[Reportbug-maint] Bug#577986: Bug#578296: closed by Sandro Tosi <morph at debian.org> (Re: Bug#577986: reportbug: Crash after current bugs list)

[Sandro Tosi]

Hello Klaus,

On Wed, Apr 21, 2010 at 17:34, Klaus Ethgen <Klaus at ethgen.de> wrote:
> I hope I do reply to the right bug now. There is a big mess with this
> bug.

I agree

> The problem did occur after the update today. However I cannot exclude
> that the bug was happen earlier as I did not try to report a bug the
> last days.

The original bug report, the one about crash for (filenames, md5sums),
is valid and I ack it's there. The reply that originated the closure
(and subsequente reopening) was a completely different matter, written
here since it was the most recent bug report against reportbug but
that has nothing to do with the original (an valid) report.

> Am Mo den 19. Apr 2010 um 22:28 schrieb Sandro Tosi:
>> > 1. Closing a bug "because it is a bug in another package" is always a
>> >   misuse  of the Debian BTS, which allows reassigning bugs to the package
>> >   that is really at fault.
>>
>> it's not a bug in "another package" but in god-only-knows-what: is it
>> in pango? or is it in gtk? or in glib? and in which of their multiple
>> libraries?
>
> I did use reportbug in a shell. I have no clue about any graphical tools
> as I do not use them.
>
> Just did a "reportbug kdm".

I see.

>> This problem only occurs to testing users, so hardly a wide-spread bug
>> (else it would have spotted also in sid and in many other packages).
>
> It happens in sid now.

I know.

>> I asked the first reporters (months ago) to ask a user support forum
>> (ml or irc) for help debugging what caused the problem, since I don't
>> have time nor the interest (since I couldn't replicate it) to follow
>> it up.
>
> Please tell me whet you need so maybe I can help out with it.

I was referring to the first that ever reported the "Floating point exception".

>> > It pisses users off, and prevents them from
>> >   filing bug reports in the future. Getting users to file bug reports
>> >   is hard enough without annoying the few who do.
>>
>> like pisses off maintainer for bugs reported already several other
>> times, that the user didn't even bother to search before filing
>> another one, new bugs without any additional information or "clue"
>> from the reporter.
>
> Well, that is always a problem. I try to report bugs as good as I could.

and *I* appreciate it!

> But sometimes there is something I miss. The worsted is that the
> maintainer is angry.

Really, I was not referring to you but to Elias ;)

>> >>   File "/usr/lib/pymodules/python2.5/reportbug/utils.py", line 644, in get_changed_config_files
>> >>     for (filename, md5sum) in conffiles:
>> >> ValueError: too many values to unpack
>> >
>> > 644:    for (filename, md5sum) in conffiles:
>> >
>> > While not a python programmer, seems like something has gotten into
>> > the conffiles list that is not a (filename, md5sum) tuple, and so
>> > python crashes. (Clearly, GTK is unrelated to this bug.)
>>
>> that's why I was asking to execute those 2 commands and report their
>> output here. that list is the one you can find running
>>
>> $ dpkg --status git-core
>>
>> for the config files, so:
>>
>> Conffiles:
>>  /etc/bash_completion.d/git ae19f90033978b8b9fdddbfce98d36d4
>>  /etc/emacs/site-start.d/50git-core.el a15bd943db6854ad0b599f2bcbbd5a1f
>
> Version: 1:1.7.0.4-2
> ...
> Conffiles:
>  /etc/emacs/site-start.d/50git-core.el a15bd943db6854ad0b599f2bcbbd5a1f obsolete
>  /etc/bash_completion.d/git 8f3d1ad39569e77e8394d4b97b254791 obsolete
>
>> so the filenames and the conffiles tuples list.
>
> This is no tuples, this is a driblet.

tuple as in the python data structure.

> And I see that obsolete flag in kdm too.

yes the third param 'obsolete' is was caused the problem; it seems I
tried this code path only with package without obsolete conf file.

> So I think that bug is a python bug. I see many such python code bugs in
> the last time, not only that one in reportbug. How can the split be that
> worse to break the complete tool? Seems that there is a coherence
> between python and bad written code. (But this is my view. Not helpful
> to fix the bug.) I think a user should never ever see a python
> traceback.

The bug is not in python, but in the assuption that Conffiles always
contains filenames and its md5sum, one per line, without any other
param.

> Ah, here is the complete traceback:
>   Traceback (most recent call last):
>     File "/usr/bin/reportbug", line 2082, in <module>
>       main()
>     File "/usr/bin/reportbug", line 1034, in main
>       return iface.user_interface()
>     File "/usr/bin/reportbug", line 1690, in user_interface
>       conffiles, self.options.nocompress)
>     File "/usr/lib/pymodules/python2.5/reportbug/utils.py", line 644, in get_changed_config_files
>       for (filename, md5sum) in conffiles:
>   ValueError: too many values to unpack

yep, it tries to map to a structure with 2 field when there is a
third, 'obsolete'.

Regards,
-- 
Sandro Tosi (aka morph, morpheus, matrixhasu)
My website: http://matrixhasu.altervista.org/
Me at Debian: http://wiki.debian.org/SandroTosi