1 // Copyright (C) 2016 and later: Unicode, Inc. and others.
2 // License & terms of use: http://www.unicode.org/copyright.html
4 *******************************************************************************
6 * Copyright (C) 2003-2014, International Business Machines
7 * Corporation and others. All Rights Reserved.
9 *******************************************************************************
10 * file name: icuswap.cpp
12 * tab size: 8 (not used)
15 * created on: 2003aug08
16 * created by: Markus W. Scherer
18 * This tool takes an ICU data file and "swaps" it, that is, changes its
19 * platform properties between big-/little-endianness and ASCII/EBCDIC charset
21 * The modified data file is written to a new file.
22 * Useful as an install-time tool for shipping only one flavor of ICU data
23 * and preparing data files for the target platform.
24 * Will not work with data DLLs (shared libraries).
27 #include "unicode/utypes.h"
28 #include "unicode/putil.h"
29 #include "unicode/udata.h"
46 #define DEFAULT_PADDING_LENGTH 15
48 static UOption options[]={
50 UOPTION_HELP_QUESTION_MARK,
51 UOPTION_DEF("type", 't', UOPT_REQUIRES_ARG)
56 OPT_HELP_QUESTION_MARK,
64 fseek(f, 0, SEEK_END);
65 size=(int32_t)ftell(f);
66 fseek(f, 0, SEEK_SET);
71 * Swap an ICU .dat package, including swapping of enclosed items.
73 U_CFUNC int32_t U_CALLCONV
74 udata_swapPackage(const char *inFilename, const char *outFilename,
75 const UDataSwapper *ds,
76 const void *inData, int32_t length, void *outData,
77 UErrorCode *pErrorCode);
80 static void U_CALLCONV
81 printError(void *context, const char *fmt, va_list args) {
82 vfprintf((FILE *)context, fmt, args);
87 printUsage(const char *pname, UBool ishelp) {
89 "%csage: %s [ -h, -?, --help ] -tl|-tb|-te|--type=b|... infilename outfilename\n",
90 ishelp ? 'U' : 'u', pname);
93 "\nOptions: -h, -?, --help print this message and exit\n"
94 " Read the input file, swap its platform properties according\n"
95 " to the -t or --type option, and write the result to the output file.\n"
96 " -tl change to little-endian/ASCII charset family\n"
97 " -tb change to big-endian/ASCII charset family\n"
98 " -te change to big-endian/EBCDIC charset family\n");
105 main(int argc, char *argv[]) {
114 const UDataInfo *pInfo;
115 UErrorCode errorCode;
117 UBool outIsBigEndian;
119 U_MAIN_INIT_ARGS(argc, argv);
121 fprintf(stderr, "Warning: icuswap is an obsolete tool and it will be removed in the next ICU release.\nPlease use the icupkg tool instead.\n");
123 /* get the program basename */
124 pname=strrchr(argv[0], U_FILE_SEP_CHAR);
126 pname=strrchr(argv[0], '/');
134 argc=u_parseArgs(argc, argv, UPRV_LENGTHOF(options), options);
135 ishelp=options[OPT_HELP_H].doesOccur || options[OPT_HELP_QUESTION_MARK].doesOccur;
136 if(ishelp || argc!=3) {
137 return printUsage(pname, ishelp);
140 /* parse the output type option */
141 data=(char *)options[OPT_OUT_TYPE].value;
142 if(data[0]==0 || data[1]!=0) {
143 /* the type must be exactly one letter */
144 return printUsage(pname, FALSE);
148 outIsBigEndian=FALSE;
149 outCharset=U_ASCII_FAMILY;
153 outCharset=U_ASCII_FAMILY;
157 outCharset=U_EBCDIC_FAMILY;
160 return printUsage(pname, FALSE);
166 /* open the input file, get its length, allocate memory for it, read the file */
167 in=fopen(argv[1], "rb");
169 fprintf(stderr, "%s: unable to open input file \"%s\"\n", pname, argv[1]);
175 if(length<DEFAULT_PADDING_LENGTH) {
176 fprintf(stderr, "%s: empty input file \"%s\"\n", pname, argv[1]);
182 * +15: udata_swapPackage() may need to add a few padding bytes to the
183 * last item if charset swapping is done,
184 * because the last item may be resorted into the middle and then needs
185 * additional padding bytes
187 data=(char *)malloc(length+DEFAULT_PADDING_LENGTH);
189 fprintf(stderr, "%s: error allocating memory for \"%s\"\n", pname, argv[1]);
194 /* set the last 15 bytes to the usual padding byte, see udata_swapPackage() */
195 uprv_memset(data+length-DEFAULT_PADDING_LENGTH, 0xaa, DEFAULT_PADDING_LENGTH);
197 if(length!=(int32_t)fread(data, 1, length, in)) {
198 fprintf(stderr, "%s: error reading \"%s\"\n", pname, argv[1]);
206 /* swap the data in-place */
207 errorCode=U_ZERO_ERROR;
208 ds=udata_openSwapperForInputData(data, length, outIsBigEndian, outCharset, &errorCode);
209 if(U_FAILURE(errorCode)) {
210 fprintf(stderr, "%s: udata_openSwapperForInputData(\"%s\") failed - %s\n",
211 pname, argv[1], u_errorName(errorCode));
216 ds->printError=printError;
217 ds->printErrorContext=stderr;
219 /* speculative cast, protected by the following length check */
220 pInfo=(const UDataInfo *)((const char *)data+4);
223 pInfo->dataFormat[0]==0x43 && /* dataFormat="CmnD" */
224 pInfo->dataFormat[1]==0x6d &&
225 pInfo->dataFormat[2]==0x6e &&
226 pInfo->dataFormat[3]==0x44
229 * swap the .dat package
230 * udata_swapPackage() needs to rename ToC name entries from the old package
231 * name to the new one.
