rpms/glibc
5
0
Fork 2
Code Issues Pull Requests Releases Activity

Support in-place file conversion in the iconv tool (RHEL-71530)

Browse Source
This commit is contained in:
Florian Weimer 2024-12-23 14:59:23 +01:00
parent f0a7516eb0
commit 071446cc90
11 changed files with 2882 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

194
glibc-RHEL-71530-1.patch Normal file
View File

@ -0,0 +1,194 @@
commit 1b0a2062c8938c7333cd118d85d9976c4e7c92af
Author: Andreas Schwab <schwab@suse.de>
Date: Mon Jun 10 12:19:17 2024 +0200
iconv: Fix matching of multi-character transliterations (bug 31859)
Only return __GCONV_INCOMPLETE_INPUT for a partial match when the end of
the input buffer is reached. Otherwise it is a non-match, and other
patterns should be tried.
diff --git a/iconv/Makefile b/iconv/Makefile
index 63afc853ff65967a..65b4a44ab86cf0dc 100644
--- a/iconv/Makefile
+++ b/iconv/Makefile
@@ -57,6 +57,10 @@ tests = \
tst-iconv-opt \
# tests
+test-srcs := \
+ tst-translit-mchar \
+ # test-srcs
+
others = iconv_prog iconvconfig
install-others-programs = $(inst_bindir)/iconv
install-sbin = iconvconfig
@@ -73,6 +77,7 @@ include $(patsubst %,$(..)libof-iterator.mk,$(cpp-srcs-left))
ifeq ($(run-built-tests),yes)
xtests-special += $(objpfx)test-iconvconfig.out
tests-special += $(objpfx)tst-iconv_prog.out
+tests-special += $(objpfx)tst-translit-mchar.out
endif
# Make a copy of the file because gconv module names are constructed
@@ -92,6 +97,8 @@ $(objpfx)tst-gconv-init-failure.out: \
$(objpfx)gconv-modules $(objpfx)tst-gconv-init-failure-mod.so
endif
+generated-dirs += tst-translit
+
include ../Rules
ifeq ($(run-built-tests),yes)
@@ -126,3 +133,11 @@ $(objpfx)tst-iconv_prog.out: tst-iconv_prog.sh $(objpfx)iconv_prog
$(BASH) $< $(common-objdir) '$(test-wrapper-env)' \
'$(run-program-env)' > $@; \
$(evaluate-test)
+
+$(objpfx)tst-translit-mchar.out: tst-translit-mchar.sh \
+ $(objpfx)tst-translit-mchar \
+ tst-translit-locale
+ $(SHELL) $< $(common-objpfx) '$(run-program-prefix-before-env)' \
+ '$(run-program-env)' '$(run-program-prefix-after-env)' \
+ $< > $@; \
+ $(evaluate-test)
diff --git a/iconv/gconv_trans.c b/iconv/gconv_trans.c
index 08b7a3f71dad5f1e..44f0fd849a3f82a3 100644
--- a/iconv/gconv_trans.c
+++ b/iconv/gconv_trans.c
@@ -150,7 +150,7 @@ __gconv_transliterate (struct __gconv_step *step,
/* Nothing found, continue searching. */
}
- else if (cnt > 0)
+ else if (cnt > 0 && winbuf + cnt == winbufend)
/* This means that the input buffer contents matches a prefix of
an entry. Since we cannot match it unless we get more input,
we will tell the caller about it. */
diff --git a/iconv/tst-translit-locale b/iconv/tst-translit-locale
new file mode 100644
index 0000000000000000..712b08628a64dc11
--- /dev/null
+++ b/iconv/tst-translit-locale
@@ -0,0 +1,10 @@
+# Test multi-character transliteration rule
+
+LC_CTYPE
+copy "POSIX"
+
+translit_start
+"ÄÄ" "AA"
+translit_end
+
+END LC_CTYPE
diff --git a/iconv/tst-translit-mchar.c b/iconv/tst-translit-mchar.c
new file mode 100644
index 0000000000000000..7d432ea6679b60fa
--- /dev/null
+++ b/iconv/tst-translit-mchar.c
@@ -0,0 +1,48 @@
+/* Test multi-character transliterations.
+ Copyright (C) 2024 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/>. */
+
+#include <locale.h>
+#include <iconv.h>
+#include <support/support.h>
+#include <support/check.h>
+
+static int
+do_test (void)
+{
+ iconv_t cd;
+ /* An input sequence that shares a common prefix with a transliteration
+ rule. */
+ char input[] = "ÄÅ";
+ char *inptr = input;
+ char outbuf[10];
+ char *outptr = outbuf;
+ size_t inlen = sizeof (input), outlen = sizeof (outbuf);
+ size_t n;
+
+ xsetlocale (LC_CTYPE, "tst-translit");
+
+ cd = iconv_open ("ASCII//TRANSLIT", "UTF-8");
+ TEST_VERIFY (cd != (iconv_t) -1);
+
+ /* This call used to loop infinitely. */
+ n = iconv (cd, &inptr, &inlen, &outptr, &outlen);
+ TEST_VERIFY (iconv_close (cd) == 0);
+ return n == 0;
+}
+
+#include <support/test-driver.c>
diff --git a/iconv/tst-translit-mchar.sh b/iconv/tst-translit-mchar.sh
new file mode 100644
index 0000000000000000..ab7a7f872911cf21
--- /dev/null
+++ b/iconv/tst-translit-mchar.sh
@@ -0,0 +1,51 @@
+#!/bin/sh
+# Testing of multi-character transliterations
+# Copyright (C) 2024 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/>.
+
+set -e
+
+common_objpfx=$1
+run_program_prefix_before_env=$2
+run_program_env=$3
+run_program_prefix_after_env=$4
+
+# Generate data files.
+# The locale only defines the LC_CTYPE category, so we expect a failure
+# due to warnings.
+ret=0
+${run_program_prefix_before_env} \
+${run_program_env} \
+I18NPATH=../localedata \
+${run_program_prefix_after_env} ${common_objpfx}locale/localedef \
+--quiet -i tst-translit-locale -f UTF-8 ${common_objpfx}iconv/tst-translit || ret=$?
+if [ $ret -gt 1 ]; then
+ echo "FAIL: Locale compilation for tst-translit-locale failed (error $ret)."
+ exit 1
+fi
+
+set -x
+
+# Run the test.
+${run_program_prefix_before_env} \
+${run_program_env} \
+LOCPATH=${common_objpfx}iconv \
+${run_program_prefix_after_env} ${common_objpfx}iconv/tst-translit-mchar
+
+# Local Variables:
+# mode:shell-script
+# End:

108
glibc-RHEL-71530-10.patch Normal file
View File

@ -0,0 +1,108 @@
commit 9a4b0eaf726f5404c6683d5c7c5e86f61c3f3fbc
Author: Aurelien Jarno <aurelien@aurel32.net>
Date: Sat Dec 14 11:44:11 2024 +0100
iconv: do not report error exit with transliteration [BZ #32448]
Commit 6cbf845fcdc7 ("iconv: Preserve iconv -c error exit on invalid
inputs (bug 32046)") changed the error exit code to report an error when
an input character has been transliterated. This looks like a bug as the
moto in the iconv program is to report an error code in the same
condition as the iconv() function.
This happens because the STANDARD_TO_LOOP_ERR_HANDLER macro sets a
default value for result and later updates it if the transliteration
succeed. With the changes, setting the default value also marks the
input as illegal.
Fix that by setting up the default value of result only when the
transliteration is not used. This works because __gconv_transliterate()
calls __gconv_mark_illegal_input() to return an error. At the same time
also fix the typo outself -> ourselves.
Fixes: 6cbf845fcdc7
Resolves: BZ #32448
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
diff --git a/iconv/loop.c b/iconv/loop.c
index 199fb283266fb9ca..7149cec9b215a918 100644
--- a/iconv/loop.c
+++ b/iconv/loop.c
@@ -141,12 +141,13 @@
points. */
#define STANDARD_TO_LOOP_ERR_HANDLER(Incr) \
{ \
- result = __gconv_mark_illegal_input (step_data); \
- \
if (irreversible == NULL) \
- /* This means we are in call from __gconv_transliterate. In this \
- case we are not doing any error recovery outself. */ \
- break; \
+ { \
+ /* This means we are in call from __gconv_transliterate. In this \
+ case we are not doing any error recovery ourselves. */ \
+ result = __gconv_mark_illegal_input (step_data); \
+ break; \
+ } \
\
/* If needed, flush any conversion state, so that __gconv_transliterate \
starts with current shift state. */ \
@@ -157,6 +158,8 @@
result = __gconv_transliterate \
(step, step_data, *inptrp, \
&inptr, inend, &outptr, irreversible); \
+ else \
+ result = __gconv_mark_illegal_input (step_data); \
\
REINIT_PARAMS; \
\
diff --git a/iconv/tst-iconv_prog.sh b/iconv/tst-iconv_prog.sh
index ca4dbd4a3a3318fe..f3a03ac062a70b05 100644
--- a/iconv/tst-iconv_prog.sh
+++ b/iconv/tst-iconv_prog.sh
@@ -211,12 +211,13 @@ hangarray=(
"\x00\x81;-c;WIN-SAMI-2;UTF-8//TRANSLIT//IGNORE"
)
-# List of option combinations that *should* lead to an error
-errorarray=(
+# List of option combinations with their expected return code
+testarray=(
# Converting from/to invalid character sets should cause error
-"\x00\x00;;INVALID;INVALID"
-"\x00\x00;;INVALID;UTF-8"
-"\x00\x00;;UTF-8;INVALID"
+"\x00\x00;;INVALID;INVALID;1"
+"\x00\x00;;INVALID;UTF-8;1"
+"\x00\x00;;UTF-8;INVALID;1"
+"\xc3\xa9;;UTF-8;ASCII//TRANSLIT;0"
)
# Requires $twobyte input, $c flag, $from, and $to to be set; sets $ret
@@ -264,7 +265,7 @@ done
check_errtest_result ()
{
- if [ "$ret" -eq "1" ]; then # we errored out as expected
+ if [ "$ret" -eq "$eret" ]; then # we got the expected return code
result="PASS"
else
result="FAIL"
@@ -277,11 +278,12 @@ check_errtest_result ()
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 "$errorcommand" | cut -d";" -f 4)"
+for testcommand in "${testarray[@]}"; do
+ twobyte="$(echo "$testcommand" | cut -d";" -f 1)"
+ c="$(echo "$testcommand" | cut -d";" -f 2)"
+ from="$(echo "$testcommand" | cut -d";" -f 3)"
+ to="$(echo "$testcommand" | cut -d";" -f 4)"
+ eret="$(echo "$testcommand" | cut -d";" -f 5)"
execute_test
check_errtest_result
done

218
glibc-RHEL-71530-2.patch Normal file
View File

@ -0,0 +1,218 @@
commit 422ed8ede312f786369e4850e47b8d32beaae4e4
Author: Florian Weimer <fweimer@redhat.com>
Date: Fri Sep 20 13:10:54 2024 +0200
iconv: Base tests for buffer management
Reviewed-by: DJ Delorie <dj@redhat.com>
diff --git a/iconv/Makefile b/iconv/Makefile
index 65b4a44ab86cf0dc..b0fa550141db5a06 100644
--- a/iconv/Makefile
+++ b/iconv/Makefile
@@ -76,8 +76,11 @@ include $(patsubst %,$(..)libof-iterator.mk,$(cpp-srcs-left))
ifeq ($(run-built-tests),yes)
xtests-special += $(objpfx)test-iconvconfig.out
-tests-special += $(objpfx)tst-iconv_prog.out
-tests-special += $(objpfx)tst-translit-mchar.out
+tests-special += \
+ $(objpfx)tst-iconv_prog-buffer.out \
+ $(objpfx)tst-iconv_prog.out \
+ $(objpfx)tst-translit-mchar.out \
+ # tests-special
endif
# Make a copy of the file because gconv module names are constructed
@@ -141,3 +144,8 @@ $(objpfx)tst-translit-mchar.out: tst-translit-mchar.sh \
'$(run-program-env)' '$(run-program-prefix-after-env)' \
$< > $@; \
$(evaluate-test)
+
+$(objpfx)tst-iconv_prog-buffer.out: \
+ tst-iconv_prog-buffer.sh $(objpfx)iconv_prog
+ $(BASH) $< $(common-objdir) '$(test-program-prefix)' > $@; \
+ $(evaluate-test)
diff --git a/iconv/tst-iconv_prog-buffer.sh b/iconv/tst-iconv_prog-buffer.sh
new file mode 100644
index 0000000000000000..a27107f02b95cdc7
--- /dev/null
+++ b/iconv/tst-iconv_prog-buffer.sh
@@ -0,0 +1,177 @@
+#!/bin/bash
+# Test for iconv (the program) buffer management.
+# Copyright (C) 2024 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/>.
+
+exec 2>&1
+set -e
+
+exec {logfd}>&1
+
+codir=$1
+test_program_prefix="$2"
+
+# Use internal converters to avoid issues with module loading.
+iconv_args="-f ASCII -t UTF-8"
+
+failure=false
+
+tmp=`mktemp -d`
+trap 'rm -rf "$tmp"' 0
+echo ABC > "$tmp/abc"
+echo DEF > "$tmp/def"
+echo GGG > "$tmp/ggg"
+echo HH > "$tmp/hh"
+echo XY > "$tmp/xy"
+echo ZT > "$tmp/zt"
+echo OUT > "$tmp/out-template"
+printf '\xff' > "$tmp/0xff"
+cat "$tmp/xy" "$tmp/0xff" "$tmp/zt" > "$tmp/0xff-wrapped"
+
+run_iconv () {
+ local c=0
+ if test "${FUNCNAME[2]}" = main; then
+ c=1
+ fi
+ echo "${BASH_SOURCE[$c]}:${BASH_LINENO[$c]}: iconv $iconv_args $@" >&$logfd
+ $test_program_prefix $codir/iconv/iconv_prog $iconv_args "$@"
+}
+
+check_out_expected () {
+ if ! cmp -s "$tmp/out" "$tmp/expected" ; then
+ echo "error: iconv output difference" >&$logfd
+ echo "*** expected ***" >&$logfd
+ cat "$tmp/expected" >&$logfd
+ echo "*** actual ***" >&$logfd
+ cat "$tmp/out" >&$logfd
+ failure=true
+ fi
+}
+
+expect_files () {
+ local f
+ ! test -z "$1"
+ cp "$tmp/$1" "$tmp/expected"
+ shift
+ for f in "$@" ; do
+ cat "$tmp/$f" >> "$tmp/expected"
+ done
+ check_out_expected
+}
+
+check_out () {
+ cat > "$tmp/expected"
+ check_out_expected
+}
+
+expect_exit () {
+ local expected=$1
+ shift
+ # Prevent failure for stopping the script.
+ if "$@" ; then
+ actual=$?
+ else
+ actual=$?
+ fi
+ if test "$actual" -ne "$expected"; then
+ echo "error: expected exit status $expected, not $actual" >&$logfd
+ exit 1
+ fi
+}
+
+ignore_failure () {
+ set +e
+ "$@"
+ status=$?
+ set -e
+}
+
+# Concatentation test.
+run_iconv -o "$tmp/out" "$tmp/abc" "$tmp/def"
+expect_files abc def
+
+# Single-file in-place conversion.
+run_iconv -o "$tmp/out" "$tmp/out"
+expect_files abc def
+
+# Multiple input files with in-place conversion.
+
+run_iconv -o "$tmp/out" "$tmp/out" "$tmp/abc"
+expect_files abc def abc
+
+# But not if we are writing to standard output.
+
+cp "$tmp/out-template" "$tmp/out"
+run_iconv </dev/null >>"$tmp/out"
+expect_files out-template
+
+cp "$tmp/out-template" "$tmp/out"
+run_iconv - </dev/null >>"$tmp/out"
+expect_files out-template
+
+cp "$tmp/out-template" "$tmp/out"
+run_iconv /dev/null >>"$tmp/out"
+expect_files out-template
+
+# Conversion errors should avoid clobbering an existing file if
+# it is also an input file.
+
+cp "$tmp/0xff" "$tmp/out"
+expect_exit 1 run_iconv -o "$tmp/out" "$tmp/out"
+expect_files 0xff
+
+cp "$tmp/0xff" "$tmp/out"
+expect_exit 1 run_iconv -o "$tmp/out" < "$tmp/out"
+expect_files 0xff
+
+cp "$tmp/0xff" "$tmp/out"
+expect_exit 1 run_iconv -o "$tmp/out" - < "$tmp/out"
+expect_files 0xff
+
+# If errors are ignored, the file should be overwritten.
+
+cp "$tmp/out-template" "$tmp/out"
+expect_exit 1 \
+ run_iconv -c -o "$tmp/out" "$tmp/abc" "$tmp/0xff" "$tmp/def" 2>"$tmp/err"
+! test -s "$tmp/err"
+expect_files abc def
+
+# FIXME: This is not correct, -c should not change the exit status.
+cp "$tmp/out-template" "$tmp/out"
+run_iconv -c -o "$tmp/out" \
+ "$tmp/abc" "$tmp/0xff-wrapped" "$tmp/def" 2>"$tmp/err"
+! test -s "$tmp/err"
+expect_files abc xy zt def
+
+# If the file does not exist yet, it should not be created on error.
+
+rm "$tmp/out"
+expect_exit 1 run_iconv -o "$tmp/out" "$tmp/0xff"
+! test -e "$tmp/out"
+
+expect_exit 1 run_iconv -o "$tmp/out" < "$tmp/0xff"
+! test -e "$tmp/out"
+
+expect_exit 1 run_iconv -o "$tmp/out" "$tmp/abc" "$tmp/0xff" "$tmp/def"
+! test -e "$tmp/out"
+
+expect_exit 1 run_iconv -o "$tmp/out" "$tmp/abc" - < "$tmp/0xff" "$tmp/def"
+! test -e "$tmp/out"
+
+if $failure ; then
+ exit 1
+fi

