IRC channel logs


back to list of logs

<civodul>guix-cran builds have been going happily for a few weeks now:
<zimoun>civodul: do you have some space in Inria for retaining the substitute artificats required for running “guix time-machine --commit=v1.X.0”?
<zimoun>I have hit 3 time-bombs when running you ReScience paper rebuilding all from scratch. And apteryx one with Python 2. Since the fixes for time-bomb are not all ready, maybe we could retain some substitutes (the ones required for running “guix time-machine --commit=v1.X.0 -- help“ could be a start).
<civodul>zimoun: yes, we could do that
<civodul>should, even
<zimoun>apteryx is proposing to add a Cuirass jobset for Berlin. Where would it be the most adequate?
<civodul>there's already one:
<civodul>but it needs to be fixed
<civodul>looks like it never actually worked (there zero jobs, even on the oldest evaluation)
<civodul>*there are
<zimoun>this one will keep the substitutes?
<civodul>well, at least for the configured 'guix publish' TTL
<apteryx>civodul: I think there's more to it than "build it". There's probably some logic in Cuirass that means adding store roots for the requested branch tem built items, otherwise they'd be gargabe collected
<apteryx>I have no idea how it works, but that's my understanding from reading doc/ in guix-maintenance
<apteryx>like, in the past, we'd have a 1.3.0 release branch being built, 1.2.0, etc.
<apteryx>and keeping these branches job specification in Cuirass would cause it to add roots for the corresponding package outputs, to guard them from being garbage collected, IIUC.