File iconv-option-parsing.patch of Package glibc.21693
From 70d585151c03ede999bd2ad5a724243914cb5f54 Mon Sep 17 00:00:00 2001
From: Arjun Shankar <arjun@redhat.com>
Date: Tue, 7 Jul 2020 20:31:48 +0200
Subject: [PATCH] Rewrite iconv option parsing [BZ #19519]
This commit replaces string manipulation during `iconv_open' and iconv_prog
option parsing with a structured, flag based conversion specification. In
doing so, it alters the internal `__gconv_open' interface and accordingly
adjusts its uses.
This change fixes several hangs in the iconv program and therefore includes
a new test to exercise iconv_prog options that originally led to these hangs.
It also includes a new regression test for option handling in the iconv
function.
Reviewed-by: Florian Weimer <fweimer@redhat.com>
Reviewed-by: Siddhesh Poyarekar <siddhesh@sourceware.org>
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
(cherry picked from commit 91927b7c76437db860cd86a7714476b56bb39d07)
---
From 32965a46ce978f271ec6d3745f951da111c8e74a Mon Sep 17 00:00:00 2001
From: Arjun Shankar <arjun@redhat.com>
Date: Fri, 25 Sep 2020 14:47:06 +0200
Subject: [PATCH] intl: Handle translation output codesets with suffixes [BZ
#26383]
Commit 91927b7c7643 (Rewrite iconv option parsing [BZ #19519]) did not
handle cases where the output codeset for translations (via the `gettext'
family of functions) might have a caller specified encoding suffix such as
TRANSLIT or IGNORE. This led to a regression where translations did not
work when the codeset had a suffix.
This commit fixes the above issue by parsing any suffixes passed to
__dcigettext and adds two new test-cases to intl/tst-codeset.c to
verify correct behaviour. The iconv-internal function __gconv_create_spec
and the static iconv-internal function gconv_destroy_spec are now visible
internally within glibc and used in intl/dcigettext.c.
(cherry picked from commit 7d4ec75e111291851620c6aa2c4460647b7fd50d)
---
Index: glibc-2.26/iconv/Makefile
===================================================================
--- glibc-2.26.orig/iconv/Makefile
+++ glibc-2.26/iconv/Makefile
@@ -26,7 +26,7 @@ headers = iconv.h gconv.h
routines = iconv_open iconv iconv_close \
gconv_open gconv gconv_close gconv_db gconv_conf \
gconv_builtin gconv_simple gconv_trans gconv_cache
-routines += gconv_dl
+routines += gconv_dl gconv_charset
vpath %.c ../locale/programs ../intl
@@ -60,6 +60,7 @@ include $(patsubst %,$(..)libof-iterator
ifeq ($(run-built-tests),yes)
xtests-special += $(objpfx)test-iconvconfig.out
+tests-special += $(objpfx)tst-iconv_prog.out
endif
include ../Rules
@@ -78,3 +79,8 @@ $(objpfx)test-iconvconfig.out: /dev/null
cmp $$tmp $(inst_gconvdir)/gconv-modules.cache; \
rm -f $$tmp) > $@; \
$(evaluate-test)
+
+$(objpfx)tst-iconv_prog.out: tst-iconv_prog.sh $(objpfx)iconv_prog
+ $(BASH) $< $(common-objdir) '$(test-wrapper-env)' \
+ '$(run-program-env)' > $@; \
+ $(evaluate-test)
Index: glibc-2.26/iconv/Versions
===================================================================
--- glibc-2.26.orig/iconv/Versions
+++ glibc-2.26/iconv/Versions
@@ -7,6 +7,9 @@ libc {
# functions shared with iconv program
__gconv_get_alias_db; __gconv_get_cache; __gconv_get_modules_db;
+ # functions used elsewhere in glibc
+ __gconv_open; __gconv_create_spec; __gconv_destroy_spec;
+
# function used by the gconv modules
__gconv_transliterate;
}
Index: glibc-2.26/iconv/gconv_charset.c
===================================================================
--- /dev/null
+++ glibc-2.26/iconv/gconv_charset.c
@@ -0,0 +1,228 @@
+/* Charset name normalization.
+ Copyright (C) 2020 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 of the License, or (at your option) any later version.
+
+ The GNU C Library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+
+#include <stdlib.h>
+#include <ctype.h>
+#include <locale.h>
+#include <stdbool.h>
+#include <string.h>
+#include <sys/stat.h>
+#include "gconv_int.h"
+#include "gconv_charset.h"
+
+
+/* This function returns a pointer to the last suffix in a conversion code
+ string. Valid suffixes matched by this function are of the form: '/' or ','
+ followed by arbitrary text that doesn't contain '/' or ','. It does not
+ edit the string in any way. The caller is expected to parse the suffix and
+ remove it (by e.g. truncating the string) before the next call. */
+static char *
+find_suffix (char *s)
+{
+ /* The conversion code is in the form of a triplet, separated by '/' chars.
