bug#64302: Guix derivation cannot be computed during pull

[Andreas Enge <andreas@enge.fr>]

Hello,

Am Mon, Jul 10, 2023 at 11:38:47PM +0200 schrieb Ludovic Courtès:
> Is there any more data you can grab?  Perhaps in
> /var/log/guix-daemon.log or similar if that’s on a foreign distro?

it is a GNU system virtual machine. And I have this in /var/log/messages:

Jul 19 11:57:51 localhost vmunix: [4619013.095327] oom-kill:constraint=CONSTRAINT_NONE,nodemask=(null),cpuset=/,mems_allowed=0,global_oom,task_memcg=/,task=69h614ii57kpgn5,pid=16408,uid=1000
Jul 19 11:57:51 localhost vmunix: [4619013.095350] Out of memory: Killed process 16408 (69h614ii57kpgn5) total-vm:540000kB, anon-rss:433096kB, file-rss:0kB, shmem-rss:0kB, UID:1000 pgtables:952kB oom_score_adj:0

> And as a last resort, ‘strace -f -o /tmp/log -s 500 guix pull’, so we
> can get an idea of what happens.

While this appears in the strace log:

16408 +++ killed by SIGKILL +++

So we have the culprit; not a Guix bug per se.

$ cat /proc/meminfo 
MemTotal:         997864 kB
MemFree:          679044 kB
MemAvailable:     817052 kB

It looks like 1GB of memory is not enough for "guix pull", which is
not nice. Has this been addressed in later commits? Is there a known
lower memory limit for using Guix?

Andreas