With latest ebuilds, prusaslicer crashes very often even without user interaction. Reproducible: Sometimes Steps to Reproduce: 1.launch prusa-slicer 2.use it somehow 3.crash Happens most of the time Actual Results: Prusa-slicer crashes. Using --loglever 5, last stderr line is usually *** longjmp causes uninitialized stack frame ***: terminated Expected Results: Should work okay
Need to grab a backtrace if possible. See https://wiki.gentoo.org/wiki/Debugging.
Created attachment 787304 [details] prusa-2.4.2_p20220506 with loglevel 5 trace
Created attachment 787307 [details] prusa-2.6.0_pre20220601 loglevel 5 trace
Created attachment 787310 [details] prusaliscer-2.6.0_pre20220620 loglevel 5 trace
Created attachment 787313 [details] emerge --info
(In reply to Sam James from comment #1) > Need to grab a backtrace if possible. See > https://wiki.gentoo.org/wiki/Debugging. Ok preparing that.
https://github.com/prusa3d/PrusaSlicer/issues/8395 maybe.
Do you still face the issue on media-gfx/prusaslicer-2.6.0_pre20220620?
(In reply to Piotr Karbowski from comment #8) > Do you still face the issue on media-gfx/prusaslicer-2.6.0_pre20220620? I have tried for a bit, but have not encountered the longjmp crash. I've had couple crashes with `Alarm clock`, but these are probably unrelated.
Were the alarm clock crashes seen only by gdb, or actually the application were crashing and closing?
(In reply to Piotr Karbowski from comment #10) > Were the alarm clock crashes seen only by gdb, or actually the application > were crashing and closing? Not through GDB. And correction: The longjmp problems are still here, only less frequent.
Please check 2.5.0_alpha2. It seems to not be based on master branch but on 2.4.2 instead, perhaps it will solve your problems. I consider it to replace 2.6.0_pre* snapshots later once stabilized. For now you will need to merge exact version like: emerge -av '=media-gfx/prusaslicer-2.5.0_alpha2'
(In reply to Piotr Karbowski from comment #8) > Do you still face the issue on media-gfx/prusaslicer-2.6.0_pre20220620? Yes, and with latest 2.6.0_pre20220620-r1 also.
Commented upstream (https://github.com/prusa3d/PrusaSlicer/issues/8395#issuecomment-1173170799): ``` 1. Can folks hitting this show me `emerge -v1p net-misc/curl` (to see the flags)? Flipping USE=threads on curl may make a difference. Also, try flipping USE=adns. 3. Yeah, I really think this is to do with the curl timeout being safe or not: - https://bugzilla.redhat.com/show_bug.cgi?id=539809 - https://curl.se/mail/lib-2008-09/0197.html - https://curl.se/mail/lib-2015-03/0136.html ```
(In reply to Sam James from comment #14) > Commented upstream > (https://github.com/prusa3d/PrusaSlicer/issues/8395#issuecomment-1173170799): > ``` > 1. Can folks hitting this show me `emerge -v1p net-misc/curl` (to see the > flags)? Flipping USE=threads on curl may make a difference. Also, try > flipping USE=adns. [ebuild U ] net-misc/curl-7.84.0::gentoo [7.83.1::gentoo] USE="ftp http2 imap openssl pop3 progress-meter smtp ssl tftp -adns -alt-svc -brotli -gnutls -gopher -hsts -idn -ipv6 -kerberos -ldap -mbedtls (-nghttp3) -nss (-quiche) -rtmp -samba -ssh -sslv3 -static-libs -telnet -test -threads -verify-sig -zstd" CURL_SSL="openssl -gnutls -mbedtls -nss" 2,420 KiB So currently no useflags threads nor adns. Setting them up, will test tomorrow (Europe time)
(In reply to xiaojie+bugzillagentoo from comment #15) > (In reply to Sam James from comment #14) > > Commented upstream > > (https://github.com/prusa3d/PrusaSlicer/issues/8395#issuecomment-1173170799): > > ``` > > 1. Can folks hitting this show me `emerge -v1p net-misc/curl` (to see the > > flags)? Flipping USE=threads on curl may make a difference. Also, try > > flipping USE=adns. > > [ebuild U ] net-misc/curl-7.84.0::gentoo [7.83.1::gentoo] USE="ftp > http2 imap openssl pop3 progress-meter smtp ssl tftp -adns -alt-svc -brotli > -gnutls -gopher -hsts -idn -ipv6 -kerberos -ldap -mbedtls (-nghttp3) -nss > (-quiche) -rtmp -samba -ssh -sslv3 -static-libs -telnet -test -threads > -verify-sig -zstd" CURL_SSL="openssl -gnutls -mbedtls -nss" 2,420 KiB > > So currently no useflags threads nor adns. Setting them up, will test > tomorrow (Europe time) Btw, seems to be either threads or adns, not both (emerge does not allow this).