+ The third component of the triplet contains suffixes. If we don't have two
+ slashes, we don't have a suffix. */
+
+ int slash_count = 0;
+ char *suffix_term = NULL;
+
+ for (int i = 0; s[i] != '\0'; i++)
+ switch (s[i])
+ {
+ case '/':
+ slash_count++;
+ /* Fallthrough */
+ case ',':
+ suffix_term = &s[i];
+ }
+
+ if (slash_count >= 2)
+ return suffix_term;
+
+ return NULL;
+}
+
+
+struct gconv_parsed_code
+{
+ char *code;
+ bool translit;
+ bool ignore;
+};
+
+
+/* This function parses an iconv_open encoding PC.CODE, strips any suffixes
+ (such as TRANSLIT or IGNORE) from it and sets corresponding flags in it. */
+static void
+gconv_parse_code (struct gconv_parsed_code *pc)
+{
+ pc->translit = false;
+ pc->ignore = false;
+
+ while (1)
+ {
+ /* First drop any trailing whitespaces and separators. */
+ size_t len = strlen (pc->code);
+ while ((len > 0)
+ && (isspace (pc->code[len - 1])
+ || pc->code[len - 1] == ','
+ || pc->code[len - 1] == '/'))
+ len--;
+
+ pc->code[len] = '\0';
+
+ if (len == 0)
+ return;
+
+ char * suffix = find_suffix (pc->code);
+ if (suffix == NULL)
+ {
+ /* At this point, we have processed and removed all suffixes from the
+ code and what remains of the code is suffix free. */
+ return;
+ }
+ else
+ {
+ /* A suffix is processed from the end of the code array going
+ backwards, one suffix at a time. The suffix is an index into the
+ code character array and points to: one past the end of the code
+ and any unprocessed suffixes, and to the beginning of the suffix
+ currently being processed during this iteration. We must process
+ this suffix and then drop it from the code by terminating the
+ preceding text with NULL.
+
+ We want to allow and recognize suffixes such as:
+
+ "/TRANSLIT" i.e. single suffix
+ "//TRANSLIT" i.e. single suffix and multiple separators
+ "//TRANSLIT/IGNORE" i.e. suffixes separated by "/"
+ "/TRANSLIT//IGNORE" i.e. suffixes separated by "//"
+ "//IGNORE,TRANSLIT" i.e. suffixes separated by ","
+ "//IGNORE," i.e. trailing ","
+ "//TRANSLIT/" i.e. trailing "/"
+ "//TRANSLIT//" i.e. trailing "//"
+ "/" i.e. empty suffix.
+
+ Unknown suffixes are silently discarded and ignored. */
+
+ if ((__strcasecmp_l (suffix,
+ GCONV_TRIPLE_SEPARATOR
+ GCONV_TRANSLIT_SUFFIX,
+ _nl_C_locobj_ptr) == 0)
+ || (__strcasecmp_l (suffix,
+ GCONV_SUFFIX_SEPARATOR
+ GCONV_TRANSLIT_SUFFIX,
+ _nl_C_locobj_ptr) == 0))
+ pc->translit = true;
+
+ if ((__strcasecmp_l (suffix,
+ GCONV_TRIPLE_SEPARATOR
+ GCONV_IGNORE_ERRORS_SUFFIX,
+ _nl_C_locobj_ptr) == 0)
+ || (__strcasecmp_l (suffix,
+ GCONV_SUFFIX_SEPARATOR
+ GCONV_IGNORE_ERRORS_SUFFIX,
+ _nl_C_locobj_ptr) == 0))
+ pc->ignore = true;
+
+ /* We just processed this suffix. We can now drop it from the
+ code string by truncating it at the suffix's position. */
+ suffix[0] = '\0';
+ }
+ }
+}
+
+
+/* This function accepts the charset names of the source and destination of the
+ conversion and populates *conv_spec with an equivalent conversion
+ specification that may later be used by __gconv_open. The charset names
+ might contain options in the form of suffixes that alter the conversion,
+ e.g. "ISO-10646/UTF-8/TRANSLIT". It processes the charset names, ignoring
+ and truncating any suffix options in fromcode, and processing and truncating
+ any suffix options in tocode. Supported suffix options ("TRANSLIT" or
+ "IGNORE") when found in tocode lead to the corresponding flag in *conv_spec
+ to be set to true. Unrecognized suffix options are silently discarded. If
+ the function succeeds, it returns conv_spec back to the caller. It returns
+ NULL upon failure. conv_spec must be allocated and freed by the caller. */
+struct gconv_spec *
+__gconv_create_spec (struct gconv_spec *conv_spec, const char *fromcode,
+ const char *tocode)
+{
+ struct gconv_parsed_code pfc, ptc;
+ struct gconv_spec *ret = NULL;
+
+ pfc.code = __strdup (fromcode);
+ ptc.code = __strdup (tocode);
+
+ if ((pfc.code == NULL)
+ || (ptc.code == NULL))
+ goto out;
+
+ gconv_parse_code (&pfc);
+ gconv_parse_code (&ptc);
+
+ /* We ignore suffixes in the fromcode because that is how the current
+ implementation has always handled them. Only suffixes in the tocode are
+ processed and handled. The reality is that invalid input in the input
+ character set should only be ignored if the fromcode specifies IGNORE.
