Re: [Banshee-List] [banshee] Re-enable AAC profile bgo#612741

["Andres G. Aragoneses" <knocte gmail com>]

On 22/08/12 18:01, Bertrand Lorentz wrote:
> On Wed, Aug 22, 2012 at 6:34 PM, Andres G. Aragoneses <knocte gmail com> wrote:
> > On 29/06/12 22:40, gnomeuser gmail com wrote:
> > > 2012/6/29 Andres G. Aragoneses <knocte gmail com>:
> > > > On 29/06/12 18:24, gnomeuser gmail com wrote:
> > > > > 2012/6/29 Andres G. Aragoneses <knocte gmail com>:
> > > > > > <snip />
> > > > > put out for OS X and Linux will of course be against the new upcoming
> > > > > release. Regardless the errors are only for certain aac files and it
> > > > > is not fatal as in eating data or crashing Banshee.
> > > >
> > > > I thought it was causing a crash, sorry.
> > >
> > > It's still not pretty though, you get files imported with bad or empty
> > > tags which leaves the visual experience rather crappy but it is not
> > > fatal and with the Last.fm fingerprinter we can typically autocorrect
> > > them if I remember correctly.
> >
> > This one was recently fixed by Alan and *was* causing a crash:
> >
> > https://github.com/mono/taglib-sharp/commit/e53ffcc4c417bac611fae80d1d0a88ecfafcfc6d
> >
> > https://bugzilla.gnome.org/show_bug.cgi?id=676934
> >
> > So should we reconsider about tagging new taglib-sharp and make banshee 2.5
> > depend on it?
>
> That fix is already included in the latest taglib-sharp release:
> http://download.banshee.fm/taglib-sharp/2.1.0.0/taglib-sharp-2.1.0.0.news

Ah right, yeah I meant just depending on it then.


> As for bumping dependencies to get bug fixes, I'm not in favor of that
> in general, for the following reasons:
> 1/ Version requirements checked by ./configure are about what is
> necessary to be able to build. Additional runtime requirements are
> about what's needed to get some features.
> The "get the latest version of the deps to get the latest bugfixes" is
> implied, kind of.

Yes but sometimes we as developers need to give packagers a kick: 
otherwise they slack and not package and then users end up using buggy 
code which we suffer because they end up filing reports in our bug tracker.


> 2/ Distros might already have the bugfix backported to the version of
> the library they ship, so bumping the dep only brings pain to the
> packagers, and no benefit to their users.

I consider the practice of backporting patches generally a big hack: 
because when they publish some package as version X.Y, they are actually 
lying if they are in the end including some patch that only made it in 
X.Y+1. The consequence of this is version hell from a traceability 
standpoint: users may say in bugs they have X.Y when they effectively 
have X.Y+1.


> 3/ Even if it's during a development phase, it makes it harder for
> potential contributor to build Banshee from git master. From what I

That's the only point where I agree with you.

But anyway that forces devs to give packagers a kick (or even: to make 
devs more inclined to start learning about packaging).


> heard, this is the main issue with contributing to the GNOME core
> apps. I'm tempted to call it the "OMG! I need to build the whole world
> with jhbuild" syndrom ;)

Been there, I can see that :-/


> As I said, it's a general guideline, not an absolute rule. If the
> bugfix prevents a systematic crash on startup, then it's another
> story.
>
> And also, this is in the context of how we currently ship apps on
> Linux, which is not perfect either...