232 * We pass it the filenames, and udata_swapPackage() will extract the
235 length=udata_swapPackage(argv[1], argv[2], ds, data, length, data, &errorCode);
236 udata_closeSwapper(ds);
237 if(U_FAILURE(errorCode)) {
238 fprintf(stderr, "%s: udata_swapPackage(\"%s\") failed - %s\n",
239 pname, argv[1], u_errorName(errorCode));
244 /* swap the data, which is not a .dat package */
245 length=udata_swap(ds, data, length, data, &errorCode);
246 udata_closeSwapper(ds);
247 if(U_FAILURE(errorCode)) {
248 fprintf(stderr, "%s: udata_swap(\"%s\") failed - %s\n",
249 pname, argv[1], u_errorName(errorCode));
255 out=fopen(argv[2], "wb");
257 fprintf(stderr, "%s: unable to open output file \"%s\"\n", pname, argv[2]);
262 if(length!=(int32_t)fwrite(data, 1, length, out)) {
263 fprintf(stderr, "%s: error writing \"%s\"\n", pname, argv[2]);
287 /* swap .dat package files -------------------------------------------------- */
290 extractPackageName(const UDataSwapper *ds, const char *filename,
291 char pkg[], int32_t capacity,
292 UErrorCode *pErrorCode) {
293 const char *basename;
296 if(U_FAILURE(*pErrorCode)) {
300 basename=findBasename(filename);
301 len=(int32_t)uprv_strlen(basename)-4; /* -4: subtract the length of ".dat" */
303 if(len<=0 || 0!=uprv_strcmp(basename+len, ".dat")) {
304 udata_printError(ds, "udata_swapPackage(): \"%s\" is not recognized as a package filename (must end with .dat)\n",
306 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
311 udata_printError(ds, "udata_swapPackage(): the package name \"%s\" is too long (>=%ld)\n",
313 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
317 uprv_memcpy(pkg, basename, len);
323 uint32_t nameOffset, inOffset, outOffset, length;
327 static int32_t U_CALLCONV
328 compareToCEntries(const void *context, const void *left, const void *right) {
329 const char *chars=(const char *)context;
330 return (int32_t)uprv_strcmp(chars+((const ToCEntry *)left)->nameOffset,
331 chars+((const ToCEntry *)right)->nameOffset);
335 U_CFUNC int32_t U_CALLCONV
336 udata_swapPackage(const char *inFilename, const char *outFilename,
337 const UDataSwapper *ds,
338 const void *inData, int32_t length, void *outData,
339 UErrorCode *pErrorCode) {
340 const UDataInfo *pInfo;
343 const uint8_t *inBytes;
346 uint32_t itemCount, offset, i;
349 const UDataOffsetTOCEntry *inEntries;
350 UDataOffsetTOCEntry *outEntries;
354 char inPkgName[32], outPkgName[32];
355 int32_t inPkgNameLength, outPkgNameLength;
357 /* udata_swapDataHeader checks the arguments */
358 headerSize=udata_swapDataHeader(ds, inData, length, outData, pErrorCode);
359 if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) {
363 /* check data format and format version */
364 pInfo=(const UDataInfo *)((const char *)inData+4);
366 pInfo->dataFormat[0]==0x43 && /* dataFormat="CmnD" */
367 pInfo->dataFormat[1]==0x6d &&
368 pInfo->dataFormat[2]==0x6e &&
369 pInfo->dataFormat[3]==0x44 &&
370 pInfo->formatVersion[0]==1
372 udata_printError(ds, "udata_swapPackage(): data format %02x.%02x.%02x.%02x (format version %02x) is not recognized as an ICU .dat package\n",
373 pInfo->dataFormat[0], pInfo->dataFormat[1],
374 pInfo->dataFormat[2], pInfo->dataFormat[3],
375 pInfo->formatVersion[0]);
376 *pErrorCode=U_UNSUPPORTED_ERROR;
381 * We need to change the ToC name entries so that they have the correct
382 * package name prefix.