91
glibc-RHEL-71530-3.patch Normal file
View File

@ -0,0 +1,91 @@
commit 0cb64617a6f691b611406427c8e24b7f04c4983f
Author: Florian Weimer <fweimer@redhat.com>
Date: Fri Sep 20 13:10:54 2024 +0200
iconv: Do not use mmap in iconv (the program) (bug 17703)
On current systems, very large files are needed before
mmap becomes beneficial. Simplify the implementation.
This exposed that inptr was not initialized correctly in
process_fd. Handling multiple input files resulted in
EFAULT in read because a null pointer was passed. This
could be observed previously if an input file was not
mappable and was reported as bug 17703.
Reviewed-by: DJ Delorie <dj@redhat.com>
diff --git a/iconv/iconv_prog.c b/iconv/iconv_prog.c
index a765b1af21d2bde0..88a928557e7afb0a 100644
--- a/iconv/iconv_prog.c
+++ b/iconv/iconv_prog.c
@@ -31,9 +31,6 @@
#include <string.h>
#include <unistd.h>
#include <libintl.h>
-#ifdef _POSIX_MAPPED_FILES
-# include <sys/mman.h>
-#endif
#include <charmap.h>
#include <gconv_int.h>
#include "iconv_prog.h"
@@ -253,10 +250,6 @@ conversions from `%s' and to `%s' are not supported"),
else
do
{
-#ifdef _POSIX_MAPPED_FILES
- struct stat64 st;
- char *addr;
-#endif
int fd, ret;
if (verbose)
@@ -276,39 +269,6 @@ conversions from `%s' and to `%s' are not supported"),
}
}
-#ifdef _POSIX_MAPPED_FILES
- /* We have possibilities for reading the input file. First try
- to mmap() it since this will provide the fastest solution. */
- if (fstat64 (fd, &st) == 0
- && ((addr = mmap (NULL, st.st_size, PROT_READ, MAP_PRIVATE,
- fd, 0)) != MAP_FAILED))
- {
- /* Yes, we can use mmap(). The descriptor is not needed
- anymore. */
- if (close (fd) != 0)
- error (EXIT_FAILURE, errno,
- _("error while closing input `%s'"),
- argv[remaining]);
-
- ret = process_block (cd, addr, st.st_size, &output,
- output_file);
-
- /* We don't need the input data anymore. */
- munmap ((void *) addr, st.st_size);
-
- if (ret != 0)
- {
- status = EXIT_FAILURE;
-
- if (ret < 0)
- /* We cannot go on with producing output since it might
- lead to problem because the last output might leave
- the output stream in an undefined state. */
- break;
- }
- }
- else
-#endif /* _POSIX_MAPPED_FILES */
{
/* Read the file in pieces. */
ret = process_fd (cd, fd, &output, output_file);
@@ -544,7 +504,7 @@ process_fd (iconv_t cd, int fd, FILE **output, const char *output_file)
process it in one step. */
static char *inbuf = NULL;
static size_t maxlen = 0;
- char *inptr = NULL;
+ char *inptr = inbuf;
size_t actlen = 0;
while (actlen < maxlen)

62
glibc-RHEL-71530-4.patch Normal file
View File

@ -0,0 +1,62 @@
commit 00ba299787c2ea9e5c4986301e2f4965dffbfded
Author: Florian Weimer <fweimer@redhat.com>
Date: Fri Sep 20 13:10:54 2024 +0200
manual: __is_last is no longer part of iconv internals
The __is_last field was replaced with a bitmask in
commit 85830c4c4688b30d3d76111aa9a26745c7b141d6 in 2000,
and multiple bits are in use today.
Reviewed-by: DJ Delorie <dj@redhat.com>
diff --git a/manual/charset.texi b/manual/charset.texi
index 427db3bc804f6244..3aaa62d088570f76 100644
--- a/manual/charset.texi
+++ b/manual/charset.texi
@@ -2422,11 +2422,11 @@ written into the buffer to signal how much output is available. If this
conversion step is not the last one, the element must not be modified.
The @code{__outbufend} element must not be modified.
-@item int __is_last
-This element is nonzero if this conversion step is the last one. This
-information is necessary for the recursion. See the description of the
-conversion function internals below. This element must never be
-modified.
+@item int __flags
+This field is a set of flags. The @code{__GCONV_IS_LAST} bit is set if
+this conversion step is the last one. This information is necessary for
+the recursion. See the description of the conversion function internals
+below. This element must never be modified.
@item int __invocation_counter
The conversion function can use this element to see how many calls of
@@ -2731,8 +2731,8 @@ Otherwise the function has to emit a byte sequence to bring the state
object into the initial state. Once this all happened the other
conversion modules in the chain of conversions have to get the same
chance. Whether another step follows can be determined from the
-@code{__is_last} element of the step data structure to which the first
-parameter points.
+@code{__GCONV_IS_LAST} flag in the @code{__flags} field of the step
+data structure to which the first parameter points.
The more interesting mode is when actual text has to be converted. The
first step in this case is to convert as much text as possible from the
@@ -2866,7 +2866,7 @@ gconv (struct __gconv_step *step, struct __gconv_step_data *data,
/* @r{Call the steps down the chain if there are any but only}
@r{if we successfully emitted the escape sequence.} */
- if (status == __GCONV_OK && ! data->__is_last)
+ if (status == __GCONV_OK && ! (data->__flags & __GCONV_IS_LAST))
status = fct (next_step, next_data, NULL, NULL,
written, 1);
@}
@@ -2892,7 +2892,7 @@ gconv (struct __gconv_step *step, struct __gconv_step_data *data,
/* @r{If this is the last step, leave the loop. There is}
@r{nothing we can do.} */
- if (data->__is_last)
+ if (data->__flags & __GCONV_IS_LAST)
@{
/* @r{Store information about how many bytes are}
@r{available.} */