+ The current implementation ignores invalid intput in the input character
+ set if the tocode contains IGNORE. We preserve this behavior for
+ backwards compatibility. In the future we may split the handling of
+ IGNORE to allow a finer grained specification of ignorning invalid input
+ and/or ignoring invalid output. */
+ conv_spec->translit = ptc.translit;
+ conv_spec->ignore = ptc.ignore;
+
+ /* 3 extra bytes because 1 extra for '\0', and 2 extra so strip might
+ be able to add one or two trailing '/' characters if necessary. */
+ conv_spec->fromcode = malloc (strlen (fromcode) + 3);
+ if (conv_spec->fromcode == NULL)
+ goto out;
+
+ conv_spec->tocode = malloc (strlen (tocode) + 3);
+ if (conv_spec->tocode == NULL)
+ {
+ free (conv_spec->fromcode);
+ conv_spec->fromcode = NULL;
+ goto out;
+ }
+
+ /* Strip unrecognized characters and ensure that the code has two '/'
+ characters as per conversion code triplet specification. */
+ strip (conv_spec->fromcode, pfc.code);
+ strip (conv_spec->tocode, ptc.code);
+ ret = conv_spec;
+
+out:
+ free (pfc.code);
+ free (ptc.code);
+
+ return ret;
+}
+libc_hidden_def (__gconv_create_spec)
+
+
+void
+__gconv_destroy_spec (struct gconv_spec *conv_spec)
+{
+ free (conv_spec->fromcode);
+ free (conv_spec->tocode);
+ return;
+}
+libc_hidden_def (__gconv_destroy_spec)
Index: glibc-2.26/iconv/gconv_charset.h
===================================================================
--- glibc-2.26.orig/iconv/gconv_charset.h
+++ glibc-2.26/iconv/gconv_charset.h
@@ -19,9 +19,41 @@
#include <ctype.h>
#include <locale.h>
+#include <stdbool.h>
+#include <string.h>
+#include <sys/stat.h>
+#include <stdlib.h>
+#include "gconv_int.h"
-static void
+/* An iconv encoding is in the form of a triplet, with parts separated by
+ a '/' character. The first part is the standard name, the second part is
+ the character set, and the third part is the error handler. If the first
+ part is sufficient to identify both the standard and the character set
+ then the second part can be empty e.g. UTF-8//. If the first part is not
+ sufficient to identify both the standard and the character set then the
+ second part is required e.g. ISO-10646/UTF8/. If neither the first or
+ second parts are provided e.g. //, then the current locale is used.
+ The actual values used in the first and second parts are not entirely
+ relevant to the implementation. The values themselves are used in a hash
+ table to lookup modules and so the naming convention of the first two parts
+ is somewhat arbitrary and only helps locate the entries in the cache.
+ The third part is the error handler and is comprised of a ',' or '/'
+ separated list of suffixes. Currently, we support "TRANSLIT" for
+ transliteration and "IGNORE" for ignoring conversion errors due to
+ unrecognized input characters. */
+#define GCONV_TRIPLE_SEPARATOR "/"
+#define GCONV_SUFFIX_SEPARATOR ","
+#define GCONV_TRANSLIT_SUFFIX "TRANSLIT"
+#define GCONV_IGNORE_ERRORS_SUFFIX "IGNORE"
+
+
+/* This function copies in-order, characters from the source 's' that are
+ either alpha-numeric or one in one of these: "_-.,:/" - into the destination
+ 'wp' while dropping all other characters. In the process, it converts all
+ alphabetical characters to upper case. It then appends up to two '/'
+ characters so that the total number of '/'es in the destination is 2. */
+static inline void __attribute__ ((unused, always_inline))
strip (char *wp, const char *s)
{
int slash_count = 0;
Index: glibc-2.26/iconv/gconv_int.h
===================================================================
--- glibc-2.26.orig/iconv/gconv_int.h
+++ glibc-2.26/iconv/gconv_int.h
@@ -92,6 +92,15 @@ struct gconv_module
};
+/* The specification of the conversion that needs to be performed. */
+struct gconv_spec
+{
+ char *fromcode;
+ char *tocode;
+ bool translit;
+ bool ignore;
+};
+
/* Flags for `gconv_open'. */
enum
{
@@ -154,10 +163,33 @@ __libc_lock_define (extern, __gconv_lock
})
-/* Return in *HANDLE decriptor for transformation from FROMSET to TOSET. */
-extern int __gconv_open (const char *toset, const char *fromset,
- __gconv_t *handle, int flags)
- internal_function;
+/* Return in *HANDLE, a decriptor for the transformation. The function expects
+ the specification of the transformation in the structure pointed to by
+ CONV_SPEC. It only reads *CONV_SPEC and does not take ownership of it. */
+extern int __gconv_open (struct gconv_spec *conv_spec,
+ __gconv_t *handle, int flags);
+libc_hidden_proto (__gconv_open)
+
+/* This function accepts the charset names of the source and destination of the
+ conversion and populates *conv_spec with an equivalent conversion
