Re: Continuous integration - automatic EMAIL

[zimoun]

Hi,

On Mon, 12 Oct 2020 at 14:06, Ludovic Courtès <ludo@gnu.org> wrote:

> >> Please, can we have the build servers send build failures to guix-devel
> >> instead of hoping that people check manually?  I have other things to do
> >> in my life than to poll random servers every few hours.
> >
> > That feature is definitely on my list. Fixing Cuirass and improving the
> > build throughput is already a hard task, but I'm getting there. In the
> > meantime, if people want to join Cuirass debugging party, they are more
> > than welcome!

[...]

> Caveat: my experience with Hydra is that you immediately receive too
> much mail.  Initially Hydra would send one message per failed build,
> which was then changed to one message at each status change (from
> “success” to “failure” and vice versa), but that was still too much.  I
> think eventually it was change to email only the committers of the
> offending commits, which is probably the best option.

If "guix weather -m" is improved, does it fill the gap?  Because it is
exactly doing what is required, isn't?  But it is not suitable for
this purpose because of UI.  I mean:

--8<---------------cut here---------------start------------->8---
$ cat /tmp/spec.scm
(specifications->manifest (list "emacs" "gcc-toolchain" "python-umap-learn"))

$ guix weather -m /tmp/spec.scm
computing 49 package derivations for x86_64-linux...
looking for 66 store items on https://ci.guix.gnu.org...
https://ci.guix.gnu.org
  97.0% substitutes available (64 out of 66)
  at least 340.7 MiB of nars (compressed)
  534.1 MiB on disk (uncompressed)
  0.001 seconds per request (0.0 seconds in total)
  1,768.0 requests per second
  'https://ci.guix.gnu.org/api/queue?nr=1000' returned 504 ("Gateway Time-out")
--8<---------------cut here---------------end--------------->8---

The information is here but I do not know which ones are fine and
which ones are failing.  If an option like "--raw" (or "--plain" or
"--exhaustive" or "--name-it" :-)) could display the status of all the
66 packages, then I think it would ease the detection of the
regresion.  For example,

  OK: emacs
  FAIL: python-umap-learn
  [...]

then it is doable to have a shell script parsing this output and I can
feed "git bisect".  Somehow.

WDYT?

All the best,
simon