1081
glibc-RHEL-71530-5.patch Normal file
View File

File diff suppressed because it is too large Load Diff

713
glibc-RHEL-71530-6.patch Normal file
View File

@ -0,0 +1,713 @@
commit 8ef3cff9d1ceafe369f982d980678d749fb93bd2
Author: Florian Weimer <fweimer@redhat.com>
Date: Fri Sep 20 13:10:54 2024 +0200
iconv: Support in-place conversions (bug 10460, bug 32033)
Check if any of the input files overlaps with the output file, and use
a temporary file in this case, so that the input is no clobbered
before it is read. This fixes bug 10460. It allows to use iconv
more easily as a functional replacement for GNU recode.
The updated output buffer management truncates the output file
if there is no input, fixing bug 32033.
Reviewed-by: DJ Delorie <dj@redhat.com>
diff --git a/iconv/Makefile b/iconv/Makefile
index 29e4f280ec4cdcbf..c9af0c4d44cae7fb 100644
--- a/iconv/Makefile
+++ b/iconv/Makefile
@@ -81,6 +81,8 @@ include $(patsubst %,$(..)libof-iterator.mk,$(cpp-srcs-left))
ifeq ($(run-built-tests),yes)
xtests-special += $(objpfx)test-iconvconfig.out
tests-special += \
+ $(objpfx)tst-iconv_prog-buffer-large.out \
+ $(objpfx)tst-iconv_prog-buffer-tiny.out \
$(objpfx)tst-iconv_prog-buffer.out \
$(objpfx)tst-iconv_prog.out \
$(objpfx)tst-translit-mchar.out \
@@ -153,3 +155,12 @@ $(objpfx)tst-iconv_prog-buffer.out: \
tst-iconv_prog-buffer.sh $(objpfx)iconv_prog
$(BASH) $< $(common-objdir) '$(test-program-prefix)' > $@; \
$(evaluate-test)
+$(objpfx)tst-iconv_prog-buffer-tiny.out: \
+ tst-iconv_prog-buffer.sh $(objpfx)iconv_prog
+ $(BASH) $< $(common-objdir) '$(test-program-prefix)' \
+ '--buffer-size=1' > $@; \
+ $(evaluate-test)
+$(objpfx)tst-iconv_prog-buffer-large.out: \
+ tst-iconv_prog-buffer.sh $(objpfx)iconv_prog
+ $(BASH) $< $(common-objdir) '$(test-program-prefix)' '' '22' > $@; \
+ $(evaluate-test)
diff --git a/iconv/iconv_prog.c b/iconv/iconv_prog.c
index 5fe4fe7a6c3776f4..3e02db7319185d45 100644
--- a/iconv/iconv_prog.c
+++ b/iconv/iconv_prog.c
@@ -47,7 +47,11 @@
static void print_version (FILE *stream, struct argp_state *state);
void (*argp_program_version_hook) (FILE *, struct argp_state *) = print_version;
-#define OPT_VERBOSE 1000
+enum
+ {
+ OPT_VERBOSE = 1000,
+ OPT_BUFFER_SIZE,
+ };
#define OPT_LIST 'l'
/* Definitions of arguments for argp functions. */
@@ -63,6 +67,10 @@ static const struct argp_option options[] =
{ "output", 'o', N_("FILE"), 0, N_("output file") },
{ "silent", 's', NULL, 0, N_("suppress warnings") },
{ "verbose", OPT_VERBOSE, NULL, 0, N_("print progress information") },
+ /* This is an internal option intended for testing only. Very small
+ buffers do not work with all character sets. */
+ { "buffer-size", OPT_BUFFER_SIZE, N_("BYTE-COUNT"), OPTION_HIDDEN,
+ N_("size of in-memory scratch buffer") },
{ NULL, 0, NULL, 0, NULL }
};
@@ -100,13 +108,20 @@ static int list;
/* If nonzero omit invalid character from output. */
int omit_invalid;
+/* Current index in argv (after command line processing) with the
+ input file name. */
+static int current_input_file_index;
+
+/* Size of the temporary, in-memory buffer. Exceeding it needs
+ spooling to disk in a temporary file. Controlled by --buffer_size. */
+static size_t output_buffer_size = 1024 * 1024;
+
/* Prototypes for the functions doing the actual work. */
-static int process_block (iconv_t cd, char *addr, size_t len, FILE **output,
- const char *output_file);
-static int process_fd (iconv_t cd, int fd, FILE **output,
- const char *output_file);
-static int process_file (iconv_t cd, FILE *input, FILE **output,
- const char *output_file);
+static void prepare_output_file (char **argv);
+static void close_output_file (int status);
+static int process_block (iconv_t cd, char *addr, size_t len);
+static int process_fd (iconv_t cd, int fd);
+static int process_file (iconv_t cd, FILE *input);
static void print_known_names (void);
@@ -114,7 +129,6 @@ int
main (int argc, char *argv[])
{
int status = EXIT_SUCCESS;
- int remaining;
__gconv_t cd;
struct charmap_t *from_charmap = NULL;
struct charmap_t *to_charmap = NULL;
@@ -126,7 +140,7 @@ main (int argc, char *argv[])
textdomain (_libc_intl_domainname);
/* Parse and process arguments. */
- argp_parse (&argp, argc, argv, 0, &remaining, NULL);
+ argp_parse (&argp, argc, argv, 0, &current_input_file_index, NULL);
/* List all coded character sets if wanted. */
if (list)
@@ -161,7 +175,8 @@ main (int argc, char *argv[])
if (from_charmap != NULL || to_charmap != NULL)
/* Construct the conversion table and do the conversion. */
status = charmap_conversion (from_code, from_charmap, to_code, to_charmap,
- argc, remaining, argv, output_file);
+ argc, current_input_file_index, argv,
+ output_file);
else
{
struct gconv_spec conv_spec;
@@ -235,16 +250,14 @@ conversions from `%s' and to `%s' are not supported"),
_("failed to start conversion processing"));
}
- /* The output file. Will be opened when we are ready to produce
- output. */
- FILE *output = NULL;
+ prepare_output_file (argv);
/* Now process the remaining files. Write them to stdout or the file
specified with the `-o' parameter. If we have no file given as
the parameter process all from stdin. */
- if (remaining == argc)
+ if (current_input_file_index == argc)
{
- if (process_file (cd, stdin, &output, output_file) != 0)
+ if (process_file (cd, stdin) != 0)
status = EXIT_FAILURE;
}
else
@@ -253,17 +266,17 @@ conversions from `%s' and to `%s' are not supported"),
int fd, ret;
if (verbose)
- fprintf (stderr, "%s:\n", argv[remaining]);
- if (strcmp (argv[remaining], "-") == 0)
- fd = 0;
+ fprintf (stderr, "%s:\n", argv[current_input_file_index]);
+ if (strcmp (argv[current_input_file_index], "-") == 0)
+ fd = STDIN_FILENO;
else
{
- fd = open (argv[remaining], O_RDONLY);
+ fd = open (argv[current_input_file_index], O_RDONLY);
if (fd == -1)
{
error (0, errno, _("cannot open input file `%s'"),
- argv[remaining]);
+ argv[current_input_file_index]);
status = EXIT_FAILURE;
continue;
}
@@ -271,7 +284,7 @@ conversions from `%s' and to `%s' are not supported"),
{
/* Read the file in pieces. */
- ret = process_fd (cd, fd, &output, output_file);
+ ret = process_fd (cd, fd);
/* Now close the file. */
close (fd);
@@ -289,7 +302,7 @@ conversions from `%s' and to `%s' are not supported"),
}
}
}
- while (++remaining < argc);
+ while (++current_input_file_index < argc);
/* Ensure that iconv -c still exits with failure if iconv (the
function) has failed with E2BIG instead of EILSEQ. */
@@ -297,8 +310,7 @@ conversions from `%s' and to `%s' are not supported"),
status = EXIT_FAILURE;
/* Close the output file now. */
- if (output != NULL && fclose (output))
- error (EXIT_FAILURE, errno, _("error while closing output file"));
+ close_output_file (status);
}
return status;
@@ -328,6 +340,14 @@ parse_opt (int key, char *arg, struct argp_state *state)
/* Omit invalid characters from output. */
omit_invalid = 1;
break;
+ case OPT_BUFFER_SIZE:
+ {
+ int i = atoi (arg);
