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

Re: ghc-7.10.3 is out...



On Tue, 15 Dec 2015 17:34:32 +0100
Sven Bartscher <sven.bartscher@weltraumschlangen.de> wrote:

> On Sun, 06 Dec 2015 23:59:13 +0100
> Joachim Breitner <nomeata@debian.org> wrote:
> 
> > Hi,
> > 
> > Am Sonntag, den 06.12.2015, 19:15 +0000 schrieb Clint Adams:  
> > > On Sun, Dec 06, 2015 at 06:51:51PM +0100, Joachim Breitner wrote:    
> > > > Likely a -1 version was added to the repository but no according upload
> > > > was made. You bumped the version to -2, causing dpkg-buildpackage not
> > > > to include the source in the upload, which then failed.    
> > > 
> > > So one recent example is haskell-parsers_0.12.3-1_amd64.changes , which
> > > is an amd64+all (no source) changes file produced by dht make-all,
> > > and I'm not sure why.    
> > 
> > the code of make-all.hs indeed passes "no-source" to sbuild. This was
> > introduced by Sven in this commit:
> > http://anonscm.debian.org/cgit/pkg-haskell/pkg-haskell-tools.git/commit/?id=f1da4e91f233d22d15fd6dcf6e850532a320b175
> > 
> > The commit message does not explain that change. Maybe it was
> > inadvertent?
> > 
> > Sven, can you comment?  
> 
> Sorry for not answering for so long, I was sick and didn't have access
> to my email.
> 
> I think the problem was that sbuild overwrites the original dsc file
> during the build and at some later point something tries to add some
> information (probably about the source) to the freshly built dsc, which
> already contained the information which resulted in some error.
> I will look deeper into this, when I've got some spare time.
> I retrospect, I should have written a better commit message.

The problem I tried to fix with that flag just happened to me.
The problem was, that after building the package, make-all uses adddsc
to add the dsc file that was generated before the build to the .changes
file. If sbuild is configured to build with source by default, that
file will already be added to the .changes file and adddsc will fail,
that's why I made make-all always build without source.
I think we should try to make make-all work regardless fo the users
sbuild configuration. I guess the best solution would be to not fail if
adddsc fails.

Regards
Sven

Attachment: pgpzGPbnmBwKe.pgp
Description: Digitale Signatur von OpenPGP


Reply to: