2 * utf.c: UTF-8 conversion routines
4 * ====================================================================
5 * Licensed to the Apache Software Foundation (ASF) under one
6 * or more contributor license agreements. See the NOTICE file
7 * distributed with this work for additional information
8 * regarding copyright ownership. The ASF licenses this file
9 * to you under the Apache License, Version 2.0 (the
10 * "License"); you may not use this file except in compliance
11 * with the License. You may obtain a copy of the License at
13 * http://www.apache.org/licenses/LICENSE-2.0
15 * Unless required by applicable law or agreed to in writing,
16 * software distributed under the License is distributed on an
17 * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
18 * KIND, either express or implied. See the License for the
19 * specific language governing permissions and limitations
21 * ====================================================================
30 #include <apr_strings.h>
32 #include <apr_xlate.h>
33 #include <apr_atomic.h>
36 #include "svn_string.h"
37 #include "svn_error.h"
38 #include "svn_pools.h"
39 #include "svn_ctype.h"
41 #include "svn_private_config.h"
42 #include "win32_xlate.h"
44 #include "private/svn_utf_private.h"
45 #include "private/svn_dep_compat.h"
46 #include "private/svn_string_private.h"
47 #include "private/svn_mutex.h"
51 /* Use these static strings to maximize performance on standard conversions.
52 * Any strings on other locations are still valid, however.
54 static const char *SVN_UTF_NTOU_XLATE_HANDLE = "svn-utf-ntou-xlate-handle";
55 static const char *SVN_UTF_UTON_XLATE_HANDLE = "svn-utf-uton-xlate-handle";
57 static const char *SVN_APR_UTF8_CHARSET = "UTF-8";
59 static svn_mutex__t *xlate_handle_mutex = NULL;
60 static svn_boolean_t assume_native_charset_is_utf8 = FALSE;
62 /* The xlate handle cache is a global hash table with linked lists of xlate
63 * handles. In multi-threaded environments, a thread "borrows" an xlate
64 * handle from the cache during a translation and puts it back afterwards.
65 * This avoids holding a global lock for all translations.
66 * If there is no handle for a particular key when needed, a new is
67 * handle is created and put in the cache after use.
68 * This means that there will be at most N handles open for a key, where N
69 * is the number of simultanous handles in use for that key. */
71 typedef struct xlate_handle_node_t {
73 /* FALSE if the handle is not valid, since its pool is being
76 /* The name of a char encoding or APR_LOCALE_CHARSET. */
77 const char *frompage, *topage;
78 struct xlate_handle_node_t *next;
79 } xlate_handle_node_t;
81 /* This maps const char * userdata_key strings to xlate_handle_node_t **
82 handles to the first entry in the linked list of xlate handles. We don't
83 store the pointer to the list head directly in the hash table, since we
84 remove/insert entries at the head in the list in the code below, and
85 we can't use apr_hash_set() in each character translation because that
86 function allocates memory in each call where the value is non-NULL.
87 Since these allocations take place in a global pool, this would be a
89 static apr_hash_t *xlate_handle_hash = NULL;
91 /* "1st level cache" to standard conversion maps. We may access these
92 * using atomic xchange ops, i.e. without further thread synchronization.
93 * If the respective item is NULL, fallback to hash lookup.
95 static void * volatile xlat_ntou_static_handle = NULL;
96 static void * volatile xlat_uton_static_handle = NULL;
98 /* Clean up the xlate handle cache. */
100 xlate_cleanup(void *arg)
102 /* We set the cache variables to NULL so that translation works in other
103 cleanup functions, even if it isn't cached then. */
104 xlate_handle_hash = NULL;
106 /* ensure no stale objects get accessed */
107 xlat_ntou_static_handle = NULL;
108 xlat_uton_static_handle = NULL;
113 /* Set the handle of ARG to NULL. */
115 xlate_handle_node_cleanup(void *arg)
117 xlate_handle_node_t *node = arg;
124 svn_utf_initialize2(svn_boolean_t assume_native_utf8,
127 if (!xlate_handle_hash)
129 /* We create our own subpool, which we protect with the mutex.
