Hi Simon, sorry for the late reply! On Fri, Jun 03, 2011 at 06:53:03PM +0200, Simon Chopin wrote: > Hi ! > > On Fri, Jun 03, 2011 at 04:35:05PM +0200, Guido Günther wrote: > > I'm actually having a bit of a hard time to understand to see what you > > want to accomplish. Are you talking about upstream using pristine-tar > > but a different naming convention for tarballs? Is there a usecase for > You nailed it. I know that it isn't a common case (in fact, I haven't > seen it yet outside of my use case), but I think this practice should be > encouraged, especially with the whole GitHub trend — while they provide > a way to download tarballs, it wouldn't necessarily be the same as one > generated by the developer. > And it would be pretty odd for an upstream to name its release archive > foo_1.0.orig.tar.gz, wouldn't it ? :-) > > In a more practical matter, I stumbled across this use case when trying > to set up the distribution vector of a project of mine hosted on > gitorious[1], while trying to package it as an exercise. pristine-tar > seemed to be the right answer. > > > this/ If so could you add some tests? If so wouldn't we want to be able > > to specify the exact format? > > I guess I could add some tests, but I cannot really wrap my head around > the concept of tests concerning more than a single tree of a git repo. I > shamefully confess I didn't read the test code already written, mea > culpa. > > As for the format specification, it is after all what this patch is > about. I did not push the bit further than what I thought was needed, > but in the end we could as well ask the user to provide a regexp with > the appropriate named groups.
Shouldn't we be able to identify everythig beforehand? Look at how we infert the data from the changelog. I don't think we even need a regexp. We know the package name (either from the changelog or from configuration) and only need to fill in the version number, right? > > > Cheers, > > -- Guido > > > > P.S.: having patches that don't contain "ugly hack" is an extra bonus ;) > > Yeaaaah, that one... Would it be OK if I delete the comment ? /me ducks. > I hesitated between this solution and one involving scanning multiple > times the string. IMHO both are ugly, but the hack is potentially more See above. I think we can get away with a cleaner solution by infering more data from the project. Cheers, -- Guido -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
