Build of app-editors/pluma-2.24.1 fails on itstool crash. Further investigation had shown that a segmentation fault was caused by a known libxml2 bug. Reproducible: Always Steps to Reproduce: 1. Build app-editors/pluma-2.24.1 Actual Results: Build fails on a segmentation fault in itstool. Expected Results: Pluma 2.24.1 built.
Created attachment 662872 [details, diff] Patch used in Red Hat and Debian builds
https://gitlab.gnome.org/GNOME/libxml2/-/issues/187 seems to be a similar but different issue. In both cases malformed document causes segfault, however this problem is with structure and not broken unicode. Looking at the same file in libxml2 source, I can see a bunch of places where strings are passed from vsnprintf() to Python with no checks. In at least two instances null termination of vsnprintf() results also has off-by-one errors in truncation detection. Fortunately, the error is to the safe side, however that truncation potentially can produce _another_ invalid unicode from valid one.
I can confirm this behaviour. I applied the patch from https://gitweb.gentoo.org/repo/gentoo.git/plain/dev-libs/libxml2/files/2.9.9-python3-unicode-errors.patch?id=47c1fed5929fd9633e535c9da15d34c1f09d065a and it solved the problem.
So that patch (the same as https://bugs.gentoo.org/attachment.cgi?id=662872) was already there but dropped in https://gitweb.gentoo.org/repo/gentoo.git/commit/dev-libs/libxml2/files?id=f016ee441ef5590db02e09198fcbf6b12a463c50 Why?
I can re-apply the old patch, but is this *actually* reported upstream?
(In reply to Sam James from comment #5) > I can re-apply the old patch, but is this *actually* reported upstream? Ah: https://gitlab.gnome.org/GNOME/libxml2/-/issues/64. (https://745162.bugs.gentoo.org/attachment.cgi?id=662872)
The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=da4038c33b2c7684f5766d6e8f1d1089e863e87c commit da4038c33b2c7684f5766d6e8f1d1089e863e87c Author: Sam James <sam@gentoo.org> AuthorDate: 2020-12-03 00:13:24 +0000 Commit: Sam James <sam@gentoo.org> CommitDate: 2020-12-03 00:14:49 +0000 dev-libs/libxml2: restore itstool-segfault patch We stopped applying this patch during a roll of a new patchset (my fault), but it seems to still be needed. Noticed when building some of MATE. Bug: https://bugs.gentoo.org/745162 Package-Manager: Portage-3.0.9, Repoman-3.0.2 Signed-off-by: Sam James <sam@gentoo.org> .../libxml2-2.9.8-python3-unicode-errors.patch | 34 ++++++++++++++++++++++ ...2-2.9.10-r3.ebuild => libxml2-2.9.10-r4.ebuild} | 3 ++ 2 files changed, 37 insertions(+)
*** Bug 744739 has been marked as a duplicate of this bug. ***
*** Bug 734968 has been marked as a duplicate of this bug. ***
*** Bug 745135 has been marked as a duplicate of this bug. ***
I'm having this issue with libxml2-2.10.3. Is the patch no longer being applied to this version of libxml2?
(In reply to gsra99 from comment #11) > I'm having this issue with libxml2-2.10.3. Is the patch no longer being > applied to this version of libxml2? When trying to build something in the tree? Or are you calling itstool yourself or is some program you work with using itstool?
(In reply to Kobboi from comment #12) > (In reply to gsra99 from comment #11) > > I'm having this issue with libxml2-2.10.3. Is the patch no longer being > > applied to this version of libxml2? > > When trying to build something in the tree? Or are you calling itstool > yourself or is some program you work with using itstool? I am trying build and install pluma-1.24.2. I get the same error as mentioned for pluma-1.24.1. However libxml2 is now at v.2.10 rather than 2.9.
This is back again, it seems, so reopening. I managed to construct a minimalish reproducer about 6 months ago but it then succeeded on a modern libxml2 :( https://forums.gentoo.org/viewtopic-t-1159343.html mentions a test case which may be useful for debugging (I can't try it right now).
*** Bug 868630 has been marked as a duplicate of this bug. ***
*** Bug 878183 has been marked as a duplicate of this bug. ***
*** Bug 878179 has been marked as a duplicate of this bug. ***
The old patch still works. building libxml2 with the patch, I was able to compile pluma and mate-applets.
The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=bee4fbd32b0225d09bf7fca8d38b83a9ac368bfd commit bee4fbd32b0225d09bf7fca8d38b83a9ac368bfd Author: David Seifert <soap@gentoo.org> AuthorDate: 2022-11-29 19:58:44 +0000 Commit: David Seifert <soap@gentoo.org> CommitDate: 2022-11-29 19:58:44 +0000 dev-libs/libxml2: add workaround patch for itstool breakage Bug: https://bugs.gentoo.org/745162 Signed-off-by: David Seifert <soap@gentoo.org> .../libxml2-2.10.3-python3-unicode-errors.patch | 35 ++++++++++++++++++++++ ...xml2-2.10.3.ebuild => libxml2-2.10.3-r1.ebuild} | 2 ++ 2 files changed, 37 insertions(+)
Created attachment 860557 [details] broken-libxml2.tar.xz In an old chroot (2019-10-30) I created, I get: ``` /var/tmp/portage/app-office/gnumeric-1.12.44/work/gnumeric-1.12.44/doc # itstool -m "/var/tmp/portage/app-office/gnumeric-1.12.44/work/gnumeric-1.12.44/doc/cs/cs.mo" C/gnumeric.xml Warning: Could not merge cs translation for msgid: b"Using <inlineequation> <_:alt-1/> <_:mathphrase-2/> </inlineequation> as the z-value of the <inlineequation> <_:alt-3/> <_:mathphrase-4/> </inlineequation> percentile of the standard normal distribution, set the initial estimate of the number of iterations required as the smallest integer <inlineequation> <_:alt-5/> <_:mathphrase-6/> </inlineequation> such that <_:equation-7/>. Note that if <inlineequation> <_:alt-8/> <_:mathphrase-9/> </inlineequation> is small, it would be more appropriate to use the student's t-distribution of <inlineequation> <_:alt-10/> <_:mathphrase-11/> </inlineequation> instead of <inlineequation> <_:alt-12/> <_:mathphrase-13/> </inlineequation>." Segmentation fault (core dumped) ``` I also ended up minimising a reproducer derived from itstool, but it doesn't reproduce w/ latest libxml2, while the crash does still happen (not for gnumeric but for the MATE stuff mentioned above on forums: https://forums.gentoo.org/viewtopic-t-1159343.html) without the patch w/ latest. This implies that my reduced itstool was too specific. Attaching a smaller version of the forums case, but we still need either a small C or Python reproducer which leverages the libxml2 API.
The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=eb72ee41457cc0b7f93c1d70f170ddc8ea877175 commit eb72ee41457cc0b7f93c1d70f170ddc8ea877175 Author: Sam James <sam@gentoo.org> AuthorDate: 2023-05-01 06:24:33 +0000 Commit: Sam James <sam@gentoo.org> CommitDate: 2023-05-01 06:24:33 +0000 dev-libs/libxml2: add note re itstool/unicode crash patch We need to check if it's still relevant w/ 2.11.0. Bug: https://bugs.gentoo.org/745162 Signed-off-by: Sam James <sam@gentoo.org> dev-libs/libxml2/libxml2-2.11.1.ebuild | 3 +++ profiles/package.mask | 2 ++ 2 files changed, 5 insertions(+)
The bug has been closed via the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=9b2ad65342b2445a38775260e7f4497d06466ee4 commit 9b2ad65342b2445a38775260e7f4497d06466ee4 Author: Sam James <sam@gentoo.org> AuthorDate: 2023-05-10 19:33:44 +0000 Commit: Sam James <sam@gentoo.org> CommitDate: 2023-05-10 19:33:44 +0000 profiles: unmask new libxml2 Seems to have fixed the python bindings issue too: https://gitlab.gnome.org/GNOME/libxml2/-/commit/76c6da420923f2721a2e16adfcef8707a2454a1b. Closes: https://bugs.gentoo.org/745162 Signed-off-by: Sam James <sam@gentoo.org> dev-libs/libxml2/libxml2-2.11.2.ebuild | 3 --- profiles/package.mask | 7 ------- 2 files changed, 10 deletions(-)