130 We can't use the pool passed to us by the caller, since we will
131 use it for xlate handle allocations, possibly in multiple threads,
132 and pool allocation is not thread-safe. */
133 apr_pool_t *subpool = svn_pool_create(pool);
135 svn_error_t *err = svn_mutex__init(&mutex, TRUE, subpool);
138 svn_error_clear(err);
142 xlate_handle_mutex = mutex;
143 xlate_handle_hash = apr_hash_make(subpool);
145 apr_pool_cleanup_register(subpool, NULL, xlate_cleanup,
146 apr_pool_cleanup_null);
149 if (!assume_native_charset_is_utf8)
150 assume_native_charset_is_utf8 = assume_native_utf8;
153 /* Return a unique string key based on TOPAGE and FROMPAGE. TOPAGE and
154 * FROMPAGE can be any valid arguments of the same name to
155 * apr_xlate_open(). Allocate the returned string in POOL. */
157 get_xlate_key(const char *topage,
158 const char *frompage,
161 /* In the cases of SVN_APR_LOCALE_CHARSET and SVN_APR_DEFAULT_CHARSET
162 * topage/frompage is really an int, not a valid string. So generate a
163 * unique key accordingly. */
164 if (frompage == SVN_APR_LOCALE_CHARSET)
165 frompage = "APR_LOCALE_CHARSET";
166 else if (frompage == SVN_APR_DEFAULT_CHARSET)
167 frompage = "APR_DEFAULT_CHARSET";
169 if (topage == SVN_APR_LOCALE_CHARSET)
170 topage = "APR_LOCALE_CHARSET";
171 else if (topage == SVN_APR_DEFAULT_CHARSET)
172 topage = "APR_DEFAULT_CHARSET";
174 return apr_pstrcat(pool, "svn-utf-", frompage, "to", topage,
175 "-xlate-handle", (char *)NULL);
178 /* Atomically replace the content in *MEM with NEW_VALUE and return
179 * the previous content of *MEM. If atomicy cannot be guaranteed,
180 * *MEM will not be modified and NEW_VALUE is simply returned to
183 static APR_INLINE void*
184 atomic_swap(void * volatile * mem, void *new_value)
187 #if APR_VERSION_AT_LEAST(1,3,0)
188 /* Cast is necessary because of APR bug:
189 https://issues.apache.org/bugzilla/show_bug.cgi?id=50731 */
190 return apr_atomic_xchgptr((volatile void **)mem, new_value);
192 /* old APRs don't support atomic swaps. Simply return the
193 * input to the caller for further proccessing. */
197 /* no threads - no sync. necessary */
198 void *old_value = (void*)*mem;
204 /* Set *RET to a newly created handle node for converting from FROMPAGE
205 to TOPAGE, If apr_xlate_open() returns APR_EINVAL or APR_ENOTIMPL, set
206 (*RET)->handle to NULL. If fail for any other reason, return the error.
207 Allocate *RET and its xlate handle in POOL. */
209 xlate_alloc_handle(xlate_handle_node_t **ret,
210 const char *topage, const char *frompage,
213 apr_status_t apr_err;
216 /* The error handling doesn't support the following cases, since we don't
217 use them currently. Catch this here. */
218 SVN_ERR_ASSERT(frompage != SVN_APR_DEFAULT_CHARSET
219 && topage != SVN_APR_DEFAULT_CHARSET
220 && (frompage != SVN_APR_LOCALE_CHARSET
221 || topage != SVN_APR_LOCALE_CHARSET));
223 /* Try to create a handle. */
225 apr_err = svn_subr__win32_xlate_open((win32_xlate_t **)&handle, topage,
228 apr_err = apr_xlate_open(&handle, topage, frompage, pool);
231 if (APR_STATUS_IS_EINVAL(apr_err) || APR_STATUS_IS_ENOTIMPL(apr_err))
233 else if (apr_err != APR_SUCCESS)
236 /* Can't use svn_error_wrap_apr here because it calls functions in
237 this file, leading to infinite recursion. */
238 if (frompage == SVN_APR_LOCALE_CHARSET)
239 errstr = apr_psprintf(pool,
240 _("Can't create a character converter from "
241 "native encoding to '%s'"), topage);
242 else if (topage == SVN_APR_LOCALE_CHARSET)
243 errstr = apr_psprintf(pool,
244 _("Can't create a character converter from "
245 "'%s' to native encoding"), frompage);
247 errstr = apr_psprintf(pool,
248 _("Can't create a character converter from "
249 "'%s' to '%s'"), frompage, topage);
251 return svn_error_create(apr_err, NULL, errstr);
254 /* Allocate and initialize the node. */
255 *ret = apr_palloc(pool, sizeof(xlate_handle_node_t));
256 (*ret)->handle = handle;
257 (*ret)->valid = TRUE;
258 (*ret)->frompage = ((frompage != SVN_APR_LOCALE_CHARSET)
259 ? apr_pstrdup(pool, frompage) : frompage);
260 (*ret)->topage = ((topage != SVN_APR_LOCALE_CHARSET)
261 ? apr_pstrdup(pool, topage) : topage);
264 /* If we are called from inside a pool cleanup handler, the just created
265 xlate handle will be closed when that handler returns by a newly
266 registered cleanup handler, however, the handle is still cached by us.
267 To prevent this, we register a cleanup handler that will reset the valid
268 flag of our node, so we don't use an invalid handle. */
270 apr_pool_cleanup_register(pool, *ret, xlate_handle_node_cleanup,
271 apr_pool_cleanup_null);
276 /* Extend xlate_alloc_handle by using USERDATA_KEY as a key in our
277 global hash map, if available.
279 Allocate *RET and its xlate handle in POOL if svn_utf_initialize()
280 hasn't been called or USERDATA_KEY is NULL. Else, allocate them
281 in the pool of xlate_handle_hash.
283 Note: this function is not thread-safe. Call get_xlate_handle_node
286 get_xlate_handle_node_internal(xlate_handle_node_t **ret,
287 const char *topage, const char *frompage,
288 const char *userdata_key, apr_pool_t *pool)
290 /* If we already have a handle, just return it. */
291 if (userdata_key && xlate_handle_hash)
293 xlate_handle_node_t *old_node = NULL;
295 /* 2nd level: hash lookup */
296 xlate_handle_node_t **old_node_p = svn_hash_gets(xlate_handle_hash,
299 old_node = *old_node_p;
302 /* Ensure that the handle is still valid. */
305 /* Remove from the list. */
306 *old_node_p = old_node->next;
307 old_node->next = NULL;
314 /* Note that we still have the mutex locked (if it is initialized), so we
315 can use the global pool for creating the new xlate handle. */
317 /* Use the correct pool for creating the handle. */
318 pool = apr_hash_pool_get(xlate_handle_hash);
320 return xlate_alloc_handle(ret, topage, frompage, pool);
323 /* Set *RET to a handle node for converting from FROMPAGE to TOPAGE,
324 creating the handle node if it doesn't exist in USERDATA_KEY.
325 If a node is not cached and apr_xlate_open() returns APR_EINVAL or
326 APR_ENOTIMPL, set (*RET)->handle to NULL. If fail for any other
327 reason, return the error.
