Not sure what (or why/since when) is happening, but "cabal update -w ghc-9.6" for example is segfaulting in Fedora s390x copr chroots at least. Even worse - not just F39+ but also F38 even. See for example https://copr.fedorainfracloud.org/coprs/petersen/cabal-install/build/6594155/ Reproducible: Always Steps to Reproduce: 1. run "cabal update -w ghc-9.6" in s390x fedora copr build (ie in spec file) Actual Results: + cabal update -v3 -w ghc-9.6.3 Config file path source is default config file. Config file not found: /builddir/.config/cabal/config Writing default configuration to /builddir/.config/cabal/config Searching for curl in path. Found curl at /usr/bin/curl Searching for powershell in path. Cannot find powershell on the path Searching for wget in path. Cannot find wget on the path Selected http transport implementation: curl File monitor 'config' changed: first run Project settings changed, reconfiguring... creating /builddir/build/BUILD/cabal-install-3.10.1.0/dist-newstyle creating /builddir/build/BUILD/cabal-install-3.10.1.0/dist-newstyle/cache /var/tmp/rpm-tmp.LV9I8Q: line 48: 7846 Segmentation fault (core dumped) cabal update -v3 -w ghc-9.6.3 Expected Results: No crash Previous builds in May were successful: eg https://download.copr.fedorainfracloud.org/results/petersen/cabal-install/fedora-rawhide-s390x/05989927-cabal-install/
Also just confirmed same segfault happening in Koji s390x rawhide (though of course cabal-install cannot normally be used there due to no network). Segfault doesn't happen with F37 s390x though.
(Just for completeness: same thing with latest cabal-3.10.2.0 just released: https://copr.fedorainfracloud.org/coprs/petersen/cabal-install/build/6601582/ ) But I confirmed in Koji that "cabal update" also segfaults
(epel9 also does not segfault of course)
I managed to reproduce in a s390x fedora:39 container: strace output newfstatat(AT_FDCWD, "/usr/bin/ghc-pkg", {st_mode=S_IFREG|0755, st_size=352, ...}, 0) = 0 getpid() = 279 openat(AT_FDCWD, "/dist-newstyle/cache/compiler279-0.tmp", O_RDWR|O_CREAT|O_EXCL|O_NOCTTY|O_NONBLOCK, 0666) = 11 newfstatat(11, "", {st_mode=S_IFREG|0644, st_size=0, ...}, AT_EMPTY_PATH) = 0 ioctl(11, TCGETS, 0x3ffc88f5988) = -1 ENOTTY (Inappropriate ioctl for device) sched_getaffinity(0, 128, [0 1]) = 8 rt_sigprocmask(SIG_BLOCK, [INT], [], 8) = 0 --- SIGSEGV {si_signo=SIGSEGV, si_code=SEGV_MAPERR, si_addr=0x3fbe4400000000} --- +++ killed by SIGSEGV (core dumped) +++ Segmentation fault (core dumped)
cabal-install-3.10.1.0-1.fc39.s390x also dumps core inside fedora:39 (gdb) bt #0 0x000000008000e008 in evacuate1 () #1 0x0000000083f3c9b6 in scavenge_loop1 () #2 0x0000000083f31cbc in scavenge_until_all_done () #3 0x0000000083f334fa in GarbageCollect () #4 0x0000000083f1dce2 in scheduleDoGC () #5 0x0000000083f1ee90 in schedule () #6 0x0000000083f1f6de in scheduleWaitThread () #7 0x0000000083f19610 in hs_main () #8 0x000000008003226e in main ()
Reverting F39 to https://kojipkgs.fedoraproject.org//packages/cabal-install/3.6.2.0/3.fc39/s390x/cabal-install-3.6.2.0-3.fc39.s390x.rpm stops `cabal update` from segfaulting and after that (~/.cabal/packages/ created) current cabal-install-3.8.1.0 no longer segfaults.
