Thursday, January 14, 2016 8:50 AM +01:00 from Florian Weimer
<fweimer(a)redhat.com>:
> How it is supposed to be debugged by upstream developers?
With GDB?
Yes, for C/C++ packages.
Fedora provides debugging information for most of its packages, and
you
can extract them from RPMs and specify “set debug-file-directory” in GDB
to use them, even without installing them.
-debuginfo should be for the same build version as a binary itself.
Most users never install -debuginfo. I'm not sure that old packages are tracked
somewhere for, say, rawhide.
Please correct me if I'm wrong.
Note that Fedora also compiles with -fexceptions, which provides
unwinding information despite missing frame pointers. In general, the
debugging experience is *much* better than on other systems.
Yeah, -fexceptions is very useful.
I enabled it long time ago.
> It would be nice to have **at least** a proper backtrace for
crashed
daemons.
> Even better to have a) coredump b) binary c) debug symbols for
this
version of binary.
> Otherwise I can't suggest to use such packages for the end
users.
coredumpctl works well enough for me.
Does it log stack traces with symbol names on crash?
Currently we manually produce stack traces on crashes in Tarantool (a in-memory
database),
because coredumps are completely overkill for some large instances.
--
WBR,
Roman Tsisyk <roman(a)tarantool.org>
http://tarantool.org/ - an efficient in-memory data store and a Lua application
server