+ if (i <= 0)
+ error (EXIT_FAILURE, 0, _("invalid buffer size: %s"), arg);
+ output_buffer_size = i;
+ }
+ break;
case OPT_VERBOSE:
verbose = 1;
break;
@@ -374,59 +394,247 @@ warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n\
fprintf (stream, gettext ("Written by %s.\n"), "Ulrich Drepper");
}
+/* Command line index of the last input file that overlaps with the
+ output file. Zero means no temporary file is ever required. */
+static int last_overlapping_file_index;
-static int
-write_output (const char *outbuf, const char *outptr, FILE **output,
- const char *output_file)
+/* This is set to true if the output is written to a temporary file. */
+static bool output_using_temporary_file;
+
+/* This is the file descriptor that will be used by write_output. */
+static int output_fd = -1;
+
+/* Pointers at the start and end of the fixed-size output buffer. */
+static char *output_buffer_start;
+
+/* Current write position in the output buffer. */
+static char *output_buffer_current;
+
+/* Remaining bytes after output_buffer_current in the output buffer. */
+static size_t output_buffer_remaining;
+
+
+/* Reduce the buffer size when writing directly to the output file, to
+ reduce cache utilization. */
+static size_t copy_buffer_size = BUFSIZ;
+
+static void
+output_error (void)
+{
+ error (EXIT_FAILURE, errno, _("cannot open output file"));
+}
+
+static void
+input_error (const char *path)
{
- /* We have something to write out. */
- int errno_save = errno;
+ error (0, errno, _("cannot open input file `%s'"), path);
+}
- if (*output == NULL)
+/* Opens output_file for writing, truncating it. */
+static void
+open_output_direct (void)
+{
+ output_fd = open64 (output_file, O_WRONLY | O_CREAT | O_TRUNC, 0777);
+ if (output_fd < 0)
+ output_error ();
+}
+
+static void
+prepare_output_file (char **argv)
+{
+ if (copy_buffer_size > output_buffer_size)
+ copy_buffer_size = output_buffer_size;
+
+ if (output_file == NULL || strcmp (output_file, "-") == 0)
{
- /* Determine output file. */
- if (output_file != NULL && strcmp (output_file, "-") != 0)
+ /* No buffering is required when writing to standard output
+ because input overlap is expected to be solved externally. */
+ output_fd = STDOUT_FILENO;
+ output_buffer_size = copy_buffer_size;
+ }
+ else
+ {
+ /* If iconv creates the output file, no overlap is possible. */
+ output_fd = open64 (output_file, O_WRONLY | O_CREAT | O_EXCL, 0777);
+ if (output_fd >= 0)
+ output_buffer_size = copy_buffer_size;
+ else
{
- *output = fopen (output_file, "w");
- if (*output == NULL)
- error (EXIT_FAILURE, errno, _("cannot open output file"));
+ /* Otherwise, check if any of the input files overlap with the
+ output file. */
+ struct statx st;
+ if (statx (AT_FDCWD, output_file, 0, STATX_INO | STATX_MODE, &st)
+ != 0)
+ output_error ();
+ uint32_t out_dev_minor = st.stx_dev_minor;
+ uint32_t out_dev_major = st.stx_dev_major;
+ uint64_t out_ino = st.stx_ino;
+
+ int idx = current_input_file_index;
+ while (true)
+ {
+ /* Special case: no input files means standard input. */
+ if (argv[idx] == NULL && idx != current_input_file_index)
+ break;
+
+ int ret;
+ if (argv[idx] == NULL || strcmp (argv[idx], "-") == 0)
+ ret = statx (STDIN_FILENO, "", AT_EMPTY_PATH, STATX_INO, &st);
+ else
+ ret = statx (AT_FDCWD, argv[idx], 0, STATX_INO, &st);
+ if (ret != 0)
+ {
+ input_error (argv[idx]);
+ exit (EXIT_FAILURE);
+ }
+ if (out_dev_minor == st.stx_dev_minor
+ && out_dev_major == st.stx_dev_major
+ && out_ino == st.stx_ino)
+ {
+ if (argv[idx] == NULL)
+ /* Corner case: index of NULL would be larger than
+ idx while converting, triggering a switch away
+ from the temporary file. */
+ last_overlapping_file_index = INT_MAX;
+ else
+ last_overlapping_file_index = idx;
+ }
+
+ if (argv[idx] == NULL)
+ break;
+ ++idx;
+ }
+
+ /* If there is no overlap, avoid using a temporary file. */
+ if (last_overlapping_file_index == 0)
+ {
+ open_output_direct ();
+ output_buffer_size = copy_buffer_size;
+ }
}
- else
- *output = stdout;
}
- if (fwrite (outbuf, 1, outptr - outbuf, *output) < (size_t) (outptr - outbuf)
- || ferror (*output))
+ output_buffer_start = malloc (output_buffer_size);
+ if (output_buffer_start == NULL)
+ output_error ();
+ output_buffer_current = output_buffer_start;
+ output_buffer_remaining = output_buffer_size;
+}
+
+/* Write out the range [first, last), terminating the process on write
+ error. */
+static void
+write_fully (int fd, const char *first, const char *last)
+{
+ while (first < last)
{
- /* Error occurred while printing the result. */
- error (0, 0, _("\
+ ssize_t ret = write (fd, first, last - first);
+ if (ret == 0)
+ {
+ errno = ENOSPC;
+ output_error ();
+ }
+ if (ret < 0)
+ error (EXIT_FAILURE, errno, _("\
conversion stopped due to problem in writing the output"));
- return -1;
+ first += ret;
+ }
+}
+
+static void
+flush_output (void)
+{
+ bool temporary_file_not_needed
+ = current_input_file_index > last_overlapping_file_index;
+ if (output_fd < 0)
+ {
+ if (temporary_file_not_needed)
+ open_output_direct ();
+ else
+ {
+ /* Create an anonymous temporary file. */
+ FILE *fp = tmpfile ();
+ if (fp == NULL)
+ output_error ();
+ output_fd = dup (fileno (fp));
+ if (output_fd < 0)
+ output_error ();
+ fclose (fp);
+ output_using_temporary_file = true;
+ }
+ /* Either way, no longer use a memory-only staging buffer. */
+ output_buffer_size = copy_buffer_size;
}
+ else if (output_using_temporary_file && temporary_file_not_needed)
+ {
+ /* The temporary file is no longer needed. Switch to direct
+ output, replacing output_fd. */
+ int temp_fd = output_fd;
+ open_output_direct ();
+
+ /* Copy over the data spooled to the temporary file. */
+ if (lseek (temp_fd, 0, SEEK_SET) < 0)
+ output_error ();
+ while (true)
+ {
+ char buf[BUFSIZ];
+ ssize_t ret = read (temp_fd, buf, sizeof (buf));
+ if (ret < 0)
+ output_error ();
+ if (ret == 0)
+ break;
+ write_fully (output_fd, buf, buf + ret);
+ }
+ close (temp_fd);
- errno = errno_save;
+ /* No longer using a temporary file from now on. */
+ output_using_temporary_file = false;
+ output_buffer_size = copy_buffer_size;
+ }
- return 0;
+ write_fully (output_fd, output_buffer_start, output_buffer_current);
+ output_buffer_current = output_buffer_start;
+ output_buffer_remaining = output_buffer_size;
}
+static void
+close_output_file (int status)
+{
+ /* Do not perform a flush if a temporary file or the in-memory
+ buffer is in use and there was an error. It would clobber the
+ overlapping input file. */
+ if (status != EXIT_SUCCESS && !omit_invalid &&
+ (output_using_temporary_file || output_fd < 0))
+ return;
+
+ /* The current_input_file_index variable is now larger than
+ last_overlapping_file_index, so the flush_output call switches
+ away from the temporary file. */
+ flush_output ();
+
+ if (output_fd == STDOUT_FILENO)
+ {
+ /* Close standard output in safe manner, to report certain
+ ENOSPC errors. */