+ specification that may later be used by __gconv_open. The charset names
+ might contain options in the form of suffixes that alter the conversion,
+ e.g. "ISO-10646/UTF-8/TRANSLIT". It processes the charset names, ignoring
+ and truncating any suffix options in fromcode, and processing and truncating
+ any suffix options in tocode. Supported suffix options ("TRANSLIT" or
+ "IGNORE") when found in tocode lead to the corresponding flag in *conv_spec
+ to be set to true. Unrecognized suffix options are silently discarded. If
+ the function succeeds, it returns conv_spec back to the caller. It returns
+ NULL upon failure. */
+extern struct gconv_spec *
+__gconv_create_spec (struct gconv_spec *conv_spec, const char *fromcode,
+ const char *tocode);
+libc_hidden_proto (__gconv_create_spec)
+
+/* This function frees all heap memory allocated by __gconv_create_spec. */
+extern void
+__gconv_destroy_spec (struct gconv_spec *conv_spec);
+libc_hidden_proto (__gconv_destroy_spec)
/* Free resources associated with transformation descriptor CD. */
extern int __gconv_close (__gconv_t cd)
Index: glibc-2.26/iconv/gconv_open.c
===================================================================
--- glibc-2.26.orig/iconv/gconv_open.c
+++ glibc-2.26/iconv/gconv_open.c
@@ -27,8 +27,7 @@
int
-internal_function
-__gconv_open (const char *toset, const char *fromset, __gconv_t *handle,
+__gconv_open (struct gconv_spec *conv_spec, __gconv_t *handle,
int flags)
{
struct __gconv_step *steps;
@@ -37,77 +36,38 @@ __gconv_open (const char *toset, const c
size_t cnt = 0;
int res;
int conv_flags = 0;
- const char *errhand;
- const char *ignore;
bool translit = false;
+ char *tocode, *fromcode;
/* Find out whether any error handling method is specified. */
- errhand = strchr (toset, '/');
- if (errhand != NULL)
- errhand = strchr (errhand + 1, '/');
- if (__glibc_likely (errhand != NULL))
- {
- if (*++errhand == '\0')
- errhand = NULL;
- else
- {
- /* Make copy without the error handling description. */
- char *newtoset = (char *) alloca (errhand - toset + 1);
- char *tok;
- char *ptr = NULL /* Work around a bogus warning */;
-
- newtoset[errhand - toset] = '\0';
- toset = memcpy (newtoset, toset, errhand - toset);
+ translit = conv_spec->translit;
- /* Find the appropriate transliteration handlers. */
- tok = strdupa (errhand);
+ if (conv_spec->ignore)
+ conv_flags |= __GCONV_IGNORE_ERRORS;
- tok = __strtok_r (tok, ",", &ptr);
- while (tok != NULL)
- {
- if (__strcasecmp_l (tok, "TRANSLIT", _nl_C_locobj_ptr) == 0)
- translit = true;
- else if (__strcasecmp_l (tok, "IGNORE", _nl_C_locobj_ptr) == 0)
- /* Set the flag to ignore all errors. */
- conv_flags |= __GCONV_IGNORE_ERRORS;
-
- tok = __strtok_r (NULL, ",", &ptr);
- }
- }
- }
-
- /* For the source character set we ignore the error handler specification.
- XXX Is this really always the best? */
- ignore = strchr (fromset, '/');
- if (ignore != NULL && (ignore = strchr (ignore + 1, '/')) != NULL
- && *++ignore != '\0')
- {
- char *newfromset = (char *) alloca (ignore - fromset + 1);
-
- newfromset[ignore - fromset] = '\0';
- fromset = memcpy (newfromset, fromset, ignore - fromset);
- }
+ tocode = conv_spec->tocode;
+ fromcode = conv_spec->fromcode;
/* If the string is empty define this to mean the charset of the
currently selected locale. */
- if (strcmp (toset, "//") == 0)
+ if (strcmp (tocode, "//") == 0)
{
const char *codeset = _NL_CURRENT (LC_CTYPE, CODESET);
size_t len = strlen (codeset);
char *dest;
- toset = dest = (char *) alloca (len + 3);
+ tocode = dest = (char *) alloca (len + 3);
memcpy (__mempcpy (dest, codeset, len), "//", 3);
}
- if (strcmp (fromset, "//") == 0)
+ if (strcmp (fromcode, "//") == 0)
{
const char *codeset = _NL_CURRENT (LC_CTYPE, CODESET);
size_t len = strlen (codeset);
char *dest;
- fromset = dest = (char *) alloca (len + 3);
+ fromcode = dest = (char *) alloca (len + 3);
memcpy (__mempcpy (dest, codeset, len), "//", 3);
}
- res = __gconv_find_transform (toset, fromset, &steps, &nsteps, flags);
+ res = __gconv_find_transform (tocode, fromcode, &steps, &nsteps, flags);
if (res == __GCONV_OK)
{
/* Allocate room for handle. */
@@ -206,3 +166,4 @@ __gconv_open (const char *toset, const c
*handle = result;
return res;
}
+libc_hidden_def (__gconv_open)
Index: glibc-2.26/iconv/iconv_open.c
===================================================================
--- glibc-2.26.orig/iconv/iconv_open.c
+++ glibc-2.26/iconv/iconv_open.c
@@ -31,49 +31,15 @@
iconv_t
iconv_open (const char *tocode, const char *fromcode)
{
- /* Normalize the name. We remove all characters beside alpha-numeric,
