On 11/11/2016 07:20 AM, Jakub Filak wrote:
I came across an Avocado trello card requesting integration with
systemd-coredumpctl. As a developer working on Automatic Bug Reporting Tool
(ABRT) I got interested in it. I contacted Lukas Doktor with a question of
what purpose the integration should be and with an offer to engage ABRT too.
After short chat he asked me to send an RFC to this mailing list. So here
you are Lukas.
First of all, thanks for taking the time!
Let me first briefly introduce ABRT. Despite its name ABRT is more
detecting software (and hardware) problems and the reporting is just a one
step in the whole processes of handling a problem. Currently ABRT is capable
of detecting core files, uncaught Python exceptions, uncaught Ruby
exceptions, uncaught Java exceptions, Kernel oopses (including those stored
in pstore), Kernel vmcores and Machine Check Exceptions. ABRT provides
several ways accessing detect problems which includes command line tools
(abrt), GUI tools (gnome-abrt), Python API, C API and D-Bus interface. If
interested, you can find more about ABRT at:
I propose to enhance Avocado to become aware of ABRT and include detected
problems in test results. Here is my initial proof-of-concept commit:
Now, I have to repeat: thanks for taking the time to write this PoC.
Some technical details:
- the implementation will capture only problems of a current user
This is not a big deal since our current implementation for catching
coredumps is pretty much limited to super users.
Still, let me get this straight: if a test exercises and crashes a
running daemon (running as a different user) ABRT won't be able to
capture the daemon's problems, right?
But, if the user is running the test as root, then ABRT will be able to
capture all the system's problems?
- a proper configuration of PolicyKit can allow any user to read all
Right. For our purposes, it's fine to ship an example policy as a
- there can be several simultaneous connections to ABRT D-Bus
If PolicyKit is configured to allow all system problems to be read, and
there are multiple connections to ABRT, will all of those receive all
problem notifications? Or is it a single queue and whomever reads first
removes it from the queue?
Please have a look at my patch and let me know if ABRT integration
something you are interested in.
I will indeed.
Should you have any question, please contact me or any ABRT developer
crash-catcher(a)lists.fedorahosted.org or on #abrt.
[ Sr Software Engineer - Virtualization Team - Red Hat ]
[ Avocado Test Framework - avocado-framework.github.io ]