329 Allocate *RET and its xlate handle in POOL if svn_utf_initialize()
330 hasn't been called or USERDATA_KEY is NULL. Else, allocate them
331 in the pool of xlate_handle_hash. */
333 get_xlate_handle_node(xlate_handle_node_t **ret,
334 const char *topage, const char *frompage,
335 const char *userdata_key, apr_pool_t *pool)
337 xlate_handle_node_t *old_node = NULL;
339 /* If we already have a handle, just return it. */
342 if (xlate_handle_hash)
344 /* 1st level: global, static items */
345 if (userdata_key == SVN_UTF_NTOU_XLATE_HANDLE)
346 old_node = atomic_swap(&xlat_ntou_static_handle, NULL);
347 else if (userdata_key == SVN_UTF_UTON_XLATE_HANDLE)
348 old_node = atomic_swap(&xlat_uton_static_handle, NULL);
350 if (old_node && old_node->valid)
359 /* We fall back on a per-pool cache instead. */
360 apr_pool_userdata_get(&p, userdata_key, pool);
362 /* Ensure that the handle is still valid. */
363 if (old_node && old_node->valid)
369 return xlate_alloc_handle(ret, topage, frompage, pool);
373 SVN_MUTEX__WITH_LOCK(xlate_handle_mutex,
374 get_xlate_handle_node_internal(ret,
383 /* Put back NODE into the xlate handle cache for use by other calls.
385 Note: this function is not thread-safe. Call put_xlate_handle_node
388 put_xlate_handle_node_internal(xlate_handle_node_t *node,
389 const char *userdata_key)
391 xlate_handle_node_t **node_p = svn_hash_gets(xlate_handle_hash, userdata_key);
394 userdata_key = apr_pstrdup(apr_hash_pool_get(xlate_handle_hash),
396 node_p = apr_palloc(apr_hash_pool_get(xlate_handle_hash),
399 svn_hash_sets(xlate_handle_hash, userdata_key, node_p);
401 node->next = *node_p;
407 /* Put back NODE into the xlate handle cache for use by other calls.
408 If there is no global cache, store the handle in POOL.
409 Ignore errors related to locking/unlocking the mutex. */
411 put_xlate_handle_node(xlate_handle_node_t *node,
412 const char *userdata_key,
415 assert(node->next == NULL);
419 /* push previous global node to the hash */
420 if (xlate_handle_hash)
422 /* 1st level: global, static items */
423 if (userdata_key == SVN_UTF_NTOU_XLATE_HANDLE)
424 node = atomic_swap(&xlat_ntou_static_handle, node);
425 else if (userdata_key == SVN_UTF_UTON_XLATE_HANDLE)
426 node = atomic_swap(&xlat_uton_static_handle, node);
430 SVN_MUTEX__WITH_LOCK(xlate_handle_mutex,
431 put_xlate_handle_node_internal(node,
436 /* Store it in the per-pool cache. */
437 apr_pool_userdata_set(node, userdata_key, apr_pool_cleanup_null, pool);
443 /* Return the apr_xlate handle for converting native characters to UTF-8. */
445 get_ntou_xlate_handle_node(xlate_handle_node_t **ret, apr_pool_t *pool)
447 return get_xlate_handle_node(ret, SVN_APR_UTF8_CHARSET,
448 assume_native_charset_is_utf8
449 ? SVN_APR_UTF8_CHARSET
450 : SVN_APR_LOCALE_CHARSET,
451 SVN_UTF_NTOU_XLATE_HANDLE, pool);
455 /* Return the apr_xlate handle for converting UTF-8 to native characters.
456 Create one if it doesn't exist. If unable to find a handle, or
457 unable to create one because apr_xlate_open returned APR_EINVAL, then
458 set *RET to null and return SVN_NO_ERROR; if fail for some other
459 reason, return error. */
461 get_uton_xlate_handle_node(xlate_handle_node_t **ret, apr_pool_t *pool)
463 return get_xlate_handle_node(ret,
464 assume_native_charset_is_utf8
465 ? SVN_APR_UTF8_CHARSET
466 : SVN_APR_LOCALE_CHARSET,
467 SVN_APR_UTF8_CHARSET,
468 SVN_UTF_UTON_XLATE_HANDLE, pool);
472 /* Copy LEN bytes of SRC, converting non-ASCII and zero bytes to ?\nnn
473 sequences, allocating the result in POOL. */
475 fuzzy_escape(const char *src, apr_size_t len, apr_pool_t *pool)
477 const char *src_orig = src, *src_end = src + len;
478 apr_size_t new_len = 0;
480 const char *new_orig;
482 /* First count how big a dest string we'll need. */
483 while (src < src_end)
485 if (! svn_ctype_isascii(*src) || *src == '\0')
486 new_len += 5; /* 5 slots, for "?\XXX" */
488 new_len += 1; /* one slot for the 7-bit char */
493 /* Allocate that amount, plus one slot for '\0' character. */
494 new = apr_palloc(pool, new_len + 1);
498 /* And fill it up. */
499 while (src_orig < src_end)
501 if (! svn_ctype_isascii(*src_orig) || src_orig == '\0')
503 /* This is the same format as svn_xml_fuzzy_escape uses, but that
504 function escapes different characters. Please keep in sync!
505 ### If we add another fuzzy escape somewhere, we should abstract
506 ### this out to a common function. */
507 apr_snprintf(new, 6, "?\\%03u", (unsigned char) *src_orig);
524 /* Convert SRC_LENGTH bytes of SRC_DATA in NODE->handle, store the result
525 in *DEST, which is allocated in POOL. */
527 convert_to_stringbuf(xlate_handle_node_t *node,
528 const char *src_data,
529 apr_size_t src_length,
530 svn_stringbuf_t **dest,
534 apr_status_t apr_err;
536 apr_err = svn_subr__win32_xlate_to_stringbuf((win32_xlate_t *) node->handle,
537 src_data, src_length,
540 apr_size_t buflen = src_length * 2;
541 apr_status_t apr_err;
542 apr_size_t srclen = src_length;
543 apr_size_t destlen = buflen;
545 /* Initialize *DEST to an empty stringbuf.
546 A 1:2 ratio of input bytes to output bytes (as assigned above)
547 should be enough for most translations, and if it turns out not
548 to be enough, we'll grow the buffer again, sizing it based on a
549 1:3 ratio of the remainder of the string. */
550 *dest = svn_stringbuf_create_ensure(buflen + 1, pool);
552 /* Not only does it not make sense to convert an empty string, but
553 apr-iconv is quite unreasonable about not allowing that. */
559 /* Set up state variables for xlate. */
560 destlen = buflen - (*dest)->len;
562 /* Attempt the conversion. */
563 apr_err = apr_xlate_conv_buffer(node->handle,
564 src_data + (src_length - srclen),
566 (*dest)->data + (*dest)->len,
569 /* Now, update the *DEST->len to track the amount of output data
570 churned out so far from this loop. */
571 (*dest)->len += ((buflen - (*dest)->len) - destlen);
572 buflen += srclen * 3; /* 3 is middle ground, 2 wasn't enough
573 for all characters in the buffer, 4 is
574 maximum character size (currently) */
577 } while (apr_err == APR_SUCCESS && srclen != 0);
580 /* If we exited the loop with an error, return the error. */
586 /* Can't use svn_error_wrap_apr here because it calls functions in
587 this file, leading to infinite recursion. */
588 if (node->frompage == SVN_APR_LOCALE_CHARSET)
