Re: FTFBS in sarge
Colin Watson <cjwatson@debian.org> writes:
> On Thu, Sep 02, 2004 at 11:05:35AM +0200, Jens Schmalzing wrote:
>> Bastian Blank writes:
>> > http://bblank.thinkmo.de/debian/build-sarge/
>>
>> > kernel-patch-powerpc-2.6.7
>> > kernel-patch-powerpc-2.6.8
>>
>> Both packages build nicely with `dpkg-buildpackage -b' on any arch.
>> However, sbuild is using `dpkg-buildpackage -B', which of course
>> doesn't work, because the source package only creates arch all and
>> arch powerpc patches. Kind of what you'd expect from the PowerPC
>> kernel package, eh?
>>
>> IMO, this is a bug in sbuild.
>
> Do you really think it's a good idea for sbuild to use
> 'dpkg-buildpackage -b'? That builds architecture-independent binaries
> too, which would be a total waste of its time (a considerable waste, for
> many packages).
I actually think it is a bug in the kernel-patch-powerpc package (for
having -all packages no !ppc user wnats) but that is a matter of
opinion.
On a more general note this could be a bug in the archive software
declaring the source to be "Architecture: any" and the deb specs. For
ppc it will work to just build _powerpc.deb instead of _all.deb but
for mips/mipsel that would double some packages. It would be nice if
one could set "Architecture: mips mipsel" for the mips(el) kernel
patch instead of all.
One should make sure that a package not ment for an arch refuses to
build or at least does not enter a loop creating a GB buildd log until
the disk is full like the kernel-image and kernel-patch packages tend
to do.
One way for this is
Build-Depends: type-handling
Builc-Conflicts: not+powerpc
another would be an explicit test in debian/rules.
I prefer the former since automatic Build-Depends checks would detect
this package automatically while the later would give a build-failure.
> --
> Colin Watson [cjwatson@debian.org]
MfG
Goswin
Reply to: