On Fri, Feb 22, 2019 at 09:58:52AM +0100, Jean-Philippe MENGUAL wrote: > > And send where? Upstrem? Where we have a different build config than > > them? (Let alone think about external libraries). Do we really want our > > LO to "phone home" on a crash? Do we want to set up a web service for > > those reports? (The latter one definitely is no) > > Well I am not a familiar with QA, but according to what I understand, in TDF > releases (I tried nightly before going back to Debian as it was useless), > when LO crashes, a crash report is somewhere on the TDF infra. It also is
I know. That needs infrastructure, though, as you say. > stored on the locl computer. From this file, a user can check wether a bug > is opened or not and if not, fill one with reproducible steps and this file > to help upstream devs. Yes, you can still do so manually... > > For a meaningful backtrace you just need the *-dbgsym's, not necessarily > > the "crash reporter". > > Yes but it also requirs skills related to gdb then, doesn't it? I mean, Yes. > -dbgsym are a base to get relevant crash reports, but how to generate them? > The only way I know is running gdb, but il Libreoffice the problem is > knowing on which binary. There's basically just soffice.bin as a binary. > My question is then: is reporting bug upstream from a Debian build relevant? > If yes, should not we write something to assist users doing it (a short wiki > page about how to help debugging LO from a Debian release)? https://wiki.documentfoundation.org/QA/BugReport/Debug_Information#GNU.2FLinux:_How_to_get_a_backtrace ? Regards, Rene
