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

Re: [GR] DD should be allowed to perform binary-only uploads

On Fri, Feb 09, 2007 at 08:23:54AM -0700, Wesley J. Landaker wrote:
> On Friday 09 February 2007 05:52, Reinhard Tartler wrote:
> > The use case I imagine at this point is that a maintainer uploads a
> > library package src+bin (e.g. src+amd64) for his private arch, and after
> > weeks he notices, that it still has not been built on e.g. sparc yet. So
> > he decides to start his spare Ultra 1 workstation, builds the package in
> > his custom environment and uploads it. My question to this use case:
> >
> > What happens with the "lost" buildlogs? Is there any possibility for a
> > maintainer who depends on this library to check the build logs for this
> > package on this particular architecture? Is the maintainer somehow
> > encouraged or force by policy to publish his buildlogs?
> Does is this different from wanting to check the amd64 build log, but it 
> can't be done because that was the initial architecture upload? This 
> scenerio is basically just equivalent to a src+amd64+sparc upload instead 
> of a src+sparc or src+amd64 upload. Already maintainers can basically 
> upload src+(any architecture of their choosing) for each version. In fact, 
> I occasionally upload my own packages as src+i386, but other times as 
> src+amd64. If I had a sparc machine, I'd probably upload my packages as 
> src+sparc every once in a while just for fun and profit.
> If we think that's a bad idea, we should propose that maintainers must do 
> src+bin uploads but that the bin will be discarded and rebuild for *every* 
> architecture. To my knowledge, this has been discussed many times before 
> but never proposed officially.

  One could also ask the maintainer to upload (or send to the right
email address) the buildd logs of their build if that's really a
problem. Note that could be a good thing anyway, as it could help to
spot environment problems in the package Maintainer build env.

  I personally would advocate such an idea.

·O·  Pierre Habouzit
··O                                                madcoder@debian.org
OOO                                                http://www.madism.org

Attachment: pgpjC0DoISoGH.pgp
Description: PGP signature

Reply to: