Guix substitute timeout

Guix substitute timeout

[Christopher Baines]

[-- Attachment #1: Type: text/plain, Size: 564 bytes --]

Hey,

I'm unsure how well the substituter copes with unreliable network
connections, anecdotally it seems to often hang indefinitely.

Looking at the code [1], there is some code that looks like it would
cancel the substitution after some amount of time, but it also looks to
be disabled (as #:timeout? is #f when fetch is called).

Is this something that was never enabled, or does it need to be enabled
in some way?

Thanks,

Chris

1: https://git.savannah.gnu.org/cgit/guix.git/tree/guix/scripts/substitute.scm?id=1719f4f1e3aba3af60b5f7e63b8db529ef777af5#n1041

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 962 bytes --]

Re: Guix substitute timeout

[Ludovic Courtès]

Hi Chris!

Christopher Baines <mail@cbaines.net> skribis:

> I'm unsure how well the substituter copes with unreliable network
> connections, anecdotally it seems to often hang indefinitely.

How often?

I experienced it at some point last year but the problem vanished soon
after I reported it here: <https://issues.guix.gnu.org/issue/31925>.

If you can reproduce it, could you try to get a C backtrace of ‘guix
substitute’, as well as the output of ‘guix’ to see if it’s fetching
gzip or lzip stuff, and the process sub-tree?

> Looking at the code [1], there is some code that looks like it would
> cancel the substitution after some amount of time, but it also looks to
> be disabled (as #:timeout? is #f when fetch is called).

That comes from commit 2207f73156e144a9349e4d395d5049119b67a896 from
2013 (!), when hydra.gnu.org was so slow and (I think) we’d rather not
time out too quickly while fetching a substitute or we’d never fetch
anything.  Good times…

We could probably enable it, though it should be a not-too-short
timeout.

Thoughts?

Ludo’.