abrt backtrace failures

[Jiri Moskovcak]

On 05/23/2012 06:06 AM, Chuck Anderson wrote:
> During F17 RC testing, twice now [1][2] I've been unable to submit
> bugs using abrt because the backtrace was considered "unusable" by
> abrt.
>
> 1. Is there a problem with the debuginfo's or abrt functionality
> during this F17 final RC stage?
>
> 2. What happened to the Retrace Server?  I don't see that option for
> F17, only Local GNU Debugger.

This is caused by:
https://bugzilla.redhat.com/show_bug.cgi?id=823812
You can just yum install abrt-retrace-client to get it back.

>
> 3. Can the abrt behavior be changed to allow the bug to be created,
> but perhaps without the "bad" backtrace, or perhaps marked in some way
> so it is known to have a "bad" backtrace?


- right now ABRT uses a scale 1-4 to rate the backtrace and allows only 
3 and 4 to be reported
- we're thinking about adding a possibility to create the whole bug 
manually like in the web interface only it would be easier (using some 
kind of wizard probably)

>
> It is still helpful to the bug reporter to have abrt create the bug
> with as much info as possible, with the expectation that better/usable
> backtraces could be attached later manually if possible.
>

- when ABRT finds a dupe in bugzilla it checks the quality of the 
attached backtrace and if the one being reported is better it attaches 
it - so this part is done
- but on the other hand developer won't start working on a bug before he 
gets a good backtrace, so I think it's better to wait for a good one and 
then create the ticket rather than creating a ticket and waiting for a 
good backtrace - in that case the bug might go out of the developer's 
radar..

> It is frustrating to be informed of a crash by abrt and go through all
> the effort to submit the bug report via abrt, only to be told "sorry,
> I'm not going to let you submit this bug report because it doesn't
> meet our standards of quality".  Frankly, it's a waste of the
> reporter's time and it conditions the user to not care to submit bugs
> in the future.

- I totally agree, but please be patient. We're trying to make the whole 
process less painful. There are already few ongoing projects trying to 
deal with it like minidebuginfo and ABRT server (not the retrace server)

Cheers,
Jirka

>
> Thanks,
> Chuck
>
> [1] https://bugzilla.redhat.com/show_bug.cgi?id=823755
> [2] https://bugzilla.redhat.com/show_bug.cgi?id=824228