On Sun, Dec 30, 2007 at 10:42:38AM -0700, [EMAIL PROTECTED] wrote: > Package: tbb > Version: 2.0r014-2 > Severity: serious > > There was an error while trying to autobuild your package: > > > Automatic build of tbb_2.0r014-2 on bld-3 by sbuild/hppa 98 > > Build started at 20071230-0701 >
This package should not be built on hppa. The error you are seeing is
deceptive because I am already patching the package to use 'uname -m' in
place of arch. Please ensure that the package is not being built on
architectures besides i386, amd64 and ia64.
Regards,
-Roberto
Here is the relevant IRC log:
14:49 < el_cubano> Could someone look at #446023 and maybe explain why a
package which is i386 ia64 amd46 all is being
built on hppa?
14:50 < Q_> el_cubano:
14:50 < Q_> Sure you're talking about the right bug number?
14:51 < el_cubano> Of course it's the wrong bug number :-)
14:51 < el_cubano> 458366
14:51 < Q_> That doesn't look like it at all. :)
14:51 < adamm> hey, tbb :) hehe
14:51 < vorlon> short answer is, buildds don't trust maintainers to set arch:
strings right, they use
Packages-arch-specific.
14:51 < el_cubano> Well, in 458366 lamont points me to 446023
14:52 < el_cubano> vorlon: So, what do I do about it. Can I just close the bug?
14:52 < vorlon> as lamont is one of the maintainers of Packages-arch-specific,
you should try to come to an agreement
with him about whether this package should be built
14:52 < phil> vorlon: And dpkg-buildpackage breaks late, not early, right?
14:52 < adamm> Architecture: any
14:53 < adamm> http://ftp.de.debian.org/debian/pool/main/t/tbb/tbb_2.0r014-2.dsc
14:53 < adamm> could that be the problem?
14:53 < vorlon> no
14:53 < phil> Well they are i386 amd64 ia64 in control.
14:53 < vorlon> adamm: well, it's /a/ problem, but it's what dpkg-source does
when you have an arch: all package plus
an arch: foo bar baz package
14:53 < Q_> I'm guessging that tbb-examples is arch all
14:53 < phil> But I thought that this is checked at some time in build and then
it just breaks.
14:53 < vorlon> Q_: yes
14:53 < el_cubano> Yes, I specify an Architecture stanza for each binary package
14:54 < el_cubano> vorlon: The package will in fact only compile for i386, ia64
or amd64. If it does not detect one of
those architectures, then the build will just fail with an
error.
14:55 < el_cubano> Q_: That's right.
14:55 < Q_> el_cubano: It would be better if it said: "Architecture not
supported" instead of "Architecture not
detected".
14:55 < vorlon> el_cubano: why does it do that? (this is the information that
should be passed on to the P-a-s
maintainers)
14:55 < el_cubano> vorlon: I see. It is a threading library written by intel
and so those are the only three arches it
supports.
14:56 < vorlon> well, ok
14:56 < el_cubano> Q_: I'll take that under advisement and consider patching
the Makefile
14:56 < adamm> The point to the other bug report is irrelevant as it doesn't
apply in this case.
14:56 < el_cubano> adamm: That is correct, espcially since I patch the
Makefiles to use uname instead of arch.
14:56 < el_cubano> lamont just saw that make clean failed and that had to do
with it since it calls arch
14:57 < adamm> dpkg-architecture may be more debian way.
14:57 < adamm> For example, I'm not sure if Hurd i386 or BSD kernels are
supported by tbb.
14:58 < adamm> but that doesn't matter as Architecture: line should suffice.
14:58 < Q_> el_cubano: I suggest you add a check in your rules file and give an
error if it's not a supported arch.
14:58 < el_cubano> Except that is already done by the makefile
15:00 < vorlon> right, I don't see any need for an additional check besides the
one already provided by upstream
15:00 -!- noshadow [EMAIL PROTECTED] has quit [Quit: ...]
15:01 < adamm> I guess there are bugs with build and install scripts, but I
have no idea where to report bugs against
that software. For example, the debian archive installer thingy
will not allow bz2 compressed deb
archives into Sid without predent on pre-Woody dpkg version.
15:01 < el_cubano> vorlon: So, email lamont about Packages-arch-specific ?
15:01 < vorlon> yeah
--
Roberto C. Sánchez
http://people.connexer.com/~roberto
http://www.connexer.com
signature.asc
Description: Digital signature
