--- Begin Message ---
|
Subject: |
Hydra ignores MIPS-specific patch for Mesa on core-updates |
|
Date: |
Mon, 31 Oct 2016 02:17:35 -0400 |
Commit f1267c872fcaed6c53d43b3ff51abb726f7418d6 on core-updates added a
patch to Mesa on MIPS only, in order to prevent unnecessary rebuilds on
other systems. 'guix' running on a MIPS system generates a derivation
that applies the patch, but the derivation generated for MIPS on Hydra
omits the patch.
The corresponding build page on Hydra is:
https://hydra.gnu.org/build/1569732
The derivation for that build is:
/gnu/store/vyqmm1zdhbzxjnaq0h6k312mm7ywi1bq-mesa-12.0.1.drv
which uses a source derivation:
/gnu/store/368ns1adxyv5zhnsn5rx40k6ghd9sr39-mesa-12.0.1.tar.xz.drv
and this derivation simply downloads the source tarball.
In contrast, corresponding derivations generated by 'guix' on a
mips64el-linux machine are:
/gnu/store/1r8m67dn07q2k75gl7rnw09i2kp4a0rd-mesa-12.0.1.drv
/gnu/store/imyjp7sx1rzwpc9agvcnjvk28i4qcgis-mesa-12.0.1.tar.xz.drv
and the latter derivation is a 'patch-and-repack' derivation that
applies 'mesa-wayland-egl-symbols-check-mips.patch', as expected.
Mark
--- End Message ---
--- Begin Message ---
|
Subject: |
Re: bug#24833: Hydra ignores MIPS-specific patch for Mesa on core-updates |
|
Date: |
Thu, 03 Nov 2016 11:29:33 +0100 |
|
User-agent: |
Gnus/5.13 (Gnus v5.13) Emacs/25.1 (gnu/linux) |
Mark H Weaver <address@hidden> skribis:
> address@hidden (Ludovic Courtès)
>> Mark H Weaver <address@hidden> skribis:
>>
>>> Commit f1267c872fcaed6c53d43b3ff51abb726f7418d6 on core-updates added a
>>> patch to Mesa on MIPS only, in order to prevent unnecessary rebuilds on
>>> other systems. 'guix' running on a MIPS system generates a derivation
>>> that applies the patch, but the derivation generated for MIPS on Hydra
>>> omits the patch.
>>
>> Did it turn out to be the same issue as <http://bugs.gnu.org/24833>, or
>> is it something entirely different?
>
> Bug 24833 is this bug, so yes, by reflexivity it's the same issue, but I
> suspect that's not what you meant to ask :)
Arf, sorry, I meant to ask if this is related to
<https://debbugs.gnu.org/cgi/bugreport.cgi?bug=24832>.
Commit f1267c872fcaed6c53d43b3ff51abb726f7418d6 doesn’t have the desired
effect: the ‘patches’ field is voluntarily unthunked, so it gets to see
the ‘%current-system’ value that it gets when gl.scm is loaded, hence
the problem we’re experiencing.
This is fixed in 4c93fe83e89473b2d16b93900c87e0da0ceecd6e by making the
patch an input depending on ‘%current-system’, within an ‘inputs’ field.
Thanks!
Ludo’.
--- End Message ---