- '_', '-', '/', '.', and ':'. */
- size_t tocode_len = strlen (tocode) + 3;
- char *tocode_conv;
- bool tocode_usealloca = __libc_use_alloca (tocode_len);
- if (tocode_usealloca)
- tocode_conv = (char *) alloca (tocode_len);
- else
- {
- tocode_conv = (char *) malloc (tocode_len);
- if (tocode_conv == NULL)
- return (iconv_t) -1;
- }
- strip (tocode_conv, tocode);
- tocode = (tocode_conv[2] == '\0' && tocode[0] != '\0'
- ? upstr (tocode_conv, tocode) : tocode_conv);
+ __gconv_t cd;
+ struct gconv_spec conv_spec;
- size_t fromcode_len = strlen (fromcode) + 3;
- char *fromcode_conv;
- bool fromcode_usealloca = __libc_use_alloca (fromcode_len);
- if (fromcode_usealloca)
- fromcode_conv = (char *) alloca (fromcode_len);
- else
- {
- fromcode_conv = (char *) malloc (fromcode_len);
- if (fromcode_conv == NULL)
- {
- if (! tocode_usealloca)
- free (tocode_conv);
- return (iconv_t) -1;
- }
- }
- strip (fromcode_conv, fromcode);
- fromcode = (fromcode_conv[2] == '\0' && fromcode[0] != '\0'
- ? upstr (fromcode_conv, fromcode) : fromcode_conv);
+ if (__gconv_create_spec (&conv_spec, fromcode, tocode) == NULL)
+ return (iconv_t) -1;
- __gconv_t cd;
- int res = __gconv_open (tocode, fromcode, &cd, 0);
+ int res = __gconv_open (&conv_spec, &cd, 0);
- if (! fromcode_usealloca)
- free (fromcode_conv);
- if (! tocode_usealloca)
- free (tocode_conv);
+ __gconv_destroy_spec (&conv_spec);
if (__builtin_expect (res, __GCONV_OK) != __GCONV_OK)
{
Index: glibc-2.26/iconv/iconv_prog.c
===================================================================
--- glibc-2.26.orig/iconv/iconv_prog.c
+++ glibc-2.26/iconv/iconv_prog.c
@@ -39,6 +39,7 @@
#include <gconv_int.h>
#include "iconv_prog.h"
#include "iconvconfig.h"
+#include "gconv_charset.h"
/* Get libc version number. */
#include "../version.h"
@@ -121,8 +122,7 @@ main (int argc, char *argv[])
{
int status = EXIT_SUCCESS;
int remaining;
- iconv_t cd;
- const char *orig_to_code;
+ __gconv_t cd;
struct charmap_t *from_charmap = NULL;
struct charmap_t *to_charmap = NULL;
@@ -142,39 +142,6 @@ main (int argc, char *argv[])
exit (EXIT_SUCCESS);
}
- /* If we have to ignore errors make sure we use the appropriate name for
- the to-character-set. */
- orig_to_code = to_code;
- if (omit_invalid)
- {
- const char *errhand = strchrnul (to_code, '/');
- int nslash = 2;
- char *newp;
- char *cp;
-
- if (*errhand == '/')
- {
- --nslash;
- errhand = strchrnul (errhand + 1, '/');
-
- if (*errhand == '/')
- {
- --nslash;
- errhand = strchr (errhand, '\0');
- }
- }
-
- newp = (char *) alloca (errhand - to_code + nslash + 7 + 1);
- cp = mempcpy (newp, to_code, errhand - to_code);
- while (nslash-- > 0)
- *cp++ = '/';
- if (cp[-1] != '/')
- *cp++ = ',';
- memcpy (cp, "IGNORE", sizeof ("IGNORE"));
-
- to_code = newp;
- }
-
/* POSIX 1003.2b introduces a silly thing: the arguments to -t anf -f
can be file names of charmaps. In this case iconv will have to read
those charmaps and use them to do the conversion. But there are
@@ -187,10 +154,10 @@ main (int argc, char *argv[])
file. */
from_charmap = charmap_read (from_code, /*0, 1*/1, 0, 0, 0);
- if (strchr (orig_to_code, '/') != NULL)
+ if (strchr (to_code, '/') != NULL)
/* The to-name might be a charmap file name. Try reading the
file. */
- to_charmap = charmap_read (orig_to_code, /*0, 1,*/1, 0, 0, 0);
+ to_charmap = charmap_read (to_code, /*0, 1,*/1, 0, 0, 0);
/* At this point we have to handle two cases. The first one is
@@ -204,9 +171,25 @@ main (int argc, char *argv[])
argc, remaining, argv, output_file);
else
{
+ struct gconv_spec conv_spec;
+ int res;
+
+ if (__gconv_create_spec (&conv_spec, from_code, to_code) == NULL)
+ {
+ error (EXIT_FAILURE, errno,
+ _("failed to start conversion processing"));
+ exit (1);
+ }
+
+ if (omit_invalid)
+ conv_spec.ignore = true;
+
/* Let's see whether we have these coded character sets. */
- cd = iconv_open (to_code, from_code);
- if (cd == (iconv_t) -1)
+ res = __gconv_open (&conv_spec, &cd, 0);
+
+ __gconv_destroy_spec (&conv_spec);
+
+ if (res != __GCONV_OK)
{
if (errno == EINVAL)
{
@@ -224,7 +207,7 @@ main (int argc, char *argv[])
const char *from_pretty =
(from_code[0] ? from_code : nl_langinfo (CODESET));
const char *to_pretty =
- (orig_to_code[0] ? orig_to_code : nl_langinfo (CODESET));
+ (to_code[0] ? to_code : nl_langinfo (CODESET));
if (from_wrong)
{
Index: glibc-2.26/iconv/tst-iconv_prog.sh
===================================================================
--- /dev/null
+++ glibc-2.26/iconv/tst-iconv_prog.sh
@@ -0,0 +1,280 @@
+#!/bin/bash
+# Test for some known iconv(1) hangs from bug 19519, and miscellaneous
+# iconv(1) program error conditions.
+# Copyright (C) 2020 Free Software Foundation, Inc.
+# This file is part of the GNU C Library.
+
+# The GNU C Library is free software; you can redistribute it and/or
+# modify it under the terms of the GNU Lesser General Public
+# License as published by the Free Software Foundation; either
+# version 2.1 of the License, or (at your option) any later version.
+
+# The GNU C Library is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+# Lesser General Public License for more details.
+
+# You should have received a copy of the GNU Lesser General Public
+# License along with the GNU C Library; if not, see
+# <https://www.gnu.org/licenses/>.
+
+codir=$1
+test_wrapper_env="$2"
+run_program_env="$3"
+
+# We have to have some directories in the library path.
+LIBPATH=$codir:$codir/iconvdata
+
+# How the start the iconv(1) program. $from is not defined/expanded yet.
+ICONV='
+$codir/elf/ld.so --library-path $LIBPATH --inhibit-rpath ${from}.so
+$codir/iconv/iconv_prog
+'
+ICONV="$test_wrapper_env $run_program_env $ICONV"
+
+# List of known hangs;
+# Gathered by running an exhaustive 2 byte input search against glibc-2.28
+hangarray=(
+"\x00\x23;-c;ANSI_X3.110;UTF-8//TRANSLIT//IGNORE"
+"\x00\xa1;-c;ARMSCII-8;UTF-8//TRANSLIT//IGNORE"
+"\x00\xa1;-c;ASMO_449;UTF-8//TRANSLIT//IGNORE"
+"\x00\x81;-c;BIG5;UTF-8//TRANSLIT//IGNORE"
+"\x00\xff;-c;BIG5HKSCS;UTF-8//TRANSLIT//IGNORE"
+"\x00\xff;-c;BRF;UTF-8//TRANSLIT//IGNORE"
+"\x00\xff;-c;BS_4730;UTF-8//TRANSLIT//IGNORE"
+"\x00\x81;-c;CP1250;UTF-8//TRANSLIT//IGNORE"
+"\x00\x98;-c;CP1251;UTF-8//TRANSLIT//IGNORE"
+"\x00\x81;-c;CP1252;UTF-8//TRANSLIT//IGNORE"
+"\x00\x81;-c;CP1253;UTF-8//TRANSLIT//IGNORE"
+"\x00\x81;-c;CP1254;UTF-8//TRANSLIT//IGNORE"
+"\x00\x81;-c;CP1255;UTF-8//TRANSLIT//IGNORE"
+"\x00\x81;-c;CP1257;UTF-8//TRANSLIT//IGNORE"
+"\x00\x81;-c;CP1258;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;CP932;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;CSA_Z243.4-1985-1;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;CSA_Z243.4-1985-2;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;DEC-MCS;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;DIN_66003;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;DS_2089;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;EBCDIC-AT-DE;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;EBCDIC-AT-DE-A;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;EBCDIC-CA-FR;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;EBCDIC-DK-NO;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;EBCDIC-DK-NO-A;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;EBCDIC-ES;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;EBCDIC-ES-A;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;EBCDIC-ES-S;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;EBCDIC-FI-SE;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;EBCDIC-FI-SE-A;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;EBCDIC-FR;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;EBCDIC-IS-FRISS;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;EBCDIC-IT;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;EBCDIC-PT;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;EBCDIC-UK;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;EBCDIC-US;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;ES;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;ES2;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;EUC-CN;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;EUC-JISX0213;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;EUC-JP;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;EUC-JP-MS;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;EUC-KR;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;EUC-TW;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;GB18030;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;GB_1988-80;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;GBK;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;GOST_19768-74;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;GREEK7;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;GREEK7-OLD;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;GREEK-CCITT;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;HP-GREEK8;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;HP-ROMAN8;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;HP-ROMAN9;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;HP-THAI8;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;HP-TURKISH8;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;IBM038;UTF-8//TRANSLIT//IGNORE"