589 errstr = apr_psprintf
590 (pool, _("Can't convert string from native encoding to '%s':"),
592 else if (node->topage == SVN_APR_LOCALE_CHARSET)
593 errstr = apr_psprintf
594 (pool, _("Can't convert string from '%s' to native encoding:"),
597 errstr = apr_psprintf
598 (pool, _("Can't convert string from '%s' to '%s':"),
599 node->frompage, node->topage);
601 err = svn_error_create(apr_err, NULL, fuzzy_escape(src_data,
603 return svn_error_create(apr_err, err, errstr);
605 /* Else, exited due to success. Trim the result buffer down to the
607 (*dest)->data[(*dest)->len] = '\0';
613 /* Return APR_EINVAL if the first LEN bytes of DATA contain anything
614 other than seven-bit, non-control (except for whitespace) ASCII
615 characters, finding the error pool from POOL. Otherwise, return
618 check_non_ascii(const char *data, apr_size_t len, apr_pool_t *pool)
620 const char *data_start = data;
622 for (; len > 0; --len, data++)
624 if ((! svn_ctype_isascii(*data))
625 || ((! svn_ctype_isspace(*data))
626 && svn_ctype_iscntrl(*data)))
628 /* Show the printable part of the data, followed by the
629 decimal code of the questionable character. Because if a
630 user ever gets this error, she's going to have to spend
631 time tracking down the non-ASCII data, so we want to help
632 as much as possible. And yes, we just call the unsafe
633 data "non-ASCII", even though the actual constraint is
634 somewhat more complex than that. */
636 if (data - data_start)
638 const char *error_data
639 = apr_pstrndup(pool, data_start, (data - data_start));
641 return svn_error_createf
643 _("Safe data '%s' was followed by non-ASCII byte %d: "
644 "unable to convert to/from UTF-8"),
645 error_data, *((const unsigned char *) data));
649 return svn_error_createf
651 _("Non-ASCII character (code %d) detected, "
652 "and unable to convert to/from UTF-8"),
653 *((const unsigned char *) data));
661 /* Construct an error with code APR_EINVAL and with a suitable message
662 * to describe the invalid UTF-8 sequence DATA of length LEN (which
663 * may have embedded NULLs). We can't simply print the data, almost
664 * by definition we don't really know how it is encoded.
667 invalid_utf8(const char *data, apr_size_t len, apr_pool_t *pool)
669 const char *last = svn_utf__last_valid(data, len);
670 const char *valid_txt = "", *invalid_txt = "";
672 size_t valid, invalid;
674 /* We will display at most 24 valid octets (this may split a leading
675 multi-byte character) as that should fit on one 80 character line. */
679 for (i = 0; i < valid; ++i)
680 valid_txt = apr_pstrcat(pool, valid_txt,
681 apr_psprintf(pool, " %02x",
682 (unsigned char)last[i-valid]),
685 /* 4 invalid octets will guarantee that the faulty octet is displayed */
686 invalid = data + len - last;
689 for (i = 0; i < invalid; ++i)
690 invalid_txt = apr_pstrcat(pool, invalid_txt,
691 apr_psprintf(pool, " %02x",
692 (unsigned char)last[i]),
695 return svn_error_createf(APR_EINVAL, NULL,
696 _("Valid UTF-8 data\n(hex:%s)\n"
697 "followed by invalid UTF-8 sequence\n(hex:%s)"),
698 valid_txt, invalid_txt);
701 /* Verify that the sequence DATA of length LEN is valid UTF-8.
702 If it is not, return an error with code APR_EINVAL. */
704 check_utf8(const char *data, apr_size_t len, apr_pool_t *pool)
706 if (! svn_utf__is_valid(data, len))
707 return invalid_utf8(data, len, pool);
711 /* Verify that the NULL terminated sequence DATA is valid UTF-8.
