[opensuse-factory] arch-dependent noarch packages
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 tl;dr: it is a bad thing and should be forbidden While working on reproducible builds for openSUSE I found this interesting class of bugs (see "Depends on" list) https://bugzilla.opensuse.org/show_bug.cgi?id=1109534 In several cases those architecture-dependent noarch packages prevented verification of official OBS builds and also produced incorrect build results in certain circumstances. 1218 packages sounds like a pretty large number, but I expect, that most can be fixed with ~20 changes throughout openSUSE (and its upstreams). E.g. most perl, python, java issues could be fixed with 1-2 patches. My checks did not cover noarch subpackages yet, so there might be some more to come. I'd like openSUSE to move to a policy like Fedora where noarch packages must produce the same build result on all architectures. This means that not only must the package build reproducibly, but it must also avoid writing %_lib in its output. Even for relatively harmless cases like xorg-docs where only the man-pages have /usr/lib64 vs /usr/lib reviewing for the impact of diffs is effort. Ciao Bernhard M. -----BEGIN PGP SIGNATURE----- iF0EARECAB0WIQRk4KvQEtfG32NHprVJNgs7HfuhZAUCW6vrnQAKCRBJNgs7Hfuh ZMaFAKCXquSFIMvx/TKno9mOMt/BHMX4lwCg6Q3RZbgnBFscNU1ecVSW4uGwA1o= =DXnZ -----END PGP SIGNATURE----- -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On 27/09/2018 07.44, Olaf Hering wrote:
Am Wed, 26 Sep 2018 20:27:27 +0000 schrieb "Bernhard M. Wiedemann" <bernhardout@lsmod.de>:
must also avoid writing %_lib in its output.
This is already rejected since years. noarch@i586==fail. It happens to work on x86_64.
How is this supposed to be rejected? I was thinking of https://bugzilla.opensuse.org/show_bug.cgi?id=1109755#c1 But there is also https://bugzilla.opensuse.org/show_bug.cgi?id=1109895 There %configure passes %_libdir to the build and some packages decide to store it.
Am Thu, 27 Sep 2018 10:00:09 +0200 schrieb "Bernhard M. Wiedemann" <bernhardout@lsmod.de>:
On 27/09/2018 07.44, Olaf Hering wrote:
Am Wed, 26 Sep 2018 20:27:27 +0000 schrieb "Bernhard M. Wiedemann" <bernhardout@lsmod.de>:
must also avoid writing %_lib in its output. This is already rejected since years. noarch@i586==fail. It happens to work on x86_64. How is this supposed to be rejected?
I was thinking of %files;/usr/lib64, not actual content within files. Sorry. Olaf
On Thursday 2018-09-27 10:13, Olaf Hering wrote:
Am Thu, 27 Sep 2018 10:00:09 +0200 schrieb "Bernhard M. Wiedemann" <bernhardout@lsmod.de>:
On 27/09/2018 07.44, Olaf Hering wrote:
Am Wed, 26 Sep 2018 20:27:27 +0000 schrieb "Bernhard M. Wiedemann" <bernhardout@lsmod.de>:
must also avoid writing %_lib in its output. This is already rejected since years. noarch@i586==fail. It happens to work on x86_64. How is this supposed to be rejected?
I was thinking of %files;/usr/lib64, not actual content within files. Sorry.
Such a noarch-isn't-noarch check was already done some years ago. IIRC by Andres Jaeger(?), and we found something like /usr/include just containing different content, and it was not even due to pathspecs in the file. It's good to see it's being picked up again. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
participants (3)
-
Bernhard M. Wiedemann
-
Jan Engelhardt
-
Olaf Hering