+ output_fd = dup (output_fd);
+ if (output_fd < 0)
+ output_error ();
+ }
+ if (close (output_fd) < 0)
+ output_error ();
+}
static int
-process_block (iconv_t cd, char *addr, size_t len, FILE **output,
- const char *output_file)
+process_block (iconv_t cd, char *addr, size_t len)
{
-#define OUTBUF_SIZE 32768
const char *start = addr;
- char outbuf[OUTBUF_SIZE];
- char *outptr;
- size_t outlen;
size_t n;
int ret = 0;
while (len > 0)
{
- outptr = outbuf;
- outlen = OUTBUF_SIZE;
- n = iconv (cd, &addr, &len, &outptr, &outlen);
+ n = iconv (cd, &addr, &len,
+ &output_buffer_current, &output_buffer_remaining);
if (n == (size_t) -1 && omit_invalid && errno == EILSEQ)
{
@@ -437,39 +645,34 @@ process_block (iconv_t cd, char *addr, size_t len, FILE **output,
errno = E2BIG;
}
- if (outptr != outbuf)
- {
- ret = write_output (outbuf, outptr, output, output_file);
- if (ret != 0)
- break;
- }
-
if (n != (size_t) -1)
{
/* All the input test is processed. For state-dependent
character sets we have to flush the state now. */
- outptr = outbuf;
- outlen = OUTBUF_SIZE;
- n = iconv (cd, NULL, NULL, &outptr, &outlen);
-
- if (outptr != outbuf)
+ n = iconv (cd, NULL, NULL,
+ &output_buffer_current, &output_buffer_remaining);
+ if (n == (size_t) -1 && errno == E2BIG)
{
- ret = write_output (outbuf, outptr, output, output_file);
- if (ret != 0)
- break;
+ /* Try again if the state flush exceeded the buffer space. */
+ flush_output ();
+ n = iconv (cd, NULL, NULL,
+ &output_buffer_current, &output_buffer_remaining);
}
+ bool errno_is_EILSEQ = errno == EILSEQ;
if (n != (size_t) -1)
break;
- if (omit_invalid && errno == EILSEQ)
+ if (omit_invalid && errno_is_EILSEQ)
{
ret = 1;
break;
}
}
- if (errno != E2BIG)
+ if (errno == E2BIG)
+ flush_output ();
+ else
{
/* iconv() ran into a problem. */
switch (errno)
@@ -500,7 +703,7 @@ incomplete character or shift sequence at end of buffer"));
static int
-process_fd (iconv_t cd, int fd, FILE **output, const char *output_file)
+process_fd (iconv_t cd, int fd)
{
/* we have a problem with reading from a descriptor since we must not
provide the iconv() function an incomplete character or shift
@@ -574,16 +777,16 @@ process_fd (iconv_t cd, int fd, FILE **output, const char *output_file)
}
/* Now we have all the input in the buffer. Process it in one run. */
- return process_block (cd, inbuf, actlen, output, output_file);
+ return process_block (cd, inbuf, actlen);
}
static int
-process_file (iconv_t cd, FILE *input, FILE **output, const char *output_file)
+process_file (iconv_t cd, FILE *input)
{
/* This should be safe since we use this function only for `stdin' and
we haven't read anything so far. */
- return process_fd (cd, fileno (input), output, output_file);
+ return process_fd (cd, fileno (input));
}
diff --git a/iconv/tst-iconv_prog-buffer.sh b/iconv/tst-iconv_prog-buffer.sh
index 5ff99a02a30370cb..54ff871d32929997 100644
--- a/iconv/tst-iconv_prog-buffer.sh
+++ b/iconv/tst-iconv_prog-buffer.sh
@@ -17,6 +17,12 @@
# License along with the GNU C Library; if not, see
# <https://www.gnu.org/licenses/>.
+# Arguments:
+# root of the build tree ($(objpfx-common))
+# test command wrapper (for running on the board/with new ld.so)
+# extra flags to pass to iconv
+# number of times to double the input files in size (default: 0)
+
exec 2>&1
set -e
@@ -26,7 +32,9 @@ codir=$1
test_program_prefix="$2"
# Use internal converters to avoid issues with module loading.
-iconv_args="-f ASCII -t UTF-8"
+iconv_args="-f ASCII -t UTF-8 $3"
+
+file_size_doublings=${4-0}
failure=false
@@ -39,7 +47,19 @@ echo HH > "$tmp/hh"
echo XY > "$tmp/xy"
echo ZT > "$tmp/zt"
echo OUT > "$tmp/out-template"
+: > "$tmp/empty"
printf '\xff' > "$tmp/0xff"
+
+# Double all files to produce larger buffers.
+for p in "$tmp"/* ; do
+ i=0
+ while test $i -lt $file_size_doublings; do
+ cat "$p" "$p" > "$tmp/scratch"
+ mv "$tmp/scratch" "$p"
+ i=$(($i + 1))
+ done
+done
+
cat "$tmp/xy" "$tmp/0xff" "$tmp/zt" > "$tmp/0xff-wrapped"
run_iconv () {
@@ -113,6 +133,38 @@ expect_files abc def
run_iconv -o "$tmp/out" "$tmp/out" "$tmp/abc"
expect_files abc def abc
+run_iconv -o "$tmp/out" "$tmp/ggg" "$tmp/out"
+expect_files ggg abc def abc
+
+run_iconv -o "$tmp/out" "$tmp/hh" "$tmp/out" "$tmp/hh"
+expect_files hh ggg abc def abc hh
+
+cp "$tmp/out-template" "$tmp/out"
+run_iconv -o "$tmp/out" "$tmp/ggg" "$tmp/out" "$tmp/out" "$tmp/ggg"
+expect_files ggg out-template out-template ggg
+
+cp "$tmp/out-template" "$tmp/out"
+run_iconv -o "$tmp/out" "$tmp/ggg" "$tmp/out" "$tmp/hh" "$tmp/out" "$tmp/ggg"
+expect_files ggg out-template hh out-template ggg
+
+# Empty output should truncate the output file if exists.
+
+cp "$tmp/out-template" "$tmp/out"
+run_iconv -o "$tmp/out" </dev/null
+expect_files empty
+
+cp "$tmp/out-template" "$tmp/out"
+run_iconv -o "$tmp/out" - </dev/null
+expect_files empty
+
+cp "$tmp/out-template" "$tmp/out"
+run_iconv -o "$tmp/out" /dev/null
+expect_files empty
+
+cp "$tmp/out-template" "$tmp/out"
+expect_exit 1 run_iconv -c -o "$tmp/out" "$tmp/0xff"
+expect_files empty
+
# But not if we are writing to standard output.
cp "$tmp/out-template" "$tmp/out"
@@ -142,8 +194,36 @@ cp "$tmp/0xff" "$tmp/out"
expect_exit 1 run_iconv -o "$tmp/out" - < "$tmp/out"
expect_files 0xff
+cp "$tmp/0xff-wrapped" "$tmp/out"
+expect_exit 1 run_iconv -o "$tmp/out" "$tmp/out"
+expect_files 0xff-wrapped
+
+cp "$tmp/0xff-wrapped" "$tmp/out"
+expect_exit 1 run_iconv -o "$tmp/out" < "$tmp/out"
+expect_files 0xff-wrapped
+
+cp "$tmp/0xff-wrapped" "$tmp/out"
+expect_exit 1 run_iconv -o "$tmp/out" - < "$tmp/out"
+expect_files 0xff-wrapped
+
+cp "$tmp/0xff-wrapped" "$tmp/out"
+expect_exit 1 run_iconv -o "$tmp/out" "$tmp/abc" "$tmp/out"
+expect_files 0xff-wrapped
+
+cp "$tmp/0xff-wrapped" "$tmp/out"
+expect_exit 1 run_iconv -o "$tmp/out" "$tmp/abc" - < "$tmp/out"
+expect_files 0xff-wrapped
+
# If errors are ignored, the file should be overwritten.
+cp "$tmp/0xff-wrapped" "$tmp/out"
+expect_exit 1 run_iconv -c -o "$tmp/out" "$tmp/out"
+expect_files xy zt
+
+cp "$tmp/0xff" "$tmp/out"
+expect_exit 1 run_iconv -c -o "$tmp/out" "$tmp/abc" "$tmp/out" "$tmp/def"
+expect_files abc def
+
cp "$tmp/out-template" "$tmp/out"
expect_exit 1 \
run_iconv -c -o "$tmp/out" "$tmp/abc" "$tmp/0xff" "$tmp/def" 2>"$tmp/err"
@@ -156,6 +236,20 @@ expect_exit 1 run_iconv -c -o "$tmp/out" \
! test -s "$tmp/err"
expect_files abc xy zt def
+cp "$tmp/0xff-wrapped" "$tmp/out"
+expect_exit 1 run_iconv -c -o "$tmp/out" "$tmp/out" "$tmp/abc" "$tmp/out" "$tmp/def"
+expect_files xy zt abc xy zt def
+
+cp "$tmp/0xff-wrapped" "$tmp/out"
+expect_exit 1 run_iconv -o "$tmp/out" \
+ "$tmp/out" "$tmp/abc" "$tmp/out" "$tmp/def"
+expect_files 0xff-wrapped
+
+cp "$tmp/0xff-wrapped" "$tmp/out"
+expect_exit 1 run_iconv -c -o "$tmp/out" \
+ "$tmp/abc" "$tmp/out" "$tmp/def" "$tmp/out"
+expect_files abc xy zt def xy zt
+
# If the file does not exist yet, it should not be created on error.
rm "$tmp/out"