Crash is also reproduced with fedora-toolbox:39 toolbox (provided there is no existing ~/.cabal/)
(In reply to Jens Petersen from comment #6) > Reverting F39 to > https://kojipkgs.fedoraproject.org//packages/cabal-install/3.6.2.0/3.fc39/s390x/cabal-install-3.6.2.0-3.fc39.s390x.rpm > stops `cabal update` from segfaulting Perhaps more interestingly downgrading from Koji cabal-install-3.8.1.0-1.fc39.s390x to https://download.copr.fedorainfracloud.org/results/petersen/cabal-install-3.8/fedora-39-s390x/04784966-cabal-install/cabal-install-3.8.1.0-1.fc38.s390x.rpm also stops the segfault. https://kojipkgs.fedoraproject.org//packages/cabal-install/3.8.1.0/1.fc39/s390x/cabal-install-3.8.1.0-1.fc39.s390x.rpm also segfaults
See https://gitlab.haskell.org/ghc/ghc/-/issues/24163
FEDORA-2023-d1dd535dda has been submitted as an update to Fedora 38. https://bodhi.fedoraproject.org/updates/FEDORA-2023-d1dd535dda
FEDORA-2023-d1dd535dda has been pushed to the Fedora 38 testing repository. Soon you'll be able to install the update with the following command: `sudo dnf upgrade --enablerepo=updates-testing --refresh --advisory=FEDORA-2023-d1dd535dda` You can provide feedback for this update here: https://bodhi.fedoraproject.org/updates/FEDORA-2023-d1dd535dda See also https://fedoraproject.org/wiki/QA:Updates_Testing for more information on how to test updates.
FEDORA-2023-d1dd535dda has been pushed to the Fedora 38 stable repository. If problem still persists, please make note of it in this bug report.
Could you also release new packages for F39 including a rebuild of cabal?
Thank you Seems I didn't push the f39 build into the updates system, ugh
FEDORA-2024-88c617181a has been submitted as an update to Fedora 39. https://bodhi.fedoraproject.org/updates/FEDORA-2024-88c617181a
FEDORA-2024-88c617181a has been pushed to the Fedora 39 testing repository. Soon you'll be able to install the update with the following command: `sudo dnf upgrade --enablerepo=updates-testing --refresh --advisory=FEDORA-2024-88c617181a` You can provide feedback for this update here: https://bodhi.fedoraproject.org/updates/FEDORA-2024-88c617181a See also https://fedoraproject.org/wiki/QA:Updates_Testing for more information on how to test updates.
FEDORA-2024-88c617181a (cabal-install-3.8.1.0-5.fc39 and ghc-9.4.5-137.fc39) has been pushed to the Fedora 39 stable repository. If problem still persists, please make note of it in this bug report.
I gave cabal a try and it still fails. Is it possible that cabal-install was still compiled with the broken GHC? $ rpm -q cabal-install cabal-install-3.8.1.0-5.fc39.s390x $ cabal update Config file path source is default config file. Config file not found: /root/.cabal/config Writing default configuration to /root/.cabal/config Segmentation fault (core dumped)
Hm, it was built with ghc-compiler-9.4.5-137.fc39.s390x, which is supposed to have the fixed... https://koji.fedoraproject.org/koji/taskinfo?taskID=112716128 Just double-checking: so "/usr/bin/cabal update" segfaults for you, right? I need to check more carefully then I guess.
Is it because I need to rebuild all the deps too? ;-( I seem to remember something like that: need to check my notes in gitlab.
(In reply to Jens Petersen from comment #19) > Hm, it was built with ghc-compiler-9.4.5-137.fc39.s390x, > which is supposed to have the fixed... > > https://koji.fedoraproject.org/koji/taskinfo?taskID=112716128 > > Just double-checking: so "/usr/bin/cabal update" segfaults for you, right? Yes, I have tested it in a fresh container, i.e., no old/potentially broken $HOME/.cabal lingering around are whatever.
As I feared: https://gitlab.haskell.org/ghc/ghc/-/merge_requests/11662#note_536725 > (Just rebuilding the Fedora cabal-install package was insufficient alas.) Sigh
I wonder if the mass-rebuild was even enough to fix this for deps in F40.
Okay I rebuilt cabal-install for Rawhide and next F40, which should fix the problem there.
I can confirm that cabal is working for me on rawhide and next (40). However, it is still segfaulting on 39 for me.
Okay, thank you. Yeah, I didn't have time yet sorry - there has been quite a lot of fall out for the F40, and also some pending pandoc CVEs. So I need more time - but let me know if you consider this is a priority, and I can try to bump it up my to-do list.
Fine for me. Just wanted to confirm that F40/Rawhide packages are good. Once F39 packages are rebuild I will give them a try, too.