(In reply to xiaojie+bugzillagentoo from comment #15) > (In reply to Sam James from comment #14) > > Commented upstream > > (https://github.com/prusa3d/PrusaSlicer/issues/8395#issuecomment-1173170799): > > ``` > > 1. Can folks hitting this show me `emerge -v1p net-misc/curl` (to see the > > flags)? Flipping USE=threads on curl may make a difference. Also, try > > flipping USE=adns. > > [ebuild U ] net-misc/curl-7.84.0::gentoo [7.83.1::gentoo] USE="ftp > http2 imap openssl pop3 progress-meter smtp ssl tftp -adns -alt-svc -brotli > -gnutls -gopher -hsts -idn -ipv6 -kerberos -ldap -mbedtls (-nghttp3) -nss > (-quiche) -rtmp -samba -ssh -sslv3 -static-libs -telnet -test -threads > -verify-sig -zstd" CURL_SSL="openssl -gnutls -mbedtls -nss" 2,420 KiB > > So currently no useflags threads nor adns. Setting them up, will test > tomorrow (Europe time) As far as the tests I did go, curl[threads] or curl[adns] do NOT trigger the crashes. Will go on using prusaslicer a little bit more to be more affirmative (or not, depends).
What do you have on mind is that, as long as curl (libcurl) is built with threads, prusa-slicer do not have random crashes for you? That's rather odd, given that those crashes are random, and we do not get any of undefined reference errors.
(In reply to Piotr Karbowski from comment #18) > What do you have on mind is that, as long as curl (libcurl) is built with > threads, prusa-slicer do not have random crashes for you? That's rather odd, > given that those crashes are random, and we do not get any of undefined > reference errors. I only followed what is hinted at https://github.com/prusa3d/PrusaSlicer/issues/8395#issuecomment-1173170799. Seems to work (at least got no crashes yesterday and today).
(In reply to Piotr Karbowski from comment #18) > What do you have on mind is that, as long as curl (libcurl) is built with > threads, prusa-slicer do not have random crashes for you? That's rather odd, > given that those crashes are random, and we do not get any of undefined > reference errors. It's not random at all -- see my explanation on the upstream bug. It's to do with timeouts when curl doesn't have asynchronous DNS support. The timeouts then interrupt unsafely.
(In reply to Sam James from comment #20) > It's not random at all -- see my explanation on the upstream bug. It's to do > with timeouts when curl doesn't have asynchronous DNS support. The timeouts > then interrupt unsafely. We're likely to have to add a dep on net-misc/curl[adns], but let's await more testing first.
The bug has been closed via the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=4affbff1644fa09e286957bd246a00b3c9dc5799 commit 4affbff1644fa09e286957bd246a00b3c9dc5799 Author: Piotr Karbowski <slashbeast@gentoo.org> AuthorDate: 2022-07-06 21:01:43 +0000 Commit: Piotr Karbowski <slashbeast@gentoo.org> CommitDate: 2022-07-06 21:25:12 +0000 media-gfx/prusaslicer: depend on curl with async DNS support. Closes: https://bugs.gentoo.org/853973 Signed-off-by: Piotr Karbowski <slashbeast@gentoo.org> ...usaslicer-2.5.0_alpha2.ebuild => prusaslicer-2.5.0_alpha2-r1.ebuild} | 2 +- ....0_pre20220601-r1.ebuild => prusaslicer-2.6.0_pre20220601-r2.ebuild} | 2 +- ....0_pre20220620-r1.ebuild => prusaslicer-2.6.0_pre20220620-r2.ebuild} | 2 +- 3 files changed, 3 insertions(+), 3 deletions(-)