+"\x00\x80;-c;IBM1004;UTF-8//TRANSLIT//IGNORE"
+"\x00\xff;-c;IBM1008;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;IBM1046;UTF-8//TRANSLIT//IGNORE"
+"\x00\x51;-c;IBM1132;UTF-8//TRANSLIT//IGNORE"
+"\x00\xa0;-c;IBM1133;UTF-8//TRANSLIT//IGNORE"
+"\x00\xce;-c;IBM1137;UTF-8//TRANSLIT//IGNORE"
+"\x00\x80;-c;IBM1161;UTF-8//TRANSLIT//IGNORE"
+"\x00\xdb;-c;IBM1162;UTF-8//TRANSLIT//IGNORE"
+"\x00\x70;-c;IBM12712;UTF-8//TRANSLIT//IGNORE"
+# These are known hangs that are yet to be fixed:
+# "\x00\x0f;-c;IBM1364;UTF-8"
+# "\x00\x0f;-c;IBM1371;UTF-8"
+# "\x00\x0f;-c;IBM1388;UTF-8"
+# "\x00\x0f;-c;IBM1390;UTF-8"
+# "\x00\x0f;-c;IBM1399;UTF-8"
+"\x00\x53;-c;IBM16804;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;IBM274;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;IBM275;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;IBM281;UTF-8//TRANSLIT//IGNORE"
+"\x00\x57;-c;IBM290;UTF-8//TRANSLIT//IGNORE"
+"\x00\x45;-c;IBM420;UTF-8//TRANSLIT//IGNORE"
+"\x00\x68;-c;IBM423;UTF-8//TRANSLIT//IGNORE"
+"\x00\x70;-c;IBM424;UTF-8//TRANSLIT//IGNORE"
+"\x00\x53;-c;IBM4517;UTF-8//TRANSLIT//IGNORE"
+"\x00\x53;-c;IBM4899;UTF-8//TRANSLIT//IGNORE"
+"\x00\xa5;-c;IBM4909;UTF-8//TRANSLIT//IGNORE"
+"\x00\xdc;-c;IBM4971;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;IBM803;UTF-8//TRANSLIT//IGNORE"
+"\x00\x91;-c;IBM851;UTF-8//TRANSLIT//IGNORE"
+"\x00\x9b;-c;IBM856;UTF-8//TRANSLIT//IGNORE"
+"\x00\xd5;-c;IBM857;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;IBM864;UTF-8//TRANSLIT//IGNORE"
+"\x00\x94;-c;IBM868;UTF-8//TRANSLIT//IGNORE"
+"\x00\x94;-c;IBM869;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;IBM874;UTF-8//TRANSLIT//IGNORE"
+"\x00\x6a;-c;IBM875;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;IBM880;UTF-8//TRANSLIT//IGNORE"
+"\x00\x80;-c;IBM891;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;IBM903;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;IBM904;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;IBM905;UTF-8//TRANSLIT//IGNORE"
+"\x00\x80;-c;IBM9066;UTF-8//TRANSLIT//IGNORE"
+"\x00\x48;-c;IBM918;UTF-8//TRANSLIT//IGNORE"
+"\x00\x57;-c;IBM930;UTF-8//TRANSLIT//IGNORE"
+"\x00\x80;-c;IBM932;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;IBM933;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;IBM935;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;IBM937;UTF-8//TRANSLIT//IGNORE"
+"\x00\x41;-c;IBM939;UTF-8//TRANSLIT//IGNORE"
+"\x00\x80;-c;IBM943;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;INIS;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;INIS-8;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;INIS-CYRILLIC;UTF-8//TRANSLIT//IGNORE"
+"\x00\xec;-c;ISIRI-3342;UTF-8//TRANSLIT//IGNORE"
+"\x00\xec;-c;ISO_10367-BOX;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;ISO-2022-CN;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;ISO-2022-CN-EXT;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;ISO-2022-JP;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;ISO-2022-JP-2;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;ISO-2022-JP-3;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;ISO-2022-KR;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;ISO_2033;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;ISO_5427;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;ISO_5427-EXT;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;ISO_5428;UTF-8//TRANSLIT//IGNORE"
+"\x00\xa4;-c;ISO_6937;UTF-8//TRANSLIT//IGNORE"
+"\x00\xa0;-c;ISO_6937-2;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;ISO-8859-11;UTF-8//TRANSLIT//IGNORE"
+"\x00\xa5;-c;ISO-8859-3;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;ISO-8859-6;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;ISO-8859-7;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;ISO-8859-8;UTF-8//TRANSLIT//IGNORE"
+"\x00\x80;-c;ISO-IR-197;UTF-8//TRANSLIT//IGNORE"
+"\x00\x80;-c;ISO-IR-209;UTF-8//TRANSLIT//IGNORE"
+"\x00\x80;-c;IT;UTF-8//TRANSLIT//IGNORE"
+"\x00\x80;-c;JIS_C6220-1969-RO;UTF-8//TRANSLIT//IGNORE"
+"\x00\x80;-c;JIS_C6229-1984-B;UTF-8//TRANSLIT//IGNORE"
+"\x00\x80;-c;JOHAB;UTF-8//TRANSLIT//IGNORE"
