Description of problem:
Apparently the new libreport package lost the ability to search for duplicates when reporting to Bugzilla.
Bug #731260 has been reported today at 10:53:26 EEST then an hour later I hit the same bug and repoted it as bug #731275 at 11:48:43 EEST.
As you can see both bugs are the same thing but libreport didn't figure that out.
Version-Release number of selected component (if applicable):
The tracebacks are different, the crash happens on different line in different function, so it's not really a duplicate, can you try it twice with the same anaconda version?
Above bugs are duplicates. The crash happens on the same line in the same function:
:Traceback (most recent call first):
: File "/usr/lib/anaconda/yuminstall.py", line 1158, in simpleDBInstalled
: mi = self.ts.ts.dbMatch('name', name)
this is where things blow up
The backtrace is a bit different at the last few lines though. One install was in gui mode and the other was text mode. However the cause of the error is in the underlying code. If you strip the last 2/3 lines you will notice that everything else is the same. This is pretty common with installer related bugs.
Can you make this request a RFE then and provide some mechanism for libreport to show the user possible duplicate bugs. I will be happy even if it told me the bug numbers which include backtrace which is 50% or more similar to mine.
Such feature is being developed, please attend tech talk:
Aug 24 - ABRT - backtrace analysis server - Karel Klíč
- as for the backtrace - yes, it's the same bug, but there is not much we can do about it right now. If the calling program (in this case anaconda) doesn't provide it's own de-duplication hash the reporting library just tries to compute the hash from all information it has and it treats all data the same way (which means it doesn't care if it's (C or python or something else). In this case python-meh was used to process the exception, so I think it should provide the de-duplication hash - the libreport is meant only for reporting not for backtrace parsing/analyzing ...
Moving to python-meh to see what other folks think.
Different tracebacks can often mean different errors. I would like to keep it so strictly different tracebacks still result in different bugs getting filed, and we can do the duping manually if they really are the same.
python-meh does provide a hash function. If it's not getting used, I wouldn't know why. Since we were told to move everything to libreport, I can't follow what you guys have done anymore.
(In reply to comment #5)
> python-meh does provide a hash function. If it's not getting used, I wouldn't
> know why. Since we were told to move everything to libreport, I can't follow
> what you guys have done anymore.
- it uses the hash it if it's provided, I just wasn't sure if python-meh provides it..