712 If it is not, return an error with code APR_EINVAL. */
714 check_cstring_utf8(const char *data, apr_pool_t *pool)
717 if (! svn_utf__cstring_is_valid(data))
718 return invalid_utf8(data, strlen(data), pool);
724 svn_utf_stringbuf_to_utf8(svn_stringbuf_t **dest,
725 const svn_stringbuf_t *src,
728 xlate_handle_node_t *node;
731 SVN_ERR(get_ntou_xlate_handle_node(&node, pool));
735 err = convert_to_stringbuf(node, src->data, src->len, dest, pool);
737 err = check_utf8((*dest)->data, (*dest)->len, pool);
741 err = check_non_ascii(src->data, src->len, pool);
743 *dest = svn_stringbuf_dup(src, pool);
746 return svn_error_compose_create(err,
747 put_xlate_handle_node
749 SVN_UTF_NTOU_XLATE_HANDLE,
755 svn_utf_string_to_utf8(const svn_string_t **dest,
756 const svn_string_t *src,
759 svn_stringbuf_t *destbuf;
760 xlate_handle_node_t *node;
763 SVN_ERR(get_ntou_xlate_handle_node(&node, pool));
767 err = convert_to_stringbuf(node, src->data, src->len, &destbuf, pool);
769 err = check_utf8(destbuf->data, destbuf->len, pool);
771 *dest = svn_stringbuf__morph_into_string(destbuf);
775 err = check_non_ascii(src->data, src->len, pool);
777 *dest = svn_string_dup(src, pool);
780 return svn_error_compose_create(err,
781 put_xlate_handle_node
783 SVN_UTF_NTOU_XLATE_HANDLE,
788 /* Common implementation for svn_utf_cstring_to_utf8,
789 svn_utf_cstring_to_utf8_ex, svn_utf_cstring_from_utf8 and
790 svn_utf_cstring_from_utf8_ex. Convert SRC to DEST using NODE->handle as
791 the translator and allocating from POOL. */
793 convert_cstring(const char **dest,
795 xlate_handle_node_t *node,
800 svn_stringbuf_t *destbuf;
801 SVN_ERR(convert_to_stringbuf(node, src, strlen(src),
803 *dest = destbuf->data;
807 apr_size_t len = strlen(src);
808 SVN_ERR(check_non_ascii(src, len, pool));
809 *dest = apr_pstrmemdup(pool, src, len);
816 svn_utf_cstring_to_utf8(const char **dest,
820 xlate_handle_node_t *node;
823 SVN_ERR(get_ntou_xlate_handle_node(&node, pool));
824 err = convert_cstring(dest, src, node, pool);
825 SVN_ERR(svn_error_compose_create(err,
826 put_xlate_handle_node
828 SVN_UTF_NTOU_XLATE_HANDLE,
830 return check_cstring_utf8(*dest, pool);
835 svn_utf_cstring_to_utf8_ex2(const char **dest,
837 const char *frompage,
840 xlate_handle_node_t *node;
842 const char *convset_key = get_xlate_key(SVN_APR_UTF8_CHARSET, frompage,
845 SVN_ERR(get_xlate_handle_node(&node, SVN_APR_UTF8_CHARSET, frompage,
847 err = convert_cstring(dest, src, node, pool);
848 SVN_ERR(svn_error_compose_create(err,
849 put_xlate_handle_node
851 SVN_UTF_NTOU_XLATE_HANDLE,
854 return check_cstring_utf8(*dest, pool);
859 svn_utf_cstring_to_utf8_ex(const char **dest,
861 const char *frompage,
862 const char *convset_key,
865 return svn_utf_cstring_to_utf8_ex2(dest, src, frompage, pool);
870 svn_utf_stringbuf_from_utf8(svn_stringbuf_t **dest,
871 const svn_stringbuf_t *src,
874 xlate_handle_node_t *node;
877 SVN_ERR(get_uton_xlate_handle_node(&node, pool));
881 err = check_utf8(src->data, src->len, pool);
883 err = convert_to_stringbuf(node, src->data, src->len, dest, pool);
887 err = check_non_ascii(src->data, src->len, pool);
