From: Stefano Tondo <stondo@gmail.com>
To: Mathieu Dubois-Briand <mathieu.dubois-briand@bootlin.com>
Cc: openembedded-core@lists.openembedded.org, stefano.tondo.ext@siemens.com
Subject: Re: [PATCH 3/4] spdx30: Add the PURLs of the source downloads to the output packages
Date: Thu, 19 Nov 2026 00:44:17 +0100 [thread overview]
Message-ID: <20261118234417.3681522-1-stondo@gmail.com> (raw)
In-Reply-To: <DMYFTUY2A6RH.1WI2PDCNCPOJY@bootlin.com>
On Wed Nov 18, 2026 at 4:22 PM CET, Mathieu Dubois-Briand wrote:
> It looks like this is using memory in a way it got killed several times
> on the autobuilder. Locally, I managed to have a do_create_spdx task
> using more than 10GB of memory, with a value continuously increasing of
> about 100MB/s.
>
> This can be reproduced with master branch + your patches, then:
> bitbake python3-uv-build-native -c do_create_spdx
[...]
> Can you have a look at the issue?
Hi Mathieu,
Thanks for the report and the reproducer, and sorry for the broken
builds.
The bug is in patch 2/4, in get_cargo_scopes(). The walk over the
cargo dependency graph did not remember which workspace members it had
already expanded. In uv, the member uv-preview is a dev dependency of
itself. So the walk never ended, and its stack grew until the OOM
killer stopped the task. I could reproduce it: the task reached a 4 GB
limit in about 90 seconds.
python3-uv-build is the only cargo recipe in OE-Core with such a
cycle. The selftests use python3-bcrypt, which has none, so they did
not show it.
v2 expands each member only once and adds unit tests for member
cycles. With v2, do_create_spdx of python3-uv-build and
python3-uv-build-native takes less than 2 seconds and about 100 MB,
with uv-build 0.13.7 on master and 0.13.8 on master-next. For the
other cargo recipes in OE-Core, the result does not change.
I will send v2 now.
Regards,
Stefano
next prev parent reply other threads:[~2026-11-18 23:44 UTC|newest]
Thread overview: 7+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-18 9:29 [PATCH 0/4] spdx30: identity and purls of dependency downloads Stefano Tondo
2026-11-18 9:29 ` [PATCH 1/4] spdx30: Take name, version and PURL of downloads from the fetcher Stefano Tondo
2026-11-18 9:29 ` [PATCH 2/4] spdx30: Record the lifecycle scope of crate dependencies Stefano Tondo
2026-11-18 9:29 ` [PATCH 3/4] spdx30: Add the PURLs of the source downloads to the output packages Stefano Tondo
2026-11-18 15:22 ` [OE-core] " Mathieu Dubois-Briand
2026-11-18 23:44 ` Stefano Tondo [this message]
2026-11-18 9:29 ` [PATCH 4/4] oeqa/selftest: spdx: Test crate identity, scope and output PURLs Stefano Tondo
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20261118234417.3681522-1-stondo@gmail.com \
--to=stondo@gmail.com \
--cc=mathieu.dubois-briand@bootlin.com \
--cc=openembedded-core@lists.openembedded.org \
--cc=stefano.tondo.ext@siemens.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox