
787 | /* for each trail byte */ | |
788 | for(j=0; j<256; ++j) { | |
789 | entry=states->stateTable[trailState][j]; | |
790 | /* copy assigned-character code units and adjust fallback offsets */ | |
791 | switch(MBCS_ENTRY_FINAL_ACTION(entry)) { | |
792 | case MBCS_STATE_VALID_16: | |
793 | offset=trailOffset+MBCS_ENTRY_FINAL_VALUE_16(entry); | |
794 | /* find the old offset according to the old state table */ | |
795 | oldOffset=oldTrailOffset+MBCS_ENTRY_FINAL_VALUE_16(oldStateTable[trailState][j]); | |
796 | unit=(*pUnicodeCodeUnits)[offset]=oldUnicodeCodeUnits[oldOffset]; | |
797 | if(unit==0xfffe && (fallback=ucm_findFallback(toUFallbacks, countToUFallbacks, oldOffset))>=0) { | |
798 | toUFallbacks[fallback].offset=0x80000000|offset; | |
799 | } | |
800 | break; | |
801 | case MBCS_STATE_VALID_16_PAIR: | |
802 | offset=trailOffset+MBCS_ENTRY_FINAL_VALUE_16(entry); | |
803 | /* find the old offset according to the old state table */ | |
804 | oldOffset=oldTrailOffset+MBCS_ENTRY_FINAL_VALUE_16(oldStateTable[trailState][j]); | |
805 | (*pUnicodeCodeUnits)[offset++]=oldUnicodeCodeUnits[oldOffset++]; | |
806 | (*pUnicodeCodeUnits)[offset]=oldUnicodeCodeUnits[oldOffset]; | |
807 | break; | |
808 | default: | |
809 | break; | |
810 | } | |
811 | } | |
812 | } | |
813 | } | |
814 | } | |
815 | } | |
816 | } | |
817 | ||
818 | /* remove temporary flags from fallback offsets that protected them from being modified twice */ | |
819 | for(i=0; i<countToUFallbacks; ++i) { | |
820 | toUFallbacks[i].offset&=0x7fffffff; | |
821 | } | |
822 | ||
823 | /* free temporary memory */ | |
824 | uprv_free(oldUnicodeCodeUnits); | |
825 | uprv_free(oldStateTable); | |
826 | } | |
827 | ||
828 | /* | |
829 | * recursive sub-function of compactToUnicodeHelper() | |
830 | * returns: | |
831 | * >0 number of bytes that are used in unicodeCodeUnits[] that could be saved, | |
832 | * if all sequences from this state are unassigned, returns the | |
833 | * <0 there are assignments in unicodeCodeUnits[] | |
834 | * 0 no use of unicodeCodeUnits[] | |
835 | */ | |
836 | static int32_t | |
837 | findUnassigned(UCMStates *states, | |
838 | uint16_t *unicodeCodeUnits, | |
839 | _MBCSToUFallback *toUFallbacks, int32_t countToUFallbacks, | |
840 | int32_t state, int32_t offset, uint32_t b) { | |
841 | int32_t i, entry, savings, localSavings, belowSavings; | |
842 | UBool haveAssigned; | |
843 | ||
844 | localSavings=belowSavings=0; | |
845 | haveAssigned=FALSE; | |
846 | for(i=0; i<256; ++i) { | |
847 | entry=states->stateTable[state][i]; | |
848 | if(MBCS_ENTRY_IS_TRANSITION(entry)) { | |
849 | savings=findUnassigned(states, | |
850 | unicodeCodeUnits, | |
851 | toUFallbacks, countToUFallbacks, | |
852 | MBCS_ENTRY_TRANSITION_STATE(entry), | |
853 | offset+MBCS_ENTRY_TRANSITION_OFFSET(entry), | |
854 | (b<<8)|(uint32_t)i); | |
855 | if(savings<0) { | |
856 | haveAssigned=TRUE; | |
857 | } else if(savings>0) { | |
858 | printf(" all-unassigned sequences from prefix 0x%02lx state %ld use %ld bytes\n", | |
859 | (unsigned long)((b<<8)|i), (long)state, (long)savings); | |
860 | belowSavings+=savings; | |
861 | } | |
862 | } else if(!haveAssigned) { | |
863 | switch(MBCS_ENTRY_FINAL_ACTION(entry)) { | |
864 | case MBCS_STATE_VALID_16: | |
865 | entry=offset+MBCS_ENTRY_FINAL_VALUE_16(entry); | |
866 | if(unicodeCodeUnits[entry]==0xfffe && ucm_findFallback(toUFallbacks, countToUFallbacks, entry)<0) { | |
867 | localSavings+=2; | |
868 | } else { | |
869 | haveAssigned=TRUE; | |
870 | } | |
871 | break; | |
872 | case MBCS_STATE_VALID_16_PAIR: | |
873 | entry=offset+MBCS_ENTRY_FINAL_VALUE_16(entry); | |
874 | if(unicodeCodeUnits[entry]==0xfffe) { | |
875 | localSavings+=4; | |
876 | } else { | |
877 | haveAssigned=TRUE; | |
878 | } | |
879 | break; | |
880 | default: | |
881 | break; | |
882 | } | |
883 | } | |
884 | } | |
885 | if(haveAssigned) { | |
886 | return -1; | |
887 | } else { | |
888 | return localSavings+belowSavings; | |
889 | } | |
890 | } | |
891 | ||
892 | /* helper function for finding compaction opportunities */ | |
893 | static void | |
894 | compactToUnicodeHelper(UCMStates *states, | |
895 | uint16_t *unicodeCodeUnits, | |
896 | _MBCSToUFallback *toUFallbacks, int32_t countToUFallbacks) { | |
897 | int32_t state, savings; | |
898 | ||
899 | /* for each initial state */ | |
900 | for(state=0; state<states->countStates; ++state) { | |
901 | if((states->stateFlags[state]&0xf)==MBCS_STATE_FLAG_DIRECT) { | |
902 | savings=findUnassigned(states, | |
903 | unicodeCodeUnits, | |
904 | toUFallbacks, countToUFallbacks, | |
905 | state, 0, 0); | |
906 | if(savings>0) { | |
907 | printf(" all-unassigned sequences from initial state %ld use %ld bytes\n", | |
908 | (long)state, (long)savings); | |
909 | } | |
910 | } | |
911 | } | |
912 | } | |
913 | ||
914 | static int32_t | |
915 | compareFallbacks(const void *context, const void *fb1, const void *fb2) { | |
916 | return ((const _MBCSToUFallback *)fb1)->offset-((const _MBCSToUFallback *)fb2)->offset; | |
917 | } | |
918 | ||
919 | U_CAPI void U_EXPORT2 | |
920 | ucm_optimizeStates(UCMStates *states, | |
921 | uint16_t **pUnicodeCodeUnits, | |
922 | _MBCSToUFallback *toUFallbacks, int32_t countToUFallbacks, | |
923 | UBool verbose) { | |
924 | UErrorCode errorCode; | |
925 | int32_t state, cell, entry; | |
926 | ||
927 | /* test each state table entry */ | |
928 | for(state=0; state<states->countStates; ++state) { | |
929 | for(cell=0; cell<256; ++cell) { | |
930 | entry=states->stateTable[state][cell]; | |
931 | /* | |
932 | * if the entry is a final one with an MBCS_STATE_VALID_DIRECT_16 action code | |
933 | * and the code point is "unassigned" (0xfffe), then change it to | |
934 | * the "unassigned" action code with bits 26..23 set to zero and U+fffe. | |
935 | */ | |
936 | if(MBCS_ENTRY_SET_STATE(entry, 0)==MBCS_ENTRY_FINAL(0, MBCS_STATE_VALID_DIRECT_16, 0xfffe)) { | |
937 | states->stateTable[state][cell]=MBCS_ENTRY_FINAL_SET_ACTION(entry, MBCS_STATE_UNASSIGNED); | |
938 | } | |
939 | } | |
940 | } | |
941 | ||
942 | /* try to compact the toUnicode tables */ | |
943 | if(states->maxCharLength==2) { | |
944 | compactToUnicode2(states, pUnicodeCodeUnits, toUFallbacks, countToUFallbacks, verbose); | |
945 | } else if(states->maxCharLength>2) { | |
946 | if(verbose) { | |
947 | compactToUnicodeHelper(states, *pUnicodeCodeUnits, toUFallbacks, countToUFallbacks); | |
948 | } | |
949 | } | |
950 | ||
951 | /* sort toUFallbacks */ | |
952 | /* | |
953 | * It should be safe to sort them before compactToUnicode2() is called, | |
954 | * because it should not change the relative order of the offset values | |
