Re: createrepo 0.9.4 traceback with certain changelog data

Hi Panu,

Am Samstag, 23. Februar 2008 schrieb Panu Matilainen:
> On Wed, 13 Feb 2008, Hans-Peter Jansen wrote:
> > Am Montag, 11. Februar 2008 schrieben Sie:
> >> On Sun, 2008-02-10 at 11:53 +0100, Hans-Peter Jansen wrote:
> >>> Hi Seth,
> >>>
> >>> I think, it's better to not clutter up the ML with such stuff..
> >>>
> >>> Attached.
> >>
> >> in fedora 8 it shows changelogtime as a list 204 items long.
> >
> > Well the problem is the other way around. Or do you say, that a rpm
> > packeage with a _single_ changelog entry appear as a 204 items long
> > list ;-)
> >
> > Please either test the supplied rpm with exactly _one_ changelog entry,
> > or create one for your one, and test that.
> >
> >> I think it is a suse rpm, thing.
> >
> > If that's the case, then earlier fedora/cent os releases are affected
> > as well, since suse won't change such basic python interface code just
> > for fun..
>
> Suse might missing this patch in their rpm (not treating changelog time
> as a list always is a bug):
>
> diff -r 6a6942188e65 -r 08400e947833 python/header-py.c
> --- a/python/header-py.c        Tue May 01 05:55:14 2007 +0100
> +++ b/python/header-py.c        Thu May 10 12:35:49 2007 +0100
> @@ -537,6 +537,7 @@ static PyObject * hdr_subscript(hdrObjec
>       case RPMTAG_CONFLICTNAME:
>       case RPMTAG_CONFLICTFLAGS:
>       case RPMTAG_CONFLICTVERSION:
> +    case RPMTAG_CHANGELOGTIME:
>          forceArray = 1;
>          break;
>       case RPMTAG_SUMMARY:

I suspected such a thing, but only investigated the suse rpm build for 
suspicious patches. I should have looked into the fedora rpm.spec..

Is it possible to relate this fix to a certain rpm/fedora version? I'm going 
to create a backwards compatibility fix for yum (since I'm not willing to 
modify rpm only to get the new createrepo going..). If it's predictably 
useful beyond suse, I may try harder to create something acceptable 
upstream. 

Anyway, thanks for this very helpful hint, Panu.

Pete