41
glibc-RHEL-71530-7.patch Normal file
View File

@ -0,0 +1,41 @@
commit 75819cdd29a193cc2db980878bec305905b22bbc
Author: Florian Weimer <fweimer@redhat.com>
Date: Fri Sep 20 13:10:54 2024 +0200
iconv: Multiple - on command line should not fail (bug 32050)
Usually, the second and subsequent - return EOF immediately
and do not contribute to the output, but this is not an error.
Reviewed-by: DJ Delorie <dj@redhat.com>
diff --git a/iconv/iconv_prog.c b/iconv/iconv_prog.c
index 3e02db7319185d45..dd4bc3a59a20799a 100644
--- a/iconv/iconv_prog.c
+++ b/iconv/iconv_prog.c
@@ -287,7 +287,8 @@ conversions from `%s' and to `%s' are not supported"),
ret = process_fd (cd, fd);
/* Now close the file. */
- close (fd);
+ if (fd != STDIN_FILENO)
+ close (fd);
if (ret != 0)
{
diff --git a/iconv/tst-iconv_prog-buffer.sh b/iconv/tst-iconv_prog-buffer.sh
index 54ff871d32929997..a9c3729d948b4679 100644
--- a/iconv/tst-iconv_prog-buffer.sh
+++ b/iconv/tst-iconv_prog-buffer.sh
@@ -265,6 +265,11 @@ expect_exit 1 run_iconv -o "$tmp/out" "$tmp/abc" "$tmp/0xff" "$tmp/def"
expect_exit 1 run_iconv -o "$tmp/out" "$tmp/abc" - < "$tmp/0xff" "$tmp/def"
! test -e "$tmp/out"
+# Listing standard input multiple times should not fail (bug 32050).
+
+run_iconv -o "$tmp/out" "$tmp/xy" - - "$tmp/zt" < "$tmp/abc"
+expect_files xy abc zt
+
if $failure ; then
exit 1
fi

323
glibc-RHEL-71530-8.patch Normal file
View File

@ -0,0 +1,323 @@
commit fa1b0d5e9f6e0353e16339430770a7a8824c0468
Author: Florian Weimer <fweimer@redhat.com>
Date: Fri Sep 20 13:10:54 2024 +0200
iconv: Input buffering for the iconv program (bug 6050)
Do not read the entire input file into memory.
Reviewed-by: DJ Delorie <dj@redhat.com>
diff --git a/iconv/iconv_prog.c b/iconv/iconv_prog.c
index dd4bc3a59a20799a..a2f1d34e4579f80f 100644
--- a/iconv/iconv_prog.c
+++ b/iconv/iconv_prog.c
@@ -118,8 +118,9 @@ static size_t output_buffer_size = 1024 * 1024;
/* Prototypes for the functions doing the actual work. */
static void prepare_output_file (char **argv);
-static void close_output_file (int status);
-static int process_block (iconv_t cd, char *addr, size_t len);
+static void close_output_file (__gconv_t cd, int status);
+static int process_block (iconv_t cd, char **addr, size_t *len,
+ off64_t file_offset, bool *incomplete);
static int process_fd (iconv_t cd, int fd);
static int process_file (iconv_t cd, FILE *input);
static void print_known_names (void);
@@ -311,7 +312,7 @@ conversions from `%s' and to `%s' are not supported"),
status = EXIT_FAILURE;
/* Close the output file now. */
- close_output_file (status);
+ close_output_file (cd, status);
}
return status;
@@ -599,7 +600,7 @@ flush_output (void)
}
static void
-close_output_file (int status)
+close_output_file (__gconv_t cd, int status)
{
/* Do not perform a flush if a temporary file or the in-memory
buffer is in use and there was an error. It would clobber the
@@ -608,10 +609,28 @@ close_output_file (int status)
(output_using_temporary_file || output_fd < 0))
return;
- /* The current_input_file_index variable is now larger than
- last_overlapping_file_index, so the flush_output call switches
+ /* All the input text is processed. For state-dependent character
+ sets we have to flush the state now.
+
+ The current_input_file_index variable is now larger than
+ last_overlapping_file_index, so the flush_output calls switch
away from the temporary file. */
+ size_t n = iconv (cd, NULL, NULL,
+ &output_buffer_current, &output_buffer_remaining);
+ if (n == (size_t) -1 && errno == E2BIG)
+ {
+ /* Try again if the state flush exceeded the buffer space. */
+ flush_output ();
+ n = iconv (cd, NULL, NULL,
+ &output_buffer_current, &output_buffer_remaining);
+ }
+ int saved_errno = errno;
flush_output ();
+ if (n == (size_t) -1 && !omit_invalid)
+ {
+ errno = saved_errno;
+ output_error ();
+ }
if (output_fd == STDOUT_FILENO)
{
@@ -625,51 +644,35 @@ close_output_file (int status)
output_error ();
}
+/* CD is the iconv handle. Input processing starts at *ADDR, and
+ consumes upto *LEN bytes. *ADDR and *LEN are updated. FILE_OFFSET
+ is the file offset of the data initially at ADDR. *INCOMPLETE is
+ set to true if conversion stops due to an incomplete input
+ sequence. */
static int
-process_block (iconv_t cd, char *addr, size_t len)
+process_block (iconv_t cd, char **addr, size_t *len, off64_t file_offset,
+ bool *incomplete)
{
- const char *start = addr;
+ const char *start = *addr;
size_t n;
int ret = 0;
- while (len > 0)
+ while (*len > 0)
{
- n = iconv (cd, &addr, &len,
+ n = iconv (cd, addr, len,
&output_buffer_current, &output_buffer_remaining);
if (n == (size_t) -1 && omit_invalid && errno == EILSEQ)
{
ret = 1;
- if (len == 0)
+ if (*len == 0)
n = 0;
else
errno = E2BIG;
}
if (n != (size_t) -1)
- {
- /* All the input test is processed. For state-dependent
- character sets we have to flush the state now. */
- n = iconv (cd, NULL, NULL,
- &output_buffer_current, &output_buffer_remaining);
- if (n == (size_t) -1 && errno == E2BIG)
- {
- /* Try again if the state flush exceeded the buffer space. */
- flush_output ();
- n = iconv (cd, NULL, NULL,
- &output_buffer_current, &output_buffer_remaining);
- }
- bool errno_is_EILSEQ = errno == EILSEQ;
-
- if (n != (size_t) -1)
- break;
-
- if (omit_invalid && errno_is_EILSEQ)
- {
- ret = 1;
- break;
- }
- }
+ break;
if (errno == E2BIG)
flush_output ();
@@ -680,13 +683,12 @@ process_block (iconv_t cd, char *addr, size_t len)
{
case EILSEQ:
if (! omit_invalid)
- error (0, 0, _("illegal input sequence at position %ld"),
- (long int) (addr - start));
+ error (0, 0, _("illegal input sequence at position %lld"),
+ (long long int) (file_offset + (*addr - start)));
break;
case EINVAL:
- error (0, 0, _("\
-incomplete character or shift sequence at end of buffer"));
- break;
+ *incomplete = true;
+ return ret;
case EBADF:
error (0, 0, _("internal error (illegal descriptor)"));
break;
@@ -706,79 +708,49 @@ incomplete character or shift sequence at end of buffer"));
static int
process_fd (iconv_t cd, int fd)
{
- /* we have a problem with reading from a descriptor since we must not
- provide the iconv() function an incomplete character or shift
- sequence at the end of the buffer. Since we have to deal with
- arbitrary encodings we must read the whole text in a buffer and
- process it in one step. */
- static char *inbuf = NULL;
- static size_t maxlen = 0;
- char *inptr = inbuf;
- size_t actlen = 0;
-
- while (actlen < maxlen)
+ char inbuf[BUFSIZ];
+ char *inbuf_end = inbuf + sizeof (inbuf);
+ size_t inbuf_used = 0;
+ off64_t file_offset = 0;
+ int status = 0;
+ bool incomplete = false;
+
+ while (true)
{
- ssize_t n = read (fd, inptr, maxlen - actlen);
-
- if (n == 0)
- /* No more text to read. */
- break;
-
- if (n == -1)
+ char *p = inbuf + inbuf_used;
+ ssize_t read_ret = read (fd, p, inbuf_end - p);
+ if (read_ret == 0)
+ {
+ /* On EOF, check if the previous iconv invocation saw an
+ incomplete sequence. */
+ if (incomplete)
+ {
+ error (0, 0, _("\
+incomplete character or shift sequence at end of buffer"));
+ return 1;
+ }
+ return 0;
+ }
+ if (read_ret < 0)
{
- /* Error while reading. */
error (0, errno, _("error while reading the input"));
return -1;
}
-
- inptr += n;
- actlen += n;
+ inbuf_used += read_ret;
+ incomplete = false;
+ p = inbuf;
+ int ret = process_block (cd, &p, &inbuf_used, file_offset, &incomplete);
+ if (ret != 0)
+ {
+ status = ret;
+ if (ret < 0)
+ break;
+ }
+ /* The next loop iteration consumes the leftover bytes. */
+ memmove (inbuf, p, inbuf_used);
+ file_offset += read_ret - inbuf_used;
}
-
- if (actlen == maxlen)
- while (1)
- {
- ssize_t n;
- char *new_inbuf;
-
- /* Increase the buffer. */
- new_inbuf = (char *) realloc (inbuf, maxlen + 32768);
- if (new_inbuf == NULL)
- {
- error (0, errno, _("unable to allocate buffer for input"));
- return -1;
- }
- inbuf = new_inbuf;
- maxlen += 32768;
- inptr = inbuf + actlen;
-
- do
- {
- n = read (fd, inptr, maxlen - actlen);
-
- if (n == 0)
- /* No more text to read. */
- break;
-
- if (n == -1)
- {
- /* Error while reading. */
- error (0, errno, _("error while reading the input"));
- return -1;
- }
-
- inptr += n;
- actlen += n;
- }
- while (actlen < maxlen);
-
- if (n == 0)
- /* Break again so we leave both loops. */
- break;
- }
-
- /* Now we have all the input in the buffer. Process it in one run. */
- return process_block (cd, inbuf, actlen);
+ return status;
}
diff --git a/iconv/tst-iconv_prog-buffer.sh b/iconv/tst-iconv_prog-buffer.sh
index a9c3729d948b4679..23098ac56a344c48 100644
--- a/iconv/tst-iconv_prog-buffer.sh
+++ b/iconv/tst-iconv_prog-buffer.sh
@@ -50,6 +50,9 @@ echo OUT > "$tmp/out-template"
: > "$tmp/empty"
printf '\xff' > "$tmp/0xff"
+# Length should be a prime number, to help with buffer alignment testing.
+printf '\xc3\xa4\xe2\x80\x94\xe2\x80\x94\xc3\xa4\n' > "$tmp/utf8-sequence"
+
# Double all files to produce larger buffers.
for p in "$tmp"/* ; do
i=0
@@ -270,6 +273,34 @@ expect_exit 1 run_iconv -o "$tmp/out" "$tmp/abc" - < "$tmp/0xff" "$tmp/def"
run_iconv -o "$tmp/out" "$tmp/xy" - - "$tmp/zt" < "$tmp/abc"
expect_files xy abc zt
+# NB: Extra iconv args are ignored after this point. Actual
+# multi-byte conversion does not work with tiny buffers.
+iconv_args="-f UTF-8 -t ASCII"
+
+printf 'x\n\xc3' > "$tmp/incomplete"
+expect_exit 1 run_iconv -o "$tmp/out" "$tmp/incomplete"
+check_out <<EOF
+x
+EOF
+
+# Test buffering behavior if the buffer ends with an incomplete
+# multi-byte sequence.
+prefix=""
+prefix_length=0
+while test $prefix_length -lt 12; do
+ echo "info: testing prefix length $prefix_length" 2>&$logfd
+ printf "%s" "$prefix" > "$tmp/prefix"
+ cat "$tmp/prefix" "$tmp/utf8-sequence" > "$tmp/tmp"
+ iconv_args="-f UTF-8 -t UCS-4"
+ run_iconv -o "$tmp/out1" "$tmp/tmp"
+ iconv_args="-f UCS-4 -t UTF-8"
+ run_iconv -o "$tmp/out" "$tmp/out1"
+ expect_files prefix utf8-sequence
+
+ prefix="$prefix@"
+ prefix_length=$(($prefix_length + 1))
+done
+
if $failure ; then
exit 1
fi

37
glibc-RHEL-71530-9.patch Normal file
View File

@ -0,0 +1,37 @@
commit 079ebf7624e7fd0ad7fe94a7176a2e132c996d86
Author: Florian Weimer <fweimer@redhat.com>
Date: Tue Sep 24 10:41:35 2024 +0200
iconv: Use $(run-program-prefix) for running iconv (bug 32197)
With --enable-hardcoded-path-in-tests, $(test-program-prefix)
does not redirect to the built glibc, but we need to run
iconv (the program) against the built glibc even with
--enable-hardcoded-path-in-tests, as it is using the ABI
path for the dynamic linker (as an installed program).
Use $(run-program-prefix) instead.
Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
diff --git a/iconv/Makefile b/iconv/Makefile
index c9af0c4d44cae7fb..de9d964ed3c762bf 100644
--- a/iconv/Makefile
+++ b/iconv/Makefile
@@ -153,14 +153,14 @@ $(objpfx)tst-translit-mchar.out: tst-translit-mchar.sh \
$(objpfx)tst-iconv_prog-buffer.out: \
tst-iconv_prog-buffer.sh $(objpfx)iconv_prog
- $(BASH) $< $(common-objdir) '$(test-program-prefix)' > $@; \
+ $(BASH) $< $(common-objdir) '$(run-program-prefix)' > $@; \
$(evaluate-test)
$(objpfx)tst-iconv_prog-buffer-tiny.out: \
tst-iconv_prog-buffer.sh $(objpfx)iconv_prog
- $(BASH) $< $(common-objdir) '$(test-program-prefix)' \
+ $(BASH) $< $(common-objdir) '$(run-program-prefix)' \
'--buffer-size=1' > $@; \
$(evaluate-test)
$(objpfx)tst-iconv_prog-buffer-large.out: \
tst-iconv_prog-buffer.sh $(objpfx)iconv_prog
- $(BASH) $< $(common-objdir) '$(test-program-prefix)' '' '22' > $@; \
+ $(BASH) $< $(common-objdir) '$(run-program-prefix)' '' '22' > $@; \
$(evaluate-test)

15
glibc.spec
View File

@ -145,7 +145,7 @@ Version: %{glibcversion}
# - It allows using the Release number without the %%dist tag in the dependency # - It allows using the Release number without the %%dist tag in the dependency
# generator to make the generated requires interchangeable between Rawhide # generator to make the generated requires interchangeable between Rawhide
# and ELN (.elnYY < .fcXX). # and ELN (.elnYY < .fcXX).
%global baserelease 32 %global baserelease 33
Release: %{baserelease}%{?dist} Release: %{baserelease}%{?dist}
# Licenses: # Licenses:
@ -485,6 +485,16 @@ Patch167: glibc-upstream-2.39-137.patch
Patch168: glibc-RHEL-12867-2.patch Patch168: glibc-RHEL-12867-2.patch
Patch169: glibc-RHEL-12867-3.patch Patch169: glibc-RHEL-12867-3.patch
Patch170: glibc-RHEL-42410.patch Patch170: glibc-RHEL-42410.patch
Patch171: glibc-RHEL-71530-1.patch
Patch172: glibc-RHEL-71530-2.patch
Patch173: glibc-RHEL-71530-3.patch
Patch174: glibc-RHEL-71530-4.patch
Patch175: glibc-RHEL-71530-5.patch
Patch176: glibc-RHEL-71530-6.patch
Patch177: glibc-RHEL-71530-7.patch
Patch178: glibc-RHEL-71530-8.patch
Patch179: glibc-RHEL-71530-9.patch
Patch180: glibc-RHEL-71530-10.patch
############################################################################## ##############################################################################
# Continued list of core "glibc" package information: # Continued list of core "glibc" package information:
@ -2480,6 +2490,9 @@ update_gconv_modules_cache ()
%endif %endif
%changelog %changelog
* Mon Dec 23 2024 Florian Weimer <fweimer@redhat.com> - 2.39-33
- Support in-place file conversion in the iconv tool (RHEL-71530)
* Mon Dec 16 2024 Florian Weimer <fweimer@redhat.com> - 2.39-32 * Mon Dec 16 2024 Florian Weimer <fweimer@redhat.com> - 2.39-32
- Make getenv thread-safe in more cases (RHEL-42410) - Make getenv thread-safe in more cases (RHEL-42410)