[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Backports for fglrx and chromium-browser (and two dependencies)



On Sun, 26 Jun 2011 21:56:03 +0200 Cyril Lavier wrote:

> On 06/26/2011 10:47 AM, Gerfried Fuchs wrote:
> >     Hi!
> >
> > * Michael Gilbert<michael.s.gilbert@gmail.com>  [2011-06-19 16:19:28 CEST]:
> >> On Sun, 19 Jun 2011 15:51:56 +0200 Alexander Wirt wrote:
> >>> Giuseppe Iuculano schrieb am Sunday, den 19. June 2011:
> >>>> On 06/19/2011 03:30 PM, Michael Gilbert wrote:
> >>>>> I've prepared squeeze backports for fglrx and chromium-browser (and
> >>>>> gyp and libv8 dependencies).  Would anyone be so kind as to review and
> >>>>> upload if these look good?
> >>>>
> >>>>   * allow squeeze's libvpx0 0.9.1 to satisfy dependencies (spot-checked
> >>>> some webm videos and they seem to work just fine with this older version).
> >>>>
> >>>>
> >>>> Don't do this, please. chromium 11 needs libvpx0>= 0.9.6, see  #618621
> >> All of the links in that bug report seem to work ok.  I wonder if
> >> that was just some backwards compatibility lost in version 10's
> >> development and fixed later.
> >>
> >>> for a complex beast like chromium I would prefer to have either the original
> >>> maintainer or a really experienced debian developer as maintainer of the
> >>> backport.
> >>>
> >>> Probably we would reject the package if we are not sure if the
> >>> uploader/maintainer isn't able to handle it.
> >> Personally, I see no problem handling this.  It will just be a matter
> >> of keeping up with the numerous updates that fix lots of security issues.
> >   Actually, it isn't, if you go wakenly against the explicit stated
> > recommendation of the original package maintainer here.
> >
> >> Perhaps Guisseppe can volunteer to mentor/watch my work.
> >   I would doubt that if you ignore his recommendation, and from that
> > perspective I can only repeat what Alexander wrote: For a package like
> > this the backporter should be someone that takes the concerns of the
> > original package maintainer seriously and not addresses them with a
> > "works for me" response.
> >
> >   Thanks,
> > Rhonda
> Hi.
> 
> I completely agree with Gerfried.
> 
> The "works for me" is quite subversive, because except you, no one knows 
> the tests you have done on chromium. Maybe you didn't test any website 
> or content which actually needs the version 0.9.6 of libvpx0 (some webm 
> elements might just need the 0.9.1 and other might need 0.9.6).
> 
> I'm not saying this with the goal of disrespecting you. But I've done 
> packages for five years (less than the majority of this mailing list, I 
> think), and in this short time, I experienced a lot of issues with the 
> "works for me" method of testing packages (I also remember the time it 
> took to do a clean test of the packages, because some users reported 
> bugs I didn't have).

Just to clear things up since assumed conclusions are starting to brew
on this, my initial message was along the lines of "it works for me;"
since at the time I wasn't aware of the reasoning behind the version
bump and my changes seemed to work without any problem. After Guisseppe
mentioned that bug #618621 was the catalyst for the change, I started
using that as a reference point on the issue.

Using that basis, I was able to find that all of the problematic webm
pages mentioned in that bug report work with my version (even though I'm
using the older libvpx).  Based on this evidence, it is reasonable to
conclude that this solution is a viable one, and that its not just an
"it works for me", but its an "it works for all of the known
problematic cases so far" also.

With that said, I'm certainly willing not saying that the older libvpx
has to be used, and I'l compromise and use the newer libvpx if it's
there is other evidence that the older one is insufficient. If that's
the case, I would be happy to see it and take it into consideration.

It might help to explain my motivation a bit as well.  Since backports
are supposed to involve minimal changes, I thought it better to not
backport a package if it turned out to be something unnecessary.  So,
that's what I'm trying to do, but if vpx needs to be backported, I'll
certainly do it.

Best wishes,
Mike


Reply to: