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

Re: Versioned Dependencies (Was: Re: libjgrapht0.6-java_0.6.0-7_amd64.changes is NEWW)



On Wed, Sep 16, 2009 at 2:05 PM, Michael Koch <konqueror@gmx.de> wrote:
> On Wed, Sep 16, 2009 at 01:59:56PM +0530, Onkar Shinde wrote:
>> On Wed, Sep 16, 2009 at 10:07 AM, Paul Wise <pabs@debian.org> wrote:
>> > On Tue, Sep 15, 2009 at 11:09 PM, Torsten Werner <twerner@debian.org> wrote:
>> >
>> >> 2. You cannot binNMU an Architecture: all package but java packages
>> >> are usually of Architecture: all.
>> >
>> > That may change at some point, there have been rumblings about
>> > throwing away maintainer-built packages (but still requiring them) and
>> > building all packages (including arch all ones) on the buildds.
>>
>> That will be great because then FTBFS will be discovered sooner than
>> later for arch:all packages.
>> By the way, why not simply upload only source packages and let them
>> get built on buildd.
>
> This saves resources on the buildds and specially guards the slower architectures.
> The main problem is that if you allow source-only uploads people will upload
> packages they dont even tried to build locally. And these are often jyst buggy.

Really. I assumed that uploaders always try building the packages in
pbuilder (or equivalent) first.
And just because the uploader build the package does not mean that it
will build on buildd. Because many of them user svn-buildpackage (or
equivalent) and there is high chance of skipping build dependencies or
having different environment.
I have come across many packages where build dependencies are not
correctly specified (missing ant from build-deps) or environment
variables set to value which do not work on buildd (wrong JAVA_HOME).

I am not saying that uploading only source package will solve all the
problems. But from my personal experience (in Ubuntu) it has worked
better. Of course same rules may not apply to Debian.


Onkar


Reply to: