--- Begin Message ---
Subject: |
Memory leak in 'guix pull' or 'make' in guix source |
Date: |
Thu, 8 Jun 2017 08:39:35 +0000 |
This has been discussed on IRC and on guix-devel mailinglist
but so far I haven't seen a formal bug report to make this
visible.
I can't find the right threads or timestamps for this
bug, so take it as is and extend with facts.
At the moment of writing, guix pull has the minimum
requirement of 3 GB RAM.
Before this started, I was able to run Guix on as
little as 512MB RAM.
People are currently finding out about this bug
the hard way, so we should point this out. Didn't
really happen so I guess someone is working on this
already.
--
ng0
OpenPG: A88C8ADD129828D7EAC02E52E22F9BBFEE348588
signature.asc
Description: PGP signature
--- End Message ---
--- Begin Message ---
Subject: |
Re: bug#27284: ‘guix pull’ builds using multiple derivations |
Date: |
Sun, 08 Apr 2018 18:37:07 +0200 |
User-agent: |
Gnus/5.13 (Gnus v5.13) Emacs/25.3 (gnu/linux) |
Hello!
address@hidden (Ludovic Courtès) skribis:
> The code in ‘wip-pull-multiple-derivations’ works like this:
>
> 1. Assume we have Guile and Guix already installed, but not
> necessarily the latest versions thereof.
>
> 2. Using the (guix …) modules that we have, build a program—a
> “trampoline”—that will use the modules of the target Guix (the
> commit we want to pull) to compute the derivation of that Guix.
>
> 3. Run that trampoline, which returns /gnu/store/…-guix.drv. The
> result should be the same regardless of the initial Guix because
> the trampoline uses exclusively modules from the target Guix.
>
> 4. ‘guix pull’ builds that derivation (actually the branch does not
> modify (guix scripts pull) at all; everything is in
> build-aux/build-self.scm.)
After quite a bit of tweaking I’ve finally pushed this patch series to
master. So if you run ‘guix pull’ now, you’ll get the new
multiple-derivation build.
There’s still room for improvement, in particular:
• The initial ‘compute-guix-derivation’ program depends on quite a few
modules, so the first time you ‘guix pull’ you build a
module-import-compiled.drv that takes a while and does not produce
any output (I’ve made it verbose in core-updates commit
d32922759bfeffa03ee189158ea00b1a0ddbe8c6 though). Perhaps we could
avoid that by using a double trampoline, which would allow us to get
substitutes for all the non-trivial bits.
• The “guix-packages” derivation is coarse-grain. We could perhaps
automatically split it in clusters of modules by analyzing the (gnu
packages …) module graph.
• The compiler (and interpreter) in Guile 2.2 is still not as fast as
we’d like. I think it should be possible to make it as fast as in
2.0, at least when optimizations are turned off. Andy made
significant improvements in 2.2.3, but hopefully we can still
improve on that. The thread at
<https://lists.gnu.org/archive/html/guile-devel/2017-10/msg00035.html>
contains some info.
Normally berlin.guixsd.org will start providing substitutes for all this
but note that, as things are, it’ll compute substitutes for Guile 2.2.2
or 2.2.3, and that’s it. hydra.gnu.org currently doesn’t, for obscure
reasons (‘hydra-eval-guile-jobs’ systematically adds “.” to the load
path, which is not what we want for build-aux/hydra/guix-modular.scm.)
This works unlocks some of the stuff in <https://bugs.gnu.org/22629>, so
I’ll try to focus on that now.
Feedback and bug reports welcome!
Ludo’.
--- End Message ---