Bug#481457: python-gtk2: 2.12.1-2 breaks matplotlib and likely other apps as well

[mmetz at astro.uni-bonn.de]

> Le jeudi 05 juin 2008 à 17:33 +0200, Manuel Metz a écrit :
>> well yes, I did. I don't know/work with "meld". So, it might be that
>> there is more that causes trouble with this software, which I can not
>> know of (but according to the bug reports the bug was resolved in
>> 2.12.1-4, wasn't it ?).
>
> No, it was not. The problem is not caused by the patch itself, but by
> the introduction of wakeupfd in pygtk, which is in use since python2.5
> 2.5.2-5.
>
>>    I was reporting the other bug because the very simple example did not
>> work. I AM MORE THAN HAPPY TO HELP TO IMPROVE Debian, but _excuse me_, I
>> can not understand why to call a patch "part of the solution" if it
>> causes a problem that was NOT apparent before. And I have not seen any
>> other bug-report that is solved by this patch. So for me, this solution
>> caused a problem, which I was reporting.
>>    There might be, of course, another problem which has not not been
>> reported, and the patch might be part of that other problem - that I can
>> not know since it was not talked about. But how can it be a solution to
>> the problem I was reporting, when the former package worked fine ???
>
> There are applications that lock without the patch. Some lock with this
> patch, and some lock with the new version that has been proposed by
> upstream. The wakeupfd stuff is dealing with the main loop's internals,
> and it is hard to understand the implications - I’m pretty sure I
> don’t
> understand half of them.

Hi, first thanks for the info - I think I understand a bit better now
what's going on.

> In all cases, I fail to see how you can call the 2.12.1-4 version
> “working”, since the memory corruption bug causes crashes in most
> pygtk
> applications with this version.

Well, that's just how it was for _me_ on _my_ machine. I had/have no
problems with 2.12.1-4 until now and that's also why I switched back to
this version and call it "working". It's the first time now that I read
about these "memory corruption problems", and when I look into the
bugreports of those meld and other reports, they were _all_ marked as
"Fixed in version 2.12.1-4" ... !?

Cheers,
  Manuel