889 *dest = svn_stringbuf_dup(src, pool);
892 err = svn_error_compose_create(
894 put_xlate_handle_node(node, SVN_UTF_UTON_XLATE_HANDLE, pool));
901 svn_utf_string_from_utf8(const svn_string_t **dest,
902 const svn_string_t *src,
905 svn_stringbuf_t *dbuf;
906 xlate_handle_node_t *node;
909 SVN_ERR(get_uton_xlate_handle_node(&node, pool));
913 err = check_utf8(src->data, src->len, pool);
915 err = convert_to_stringbuf(node, src->data, src->len,
918 *dest = svn_stringbuf__morph_into_string(dbuf);
922 err = check_non_ascii(src->data, src->len, pool);
924 *dest = svn_string_dup(src, pool);
927 err = svn_error_compose_create(
929 put_xlate_handle_node(node, SVN_UTF_UTON_XLATE_HANDLE, pool));
936 svn_utf_cstring_from_utf8(const char **dest,
940 xlate_handle_node_t *node;
943 SVN_ERR(check_cstring_utf8(src, pool));
945 SVN_ERR(get_uton_xlate_handle_node(&node, pool));
946 err = convert_cstring(dest, src, node, pool);
947 err = svn_error_compose_create(
949 put_xlate_handle_node(node, SVN_UTF_UTON_XLATE_HANDLE, pool));
956 svn_utf_cstring_from_utf8_ex2(const char **dest,
961 xlate_handle_node_t *node;
963 const char *convset_key = get_xlate_key(topage, SVN_APR_UTF8_CHARSET,
966 SVN_ERR(check_cstring_utf8(src, pool));
968 SVN_ERR(get_xlate_handle_node(&node, topage, SVN_APR_UTF8_CHARSET,
970 err = convert_cstring(dest, src, node, pool);
971 err = svn_error_compose_create(
973 put_xlate_handle_node(node, convset_key, pool));
980 svn_utf_cstring_from_utf8_ex(const char **dest,
983 const char *convset_key,
986 return svn_utf_cstring_from_utf8_ex2(dest, src, topage, pool);
991 svn_utf__cstring_from_utf8_fuzzy(const char *src,
993 svn_error_t *(*convert_from_utf8)
994 (const char **, const char *, apr_pool_t *))
996 const char *escaped, *converted;
999 escaped = fuzzy_escape(src, strlen(src), pool);
1001 /* Okay, now we have a *new* UTF-8 string, one that's guaranteed to
1002 contain only 7-bit bytes :-). Recode to native... */
1003 err = convert_from_utf8(((const char **) &converted), escaped, pool);
1007 svn_error_clear(err);
1013 /* ### Check the client locale, maybe we can avoid that second
1014 * conversion! See Ulrich Drepper's patch at
1015 * http://subversion.tigris.org/issues/show_bug.cgi?id=807.
1021 svn_utf_cstring_from_utf8_fuzzy(const char *src,
1024 return svn_utf__cstring_from_utf8_fuzzy(src, pool,
1025 svn_utf_cstring_from_utf8);
1030 svn_utf_cstring_from_utf8_stringbuf(const char **dest,
1031 const svn_stringbuf_t *src,
1034 svn_stringbuf_t *destbuf;
1036 SVN_ERR(svn_utf_stringbuf_from_utf8(&destbuf, src, pool));
1037 *dest = destbuf->data;
1039 return SVN_NO_ERROR;
1044 svn_utf_cstring_from_utf8_string(const char **dest,
1045 const svn_string_t *src,
1048 svn_stringbuf_t *dbuf;
1049 xlate_handle_node_t *node;
1052 SVN_ERR(get_uton_xlate_handle_node(&node, pool));
1056 err = check_utf8(src->data, src->len, pool);
1058 err = convert_to_stringbuf(node, src->data, src->len,
1065 err = check_non_ascii(src->data, src->len, pool);
1067 *dest = apr_pstrmemdup(pool, src->data, src->len);
1070 err = svn_error_compose_create(
1072 put_xlate_handle_node(node, SVN_UTF_UTON_XLATE_HANDLE, pool));