Re: [gentoo-dev] Re: [gentoo-dev-announce] Last rites: dev-build/bazel, sci-libs/keras, sci-libs/tensorflow, sci-libs/tensorflow-estimator, sci-visualization/tensorboard

[Ionen Wolkens <ionen@gentoo.org>]

[-- Attachment #1: Type: text/plain, Size: 1316 bytes --]

On Thu, Jan 18, 2024 at 02:33:08PM +0100, Maciej Barć wrote:
> A lot of Bazel bugs were just left to rot, even though they are invalid.
> There are work from users to get Bazel to a reasonable state, see:
> https://bugs.gentoo.org/918703 (plus comment #1)
> 
> > # Unmasking this requires a sign-off from QA and treecleaners, since
> > # these packages require a ton of mental bandwidth to keep up to date
> 
> I would urge to reconsider leaving Bazel out of this; I do not think 
> that one is lost for now. Since I sometimes use Bazel myself, I will 
> work on adding a bazel-bin + virtual.
> 
> I will NOT however use/support Bazel for Gentoo packages nor eclasses.

Note that the current bazel ebuild uses the eclass. So if that's not
supported by anyone... Albeit seems(?) to only be for bazel_get_flags
so it should be easy to make it go standalone.

Also a virtual for -bin won't matter much if it's not intended for
use with Gentoo packaging. Users don't overly need to go through a
virtual and can just emerge bazel{,-bin} as wanted.

For all I care it could be reduced to a bin-only package if that's
easier and sufficient, not that I use bazel myself for anything to
have much of an opinion here.

That aside, bazel does sound far less bad than keeping tensorflow.
-- 
ionen

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]