all messages for Guix-related lists mirrored at yhetil.org
 help / color / mirror / code / Atom feed
From: Andreas Enge <andreas@enge.fr>
To: 63413@debbugs.gnu.org
Subject: bug#63413: Stop and restart builds in cuirass
Date: Wed, 10 May 2023 12:28:55 +0200	[thread overview]
Message-ID: <ZFtx50J0ibqqKcaJ@jurong> (raw)

Stopping and starting builds in cuirass is not very comfortable.

When working on core-updates, I stopped builds for some old evaluations
on aarch64, where our build power would not be enough, assuming that many
of the old builds were made obsolete by newer changes. (For instance,
these could be builds in branches that were already merged to master.)

However, this also stopped the builds that were not obsolete. And if the
derivation was unchanged in a later evaluation (or a different branch),
it would be kept as failed and not be restarted. I would argue that the
desirable behaviour would be to try all derivations in a new evaluation,
regardless of whether they were stopped in a previous evaluation.

A use case are feature branches:
Assume x-team and y-team are branches that are developed simultaneously;
now x-team is ready first and gets merged to master. All builds of y-team
are stopped (since there is no point in continuing with builds made obsolete
by changes in x-team), the branch is rebased on master, and now all builds
need to be restarted on the rebuilt branch. There is a button for this,
but it also restarts genuinely failed builds, if I remember correctly.
(In any case it did not solve the problem, but I have forgotten the
details.)

Restarting builds manually requires to take the dependency order into
account. If x->y->z is a dependency chain, x has been stopped, and y or z
are restarted, they will immediately fail since x is not available;
otherwise said, restarting manually requires a manual traversal of the
dependency graph, which should be automated.

It would also be a useful feature to restart all builds of dependent
packages: If x is "repaired" and manually started, it would be nice to
be able to start all its dependents (ordered automatically by cuirass)
to see whether they now succeed, instead of forcing us to manually
traverse the graph again.
For instance, we recently got a report on guix-devel of "no space left
on device" for icedtea@3. This would be easy to solve by a "guix gc"
on the build machines, but we currently have no way of restarting the
many builds dependent on icedtea@3 other than restarting all packages.

Andreas





             reply	other threads:[~2023-05-10 10:30 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-05-10 10:28 Andreas Enge [this message]
2024-06-04 13:59 ` bug#63413: Stop and restart builds in cuirass Ludovic Courtès
2024-06-05 13:29   ` Andreas Enge

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=ZFtx50J0ibqqKcaJ@jurong \
    --to=andreas@enge.fr \
    --cc=63413@debbugs.gnu.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
Code repositories for project(s) associated with this external index

	https://git.savannah.gnu.org/cgit/guix.git

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.