Description of problem: Thanks to the upstream developer the following stack trace exists now: #0 0x7dbfc6c8 in __GI__IO_wsetb () from /lib/libc.so.6 #1 0x7dc07458 in _IO_cleanup () from /lib/libc.so.6 #2 0x7dbc797a in __run_exit_handlers () from /lib/libc.so.6 #3 0x7dbc7a20 in exit () from /lib/libc.so.6 #4 0x55555d58 in main (argc=6, argv=<optimized out>) at xorriso/xorriso_main.c:291 Version-Release number of selected component (if applicable): libisoburn-1.4.2-1.fc24 libisoburn-1.4.2-2.fc24 How reproducible: Everytime on Fedora 24 and Rawhide on s390, however not on Fedora 23 and/ or s390x. Steps to Reproduce: Just rebuild the source RPM of libisoburn on s390 or run test suite. Actual results: Test suite of xorriso crashes only on s390 with SIGSEGV, Segmentation fault. Expected results: No crash of test suite of xorriso on s390 (like on all other archs).
Bug #1283284 looks similar according to upstream's investigation.
A few remarks from upstream: The younger levels of the call stack shown at https://bugzilla.redhat.com/show_bug.cgi?id=1283284#c2 resemble much the call stack here. But in bug 1283284 it seems to happen by an emergency exit at start time, whereas with libisoburn's xorriso it happens at regular exit after all intended work has succeeded. Maybe there is a hint in the fact that GNU xorriso of the same revision does not crash at exit. Same souce code involved. The libisoburn xorriso packaged by Fedora is linked with three dynamic libraries, which GNU xorriso links statically as .o files. libisoburn.so.1 => /tmp/libisoburn-buildroot/usr/lib/libisoburn.so.1 (0x7d072000) libisofs.so.6 => /lib/libisofs.so.6 (0x7cff1000) libburn.so.4 => /lib/libburn.so.4 (0x7cf22000) Further, the crash does not happen after all the xorriso runs in the test. But it happens reproducibly with the same xorriso runs if the overall test script is executed again. The tasks which fail and those which succeed give me no clue where in my source a problem might hide. This pattern somehow points to a subtle memory corruption while dynamic linking. Small changes in code composition cause the problem to occur or to vanish. I lack an explanation why only two packages are known to be affected. (I came to bug 1283284 by the rareness of identifier "__GI__IO_wsetb" in the web.)
Matthias Klose today fixed a similar failure of xorriso on Ubuntu powerpc by preventing option --version-script=libisoburn/libisoburn.ver from being applied by the linker run of the xorriso binary. https://bugs.launchpad.net/ubuntu/+source/libisoburn/+bug/1571684 Observed symptom after run of xorriso -version : Program received signal SIGSEGV, Segmentation fault. __GI__IO_wsetb (f=f@entry=0x1feccca0 <_IO_stdout_>, b=b@entry=0x0, eb=eb@entry=0x0, a=a@entry=0) at wgenops.c:105 -------------------------------------------------------------------- I made a quick test: ssh -l scdbackup lfedora1.lf-dev.marist.edu mock -r fedora-rawhide-s390 shell cd /tmp/libisoburn-1.4.2 make clean ./configure --disable-versioned-libs make ... it is quite slow tonight ... cd releng ./auto_isocontent -x ../xorriso/xorriso No crashes with this test run. To verify that it is indeed an improvement, i went back to default configuration: cd .. make clean ./configure --enable-versioned-libs make cd releng ./auto_isocontent -x ../xorriso/xorriso Now two of the xorriso runs in ./auto_isocontent fail with SIGSEGV. -------------------------------------------------------------------- I will probably endorse the patch of Mattias Klose in libisoburn, libburn, and libisofs. cdrskin from libburn seems prone to the same problem. Given the volatile nature of the problem, i am not completely sure that --version-script=... with non-lib binaries is really the original cause. But i understand it is wrongly applied in my build system.
What is the contents of the problematic version script?
The file contains the API symbols of libisoburn http://libburnia-project.org/browser/libisoburn/trunk/libisoburn/libisoburn.ver in order to keep applications from using non-public calls. Matthias Klose pointed out that it is inappropriate to apply it to executables. I hopefully avoid this now by http://libburnia-project.org/changeset/5692 which is essentially his Ubuntu patch. He stated "this happened after updating glibc 2.21 to 2.23". The file libisoburn.ver was introduced in juli 2010. Only after nearly six years and only on rather exotic processor architectures it seems to make trouble when applied to the xorriso binary. With Ubuntu powerpc it still gets applied to libisoburn (without negative effects). In my rough test on Fedora s390 it was not applied at all. I now refined the test by applying the change manually to the source in my mock-chroot. libisoburn.ver is applied to /bin/sh ./libtool [...] --mode=link [...] -o libisoburn/libisoburn.la but not to /bin/sh ./libtool [...] --mode=link [...] -o xorriso/xorriso The test script then runs without crash.
It seems like Fedora does not build s390 anymore, just s390x - at least when looking to http://s390.koji.fedoraproject.org/koji/packageinfo?packageID=12555 Thus I'm tempted to close this...opinions?
To my opinion it should be fixed since release of xorriso-1.4.4 in juli 2016 which included the mentioned changeset 5692. (There is the suspicion that dynamically linked xorriso-1.4.2 on FreeBSD 11 suffers from the same problem. Diffuse SIGSEGV. GNU xorriso-1.4.2 works fine.) Thanks for your swift work, Robert, when packaging xorriso-1.4.6 this weekend. :)
Potentially related: https://sourceware.org/ml/libc-alpha/2016-08/msg00820.html At this point, the only possible remedy is to adjust the version script. It should not filter out symbols in the implementation namespace. This applies to everything, both DSOs and application binaries.
If there is a problem with version scripts and .so libraries then this has not shown up yet. On Ubuntu powerpc and on Fedora s390 it sufficed to omit the script with the linker run of the xorriso binary. I use version scripts to keep library users from calling inner functions, which are not part of the API. So if it turns out that version scripts become a problem with the libraries, i would omit them by default. But first i would want to see an actual crash with xorriso 1.4.4 or 1.4.6.
This message is a reminder that Fedora 24 is nearing its end of life. Approximately 2 (two) weeks from now Fedora will stop maintaining and issuing updates for Fedora 24. It is Fedora's policy to close all bug reports from releases that are no longer maintained. At that time this bug will be closed as EOL if it remains open with a Fedora 'version' of '24'. Package Maintainer: If you wish for this bug to remain open because you plan to fix it in a currently maintained version, simply change the 'version' to a later Fedora version. Thank you for reporting this issue and we are sorry that we were not able to fix it before Fedora 24 is end of life. If you would still like to see this bug fixed and are able to reproduce it against a later version of Fedora, you are encouraged change the 'version' to a later Fedora version prior this bug is closed as described in the policy above. Although we aim to fix as many bugs as possible during every release's lifetime, sometimes those efforts are overtaken by events. Often a more recent Fedora release includes newer upstream software that fixes bugs or makes them obsolete.
Fedora 24 changed to end-of-life (EOL) status on 2017-08-08. Fedora 24 is no longer maintained, which means that it will not receive any further security or bug fix updates. As a result we are closing this bug. If you can reproduce this bug against a currently maintained version of Fedora please feel free to reopen this bug against that version. If you are unable to reopen this bug, please file a new report against the current release. If you experience problems, please add a comment to this bug. Thank you for reporting this bug and we are sorry it could not be fixed.