GNU bug report logs - #70303
[PATCH 0/2] Use guile-final for grafting.

Previous Next

Full log

Message #14 received at 70303 <at> debbugs.gnu.org (full text, mbox):

From: Ludovic Courtès <ludo <at> gnu.org>
To: Efraim Flashner <efraim <at> flashner.co.il>
Cc: Ricardo Wurmus <rekado <at> elephly.net>, Josselin Poiret <dev <at> jpoiret.xyz>,
 Simon Tournier <zimon.toutoune <at> gmail.com>, Mathieu Othacehe <othacehe <at> gnu.org>,
 Tobias Geerinckx-Rice <me <at> tobias.gr>, 70303 <at> debbugs.gnu.org,
 Christopher Baines <guix <at> cbaines.net>
Subject: Re: [bug#70303] [PATCH 0/2] Use guile-final for grafting.
Date: Mon, 15 Apr 2024 22:28:17 +0200

Howdy!

Efraim Flashner <efraim <at> flashner.co.il> skribis:

> Over the years the grafting code has changed bit by bit, with various
> attempts to speed it up.  By switching the grafts to not use parallelism
> in rewriting the leaves we can finally switch the guile-for-grafts to be
> guile-final.  The segfault is still there if we perform the grafts in
> parallel, but I believe it is fast enough that it should be fine to do
> them sequentially.

Could you time the grafting derivation of, say, libreoffice or
ungoogled-chromium?

Typically I’d do it along these lines:

--8<---------------cut here---------------start------------->8---
$ guix build libreoffice
/gnu/store/24is7ypdx6sm56mkclxdx4hyj7yg4smb-libreoffice-7.6.3.1
$ sudo sh -c 'echo 3 > /proc/sys/vm/drop_caches'
$ time guix build libreoffice --check
The following graft will be made:
   /gnu/store/wdjqpxm2kbdvq7qlrzyjxb244zn3s3bv-libreoffice-7.6.3.1.drv
applying 137 grafts for libreoffice-7.6.3.1 ...
grafting '/gnu/store/5flppg3h8y235di2ilr3sx878gfl82db-libreoffice-7.6.3.1' -> '/gnu/store/24is7ypdx6sm56mkclxdx4hyj7yg4smb-libreoffice-7.6.3.1'...
successfully built /gnu/store/wdjqpxm2kbdvq7qlrzyjxb244zn3s3bv-libreoffice-7.6.3.1.drv
successfully built /gnu/store/wdjqpxm2kbdvq7qlrzyjxb244zn3s3bv-libreoffice-7.6.3.1.drv
/gnu/store/24is7ypdx6sm56mkclxdx4hyj7yg4smb-libreoffice-7.6.3.1

real    0m14.921s
user    0m7.588s
sys     0m0.389s
--8<---------------cut here---------------end--------------->8---

(That’s on my 4-core i7.)

It’s a bummer that the segfault is still there.  I remember week-long
‘rr’ debugging sessions in the past, where I did find a few issues; I
should try again but uh…

> There's probably room in replace-store-references in (guix build graft)
> for changes if we're not going to use guile-2.0 here anymore.

I believe ‘tests/grafts.scm’ may run some of this code under Guile 2.0
(using the ‘guile-bootstrap’ tarball provided for the tests).

Thanks,
Ludo’.

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.