+"\x00\x80;-c;JUS_I.B1.002;UTF-8//TRANSLIT//IGNORE"
+"\x00\x80;-c;KOI-8;UTF-8//TRANSLIT//IGNORE"
+"\x00\x88;-c;KOI8-T;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;KSC5636;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;LATIN-GREEK;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;LATIN-GREEK-1;UTF-8//TRANSLIT//IGNORE"
+"\x00\xf6;-c;MAC-IS;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;MSZ_7795.3;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;NATS-DANO;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;NATS-SEFI;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;NC_NC00-10;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;NF_Z_62-010;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;NF_Z_62-010_1973;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;NS_4551-1;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;NS_4551-2;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;PT;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;PT2;UTF-8//TRANSLIT//IGNORE"
+"\x00\x98;-c;RK1048;UTF-8//TRANSLIT//IGNORE"
+"\x00\x98;-c;SEN_850200_B;UTF-8//TRANSLIT//IGNORE"
+"\x00\x98;-c;SEN_850200_C;UTF-8//TRANSLIT//IGNORE"
+"\x00\x80;-c;Shift_JISX0213;UTF-8//TRANSLIT//IGNORE"
+"\x00\x80;-c;SJIS;UTF-8//TRANSLIT//IGNORE"
+"\x00\x23;-c;T.61-8BIT;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;TIS-620;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;TSCII;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;UHC;UTF-8//TRANSLIT//IGNORE"
+"\x00\xd8;-c;UNICODE;UTF-8//TRANSLIT//IGNORE"
+"\x00\xdc;-c;UTF-16;UTF-8//TRANSLIT//IGNORE"
+"\xdc\x00;-c;UTF-16BE;UTF-8//TRANSLIT//IGNORE"
+"\x00\xdc;-c;UTF-16LE;UTF-8//TRANSLIT//IGNORE"
+"\xff\xff;-c;UTF-7;UTF-8//TRANSLIT//IGNORE"
+"\x00\x81;-c;WIN-SAMI-2;UTF-8//TRANSLIT//IGNORE"
+)
+
+# List of option combinations that *should* lead to an error
+errorarray=(
+# Converting from/to invalid character sets should cause error
+"\x00\x00;;INVALID;INVALID"
+"\x00\x00;;INVALID;UTF-8"
+"\x00\x00;;UTF-8;INVALID"
+)
+
+# Requires $twobyte input, $c flag, $from, and $to to be set; sets $ret
+execute_test ()
+{
+ eval PROG=\"$ICONV\"
+ echo -en "$twobyte" \
+ | timeout -k 4 3 $PROG $c -f $from -t "$to" &>/dev/null
+ ret=$?
+}
+
+check_hangtest_result ()
+{
+ if [ "$ret" -eq "124" ] || [ "$ret" -eq "137" ]; then # timeout/hang
+ result="HANG"
+ else
+ if [ "$ret" -eq "139" ]; then # segfault
+ result="SEGFAULT"
+ else
+ if [ "$ret" -gt "127" ]; then # unexpected error
+ result="UNEXPECTED"
+ else
+ result="OK"
+ fi
+ fi
+ fi
+
+ echo -n "$result: from: \"$from\", to: \"$to\","
+ echo " input \"$twobyte\", flags \"$c\""
+
+ if [ "$result" != "OK" ]; then
+ exit 1
+ fi
+}
+
+for hangcommand in "${hangarray[@]}"; do
+ twobyte="$(echo "$hangcommand" | cut -d";" -f 1)"
+ c="$(echo "$hangcommand" | cut -d";" -f 2)"
+ from="$(echo "$hangcommand" | cut -d";" -f 3)"
+ to="$(echo "$hangcommand" | cut -d";" -f 4)"
+ execute_test
+ check_hangtest_result
+done
+
+check_errtest_result ()
+{
+ if [ "$ret" -eq "1" ]; then # we errored out as expected
+ result="PASS"
+ else
+ result="FAIL"
+ fi
+ echo -n "$result: from: \"$from\", to: \"$to\","
+ echo " input \"$twobyte\", flags \"$c\", return code $ret"
+
+ if [ "$result" != "PASS" ]; then
+ exit 1
+ fi
+}
+
+for errorcommand in "${errorarray[@]}"; do
+ twobyte="$(echo "$errorcommand" | cut -d";" -f 1)"
+ c="$(echo "$errorcommand" | cut -d";" -f 2)"
+ from="$(echo "$errorcommand" | cut -d";" -f 3)"
+ to="$(echo "$errorcommand" | cut -d";" -f 4)"
+ execute_test
+ check_errtest_result
+done
Index: glibc-2.26/intl/dcigettext.c
===================================================================
--- glibc-2.26.orig/intl/dcigettext.c
+++ glibc-2.26/intl/dcigettext.c
@@ -1126,11 +1126,19 @@ _nl_find_msg (struct loaded_l10nfile *do
outcharset = encoding;
# ifdef _LIBC
+
+ struct gconv_spec conv_spec;
+
+ __gconv_create_spec (&conv_spec, charset, outcharset);
+
/* We always want to use transliteration. */
- outcharset = norm_add_slashes (outcharset, "TRANSLIT");
- charset = norm_add_slashes (charset, "");
- int r = __gconv_open (outcharset, charset, &convd->conv,
- GCONV_AVOID_NOCONV);
+ conv_spec.translit = true;
+
+ int r = __gconv_open (&conv_spec, &convd->conv,
+ GCONV_AVOID_NOCONV);
+
+ __gconv_destroy_spec (&conv_spec);
+
if (__builtin_expect (r != __GCONV_OK, 0))
{
/* If the output encoding is the same there is
