| *** bauzas_ is now known as bauzas | 00:45 | |
| *** bauzas_ is now known as bauzas | 01:10 | |
| *** bauzas_ is now known as bauzas | 05:22 | |
| opendevreview | Rodolfo Alonso proposed openstack/releases master: Release os-ken 2.9.0 https://review.opendev.org/c/openstack/releases/+/922792 | 09:15 |
|---|---|---|
| *** mgoddard- is now known as mgoddard | 10:12 | |
| opendevreview | Merged openstack/releases master: Release Neutron 25.0.0.0b1 (Neutron Dalmatian beta 1) https://review.opendev.org/c/openstack/releases/+/922742 | 11:40 |
| *** tosky_ is now known as tosky | 12:37 | |
| opendevreview | Merged openstack/releases master: Release virtualpdu 2.1.0 https://review.opendev.org/c/openstack/releases/+/922588 | 12:48 |
| frickler | release-team: rpittau: annonce-release failure: "pkg-config binary is required to compile libvirt-python" not sure why that only failed there and not elsewhere? https://zuul.opendev.org/t/openstack/build/fd4b379931934706a64663d1c75f95af | 15:16 |
| frickler | hmm, maybe that job needs to include bindep? https://opendev.org/openstack/virtualpdu/src/branch/master/bindep.txt#L9 | 15:18 |
| rpittau | mmm looks like it | 15:19 |
| frickler | hmm, that job uses the releases bindep. that's not too helpful in this case https://zuul.opendev.org/t/openstack/build/fd4b379931934706a64663d1c75f95af/console#1/0/14/ubuntu-jammy | 15:20 |
| frickler | but I'd be surprised if virtualpdu is the only deliverable with such a dependency. or maybe it is? | 15:21 |
| rpittau | from ironic also virtualbmc and sushy-tools | 15:33 |
| frickler | ah, I think I found the issue. u-c still has libvirt-python===10.3.0, so we have no wheel built for libvirt-python===10.4.0 which is the latest version. not sure if it would be better if that job would actually use upper-constraints though? or rather fix to use the correct bindep? | 15:52 |
| frickler | clarkb: fungi: ^^ maybe you have some idea? | 15:53 |
| clarkb | pip's default is to prefer the newest version of a pacakge even if wheels don't exist for it. You can set a binary only flag to prefer wheel versions. In this case I think that the bindep file is deficient because anyone running this at home won't/shouldn't be using our wheel mirror and they will fail | 15:54 |
| clarkb | its probably best to ensure you can build the wheel locally via bindep deps and then let the wheel caching catch up later | 15:55 |
| fungi | yeah, since libvirt-python is only distributed in sdist/source form, making bindep include correct dependencies for building wheels of it is "the right way" | 15:56 |
| fungi | we pre-build and cache wheels of things in constraints in order to speed up jobs, because there's no point in repeatedly testing that you can build a wheel of the same version of libvirt-python, but it's only meant as a convenience | 15:56 |
| fungi | a developer trying to run tox on their workstation very likely won't be pulling our pre-built wheels from our mirror farm/afs | 15:57 |
| frickler | well the bindep file for virtualpdu is correct, the job however is using bindep from the release repo instead, which seems wrong | 15:59 |
| frickler | or maybe it even needs to use both bindep files in this scenario? | 15:59 |
| fungi | probably, yes. or else the release repo needs a bindep file that contains the union of bindep entries for all projects it's handling | 16:01 |
| *** bauzas_ is now known as bauzas | 17:17 | |
| *** bauzas_ is now known as bauzas | 19:56 | |
| *** bauzas_ is now known as bauzas | 22:55 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!