383 * Extract the package names from the in/out filenames.
385 inPkgNameLength=extractPackageName(
387 inPkgName, (int32_t)sizeof(inPkgName),
389 outPkgNameLength=extractPackageName(
391 outPkgName, (int32_t)sizeof(outPkgName),
393 if(U_FAILURE(*pErrorCode)) {
398 * It is possible to work with inPkgNameLength!=outPkgNameLength,
399 * but then the length of the data file would change more significantly,
400 * which we are not currently prepared for.
402 if(inPkgNameLength!=outPkgNameLength) {
403 udata_printError(ds, "udata_swapPackage(): the package names \"%s\" and \"%s\" must have the same length\n",
404 inPkgName, outPkgName);
405 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
409 inBytes=(const uint8_t *)inData+headerSize;
410 inEntries=(const UDataOffsetTOCEntry *)(inBytes+4);
414 itemCount=ds->readUInt32(*(const uint32_t *)inBytes);
416 /* no items: count only the item count and return */
420 /* read the last item's offset and preflight it */
421 offset=ds->readUInt32(inEntries[itemCount-1].dataOffset);
422 itemLength=udata_swap(ds, inBytes+offset, -1, NULL, pErrorCode);
424 if(U_SUCCESS(*pErrorCode)) {
425 return headerSize+offset+(uint32_t)itemLength;
430 /* check that the itemCount fits, then the ToC table, then at least the header of the last item */
433 /* itemCount does not fit */
435 itemCount=0; /* make compilers happy */
437 itemCount=ds->readUInt32(*(const uint32_t *)inBytes);
440 } else if((uint32_t)length<(4+8*itemCount)) {
441 /* ToC table does not fit */
444 /* offset of the last item plus at least 20 bytes for its header */
445 offset=20+ds->readUInt32(inEntries[itemCount-1].dataOffset);
448 if((uint32_t)length<offset) {
449 udata_printError(ds, "udata_swapPackage(): too few bytes (%d after header) for a .dat package\n",
451 *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
455 outBytes=(uint8_t *)outData+headerSize;
457 /* swap the item count */
458 ds->swapArray32(ds, inBytes, 4, outBytes, pErrorCode);
461 /* no items: just return now */
465 /* swap the item name strings */
466 offset=4+8*itemCount;
467 itemLength=(int32_t)(ds->readUInt32(inEntries[0].dataOffset)-offset);
468 udata_swapInvStringBlock(ds, inBytes+offset, itemLength, outBytes+offset, pErrorCode);
469 if(U_FAILURE(*pErrorCode)) {
470 udata_printError(ds, "udata_swapPackage() failed to swap the data item name strings\n");
473 /* keep offset and itemLength in case we allocate and copy the strings below */
475 /* swap the package names into the output charset */
476 if(ds->outCharset!=U_CHARSET_FAMILY) {
478 ds2=udata_openSwapper(TRUE, U_CHARSET_FAMILY, TRUE, ds->outCharset, pErrorCode);
479 ds2->swapInvChars(ds2, inPkgName, inPkgNameLength, inPkgName, pErrorCode);
480 ds2->swapInvChars(ds2, outPkgName, outPkgNameLength, outPkgName, pErrorCode);
481 udata_closeSwapper(ds2);
482 if(U_FAILURE(*pErrorCode)) {
483 udata_printError(ds, "udata_swapPackage() failed to swap the input/output package names\n");
487 /* change the prefix of each ToC entry name from the old to the new package name */
491 for(i=0; i<itemCount; ++i) {
492 entryName=(char *)inBytes+ds->readUInt32(inEntries[i].nameOffset);
494 if(0==uprv_memcmp(entryName, inPkgName, inPkgNameLength)) {
495 uprv_memcpy(entryName, outPkgName, inPkgNameLength);
497 udata_printError(ds, "udata_swapPackage() failed: ToC item %ld does not have the input package name as a prefix\n",
499 *pErrorCode=U_INVALID_FORMAT_ERROR;
506 * Allocate the ToC table and, if necessary, a temporary buffer for
507 * pseudo-in-place swapping.
509 * We cannot swap in-place because:
511 * 1. If the swapping of an item fails mid-way, then in-place swapping
512 * has destroyed its data.
513 * Out-of-place swapping allows us to then copy its original data.
515 * 2. If swapping changes the charset family, then we must resort
516 * not only the ToC table but also the data items themselves.
517 * This requires a permutation and is best done with separate in/out
520 * We swapped the strings above to avoid the malloc below if string swapping fails.
522 if(inData==outData) {
523 /* +15: prepare for extra padding of a newly-last item */
524 table=(ToCEntry *)uprv_malloc(itemCount*sizeof(ToCEntry)+length+DEFAULT_PADDING_LENGTH);
526 outBytes=(uint8_t *)(table+itemCount);
528 /* copy the item count and the swapped strings */
529 uprv_memcpy(outBytes, inBytes, 4);
530 uprv_memcpy(outBytes+offset, inBytes+offset, itemLength);
533 table=(ToCEntry *)uprv_malloc(itemCount*sizeof(ToCEntry));
536 udata_printError(ds, "udata_swapPackage(): out of memory allocating %d bytes\n",
538 itemCount*sizeof(ToCEntry)+length+DEFAULT_PADDING_LENGTH :
539 itemCount*sizeof(ToCEntry));
540 *pErrorCode=U_MEMORY_ALLOCATION_ERROR;
543 outEntries=(UDataOffsetTOCEntry *)(outBytes+4);
545 /* read the ToC table */
546 for(i=0; i<itemCount; ++i) {
547 table[i].nameOffset=ds->readUInt32(inEntries[i].nameOffset);
548 table[i].inOffset=ds->readUInt32(inEntries[i].dataOffset);
550 table[i-1].length=table[i].inOffset-table[i-1].inOffset;
553 table[itemCount-1].length=(uint32_t)length-table[itemCount-1].inOffset;
555 if(ds->inCharset==ds->outCharset) {
556 /* no charset swapping, no resorting: keep item offsets the same */
557 for(i=0; i<itemCount; ++i) {
558 table[i].outOffset=table[i].inOffset;
561 /* charset swapping: resort items by their swapped names */
564 * Before the actual sorting, we need to make sure that each item
565 * has a length that is a multiple of 16 bytes so that all items
567 * Only the old last item may be missing up to 15 padding bytes.
568 * Add padding bytes for it.
569 * Since the icuswap main() function has already allocated enough
570 * input buffer space and set the last 15 bytes there to 0xaa,
571 * we only need to increase the total data length and the length
572 * of the last item here.
574 if((length&0xf)!=0) {
575 int32_t delta=16-(length&0xf);
577 table[itemCount-1].length+=(uint32_t)delta;
580 /* Save the offset before we sort the TOC. */
581 offset=table[0].inOffset;
582 /* sort the TOC entries */
583 uprv_sortArray(table, (int32_t)itemCount, (int32_t)sizeof(ToCEntry),
584 compareToCEntries, outBytes, FALSE, pErrorCode);
587 * Note: Before sorting, the inOffset values were in order.
588 * Now the outOffset values are in order.
591 /* assign outOffset values */
592 for(i=0; i<itemCount; ++i) {
593 table[i].outOffset=offset;
594 offset+=table[i].length;
598 /* write the output ToC table */
599 for(i=0; i<itemCount; ++i) {
600 ds->writeUInt32(&outEntries[i].nameOffset, table[i].nameOffset);
601 ds->writeUInt32(&outEntries[i].dataOffset, table[i].outOffset);
604 /* swap each data item */
605 for(i=0; i<itemCount; ++i) {
606 /* first copy the item bytes to make sure that unreachable bytes are copied */
607 uprv_memcpy(outBytes+table[i].outOffset, inBytes+table[i].inOffset, table[i].length);
610 udata_swap(ds, inBytes+table[i].inOffset, (int32_t)table[i].length,
611 outBytes+table[i].outOffset, pErrorCode);
613 if(U_FAILURE(*pErrorCode)) {
614 if(ds->outCharset==U_CHARSET_FAMILY) {
615 udata_printError(ds, "warning: udata_swapPackage() failed to swap item \"%s\"\n"
616 " at inOffset 0x%x length 0x%x - %s\n"
617 " the data item will be copied, not swapped\n\n",
618 (char *)outBytes+table[i].nameOffset,
619 table[i].inOffset, table[i].length, u_errorName(*pErrorCode));
621 udata_printError(ds, "warning: udata_swapPackage() failed to swap an item\n"
622 " at inOffset 0x%x length 0x%x - %s\n"
623 " the data item will be copied, not swapped\n\n",
624 table[i].inOffset, table[i].length, u_errorName(*pErrorCode));
626 /* reset the error code, copy the data item, and continue */
627 *pErrorCode=U_ZERO_ERROR;
628 uprv_memcpy(outBytes+table[i].outOffset, inBytes+table[i].inOffset, table[i].length);
632 if(inData==outData) {
633 /* copy the data from the temporary buffer to the in-place buffer */
634 uprv_memcpy((uint8_t *)outData+headerSize, outBytes, length);
638 return headerSize+length;
643 * Hey, Emacs, please set the following:
646 * indent-tabs-mode: nil