955 | * that it adjusts, but they need to be sorted at some point, and | |
956 | * it is safest here. | |
957 | */ | |
958 | if(countToUFallbacks>0) { | |
959 | errorCode=U_ZERO_ERROR; /* nothing bad will happen... */ | |
960 | uprv_sortArray(toUFallbacks, countToUFallbacks, | |
961 | sizeof(_MBCSToUFallback), | |
962 | compareFallbacks, NULL, FALSE, &errorCode); | |
963 | } | |
964 | } | |
965 | ||
966 | /* use a complete state table ----------------------------------------------- */ | |
967 | ||
968 | U_CAPI int32_t U_EXPORT2 | |
969 | ucm_countChars(UCMStates *states, | |
970 | const uint8_t *bytes, int32_t length) { | |
971 | uint32_t offset; | |
972 | int32_t i, entry, count; | |
973 | uint8_t state; | |
974 | ||
975 | offset=0; | |
976 | i=count=0; | |
977 | state=0; | |
978 | ||
979 | if(states->countStates==0) { | |
980 | fprintf(stderr, "ucm error: there is no state information!\n"); | |
981 | return -1; | |
982 | } | |
983 | ||
984 | /* for SI/SO (like EBCDIC-stateful), double-byte sequences start in state 1 */ | |
985 | if(length==2 && states->outputType==MBCS_OUTPUT_2_SISO) { | |
986 | state=1; | |
987 | } | |
988 | ||
989 | /* | |
990 | * Walk down the state table like in conversion, | |
991 | * much like getNextUChar(). | |
992 | * We assume that c<=0x10ffff. | |
993 | */ | |
994 | for(i=0; i<length; ++i) { | |
995 | entry=states->stateTable[state][bytes[i]]; | |
996 | if(MBCS_ENTRY_IS_TRANSITION(entry)) { | |
997 | state=(uint8_t)MBCS_ENTRY_TRANSITION_STATE(entry); | |
998 | offset+=MBCS_ENTRY_TRANSITION_OFFSET(entry); | |
999 | } else { | |
1000 | switch(MBCS_ENTRY_FINAL_ACTION(entry)) { | |
1001 | case MBCS_STATE_ILLEGAL: | |
1002 | fprintf(stderr, "ucm error: byte sequence ends in illegal state\n"); | |
1003 | return -1; | |
1004 | case MBCS_STATE_CHANGE_ONLY: | |
1005 | fprintf(stderr, "ucm error: byte sequence ends in state-change-only\n"); | |
1006 | return -1; | |
1007 | case MBCS_STATE_UNASSIGNED: | |
1008 | case MBCS_STATE_FALLBACK_DIRECT_16: | |
1009 | case MBCS_STATE_VALID_DIRECT_16: | |
1010 | case MBCS_STATE_FALLBACK_DIRECT_20: | |
1011 | case MBCS_STATE_VALID_DIRECT_20: | |
1012 | case MBCS_STATE_VALID_16: | |
1013 | case MBCS_STATE_VALID_16_PAIR: | |
1014 | /* count a complete character and prepare for a new one */ | |
1015 | ++count; | |
1016 | state=(uint8_t)MBCS_ENTRY_FINAL_STATE(entry); | |
1017 | offset=0; | |
1018 | break; | |
1019 | default: | |
1020 | /* reserved, must never occur */ | |
1021 | fprintf(stderr, "ucm error: byte sequence reached reserved action code, entry: 0x%02lx\n", (unsigned long)entry); | |
1022 | return -1; | |
1023 | } | |
1024 | } | |
1025 | } | |
1026 | ||
1027 | if(offset!=0) { | |
1028 | fprintf(stderr, "ucm error: byte sequence too short, ends in non-final state %hu\n", state); | |
1029 | return -1; | |
1030 | } | |
1031 | ||
1032 | /* | |
1033 | * for SI/SO (like EBCDIC-stateful), multiple-character results | |
1034 | * must consist of only double-byte sequences | |
1035 | */ | |
1036 | if(count>1 && states->outputType==MBCS_OUTPUT_2_SISO && length!=2*count) { | |
1037 | fprintf(stderr, "ucm error: SI/SO (like EBCDIC-stateful) result with %d characters does not contain all DBCS\n", (int)count); | |
1038 | return -1; | |
1039 | } | |
1040 | ||
1041 | return count; | |
1042 | } |