--- Begin Message ---
Subject: |
[PATCH 0/2] Split guix build into more steps for 32bit hosts. |
Date: |
Tue, 22 Aug 2023 19:17:59 +0200 |
Hi!
Initially writter for the Hurd (see hurd-team
<https://git.savannah.gnu.org/cgit/guix.git/commit/?h=hurd-team&id=06438c2a7edf1f0c90392858fc86480c6e5c2195>,
<https://git.savannah.gnu.org/cgit/guix.git/commit/?h=hurd-team&id=5623ab503fa70eab95e6f859e82c89581c3f69f4>)
This afternoon on IRC, Maxim confirmed my suspicion that this could be 32bit
issue, rather than a Hurd problem
(<https://logs.guix.gnu.org/guix/2023-08-22.log#182716>,
<https://ci.guix.gnu.org/build/1870536/log/raw>).
I didn't submit these for the Hurd yet, as I didn't really feel comfortable
with the 26-way split. I tried a 5-way split and that still gave OOM errors
on the Hurd and I didn't feel like hand-coding a 10-way split, so yeah.
Also, the 5-way split in the Makefile.am will produce percentages over
100 which suggests a typo...but I couldn't spot it.
Lastly, these patches feel like a workaround for Guile / libgc memory
management? Otoh, being able to build Guix on 32bit hosts is kinda nice
too... So yeah.
Greetings,
Janneke
Janneke Nieuwenhuizen (2):
build: Build gnu/packages/*.go in five steps.
self: Build gnu/packages/*.go in 26 steps on 32bit.
Makefile.am | 62 ++++++++++++++++++++++++++++++++++++++++++++-------
guix/self.scm | 31 ++++++++++++++++++++++++--
2 files changed, 83 insertions(+), 10 deletions(-)
base-commit: c655231b72ac28b5a433069fcf86a835c9c83691
--
2.41.0
--- End Message ---
--- Begin Message ---
Subject: |
Re: bug#65456: [PATCH 0/2] Split guix build into more steps for 32bit hosts. |
Date: |
Mon, 18 Sep 2023 06:52:51 +0200 |
User-agent: |
Gnus/5.13 (Gnus v5.13) Emacs/28.2 (gnu/linux) |
Janneke Nieuwenhuizen writes:
Hi!
> Janneke Nieuwenhuizen writes:
>
>> Ludovic Courtès writes:
> If there are no objections I'll go forward and install this patch by the
> end of the weekend.
So, I've been running guix pull from hurd-team on the Hurd all day.
Starting with the max-25 files split without (gc), it failed 5
consecutive times building guix-packages-base. Because the initial
patch worked pretty well and had the (gc) call (and yes, also used the
26-way split), I re-added (gc), and now the fourth attempt succeeded.
After `guix-packages-base' built, `guix-packages' then built right away.
I seem to remember guix-packages was the main problem when I wrote the
first version of this patch some four months ago, but that one split the
list of packages 26-way of course. That suggests that the (combined)
compilation of certain files is especially problematic, or that it has
since become more problematic (<https://issues.guix.gnu.org/66063>).
On the other hand, we may be able to use more than 4GiB of memory on the
Hurd some time soon (and after we upgrade and compile using
--enable-pae)
<https://lists.gnu.org/archive/html/bug-hurd/2023-08/msg00090.html>.
Pushed to master as 658de25e990a56265e2398a3737e9cf1eb57af68
Greetings,
Janneke
--
Janneke Nieuwenhuizen <janneke@gnu.org> | GNU LilyPond https://LilyPond.org
Freelance IT https://www.JoyOfSource.com | Avatar® https://AvatarAcademy.com
--- End Message ---