[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
bug#27551: guile-cairo switch to Guile 2.2 breaks (gnu build svg)
From: |
Ludovic Courtès |
Subject: |
bug#27551: guile-cairo switch to Guile 2.2 breaks (gnu build svg) |
Date: |
Sun, 02 Jul 2017 23:01:34 +0200 |
User-agent: |
Gnus/5.13 (Gnus v5.13) Emacs/25.2 (gnu/linux) |
Leo Famulari <address@hidden> skribis:
> On Fri, Jun 30, 2017 at 05:21:08PM -0400, Ludovic Court�s wrote:
>> civodul pushed a commit to branch master
>> in repository guix.
>>
>> commit e3ddb1e83296c10338d35bc687772242b2f5eac6
>> Author: Ludovic Courtès <address@hidden>
>> Date: Fri Jun 30 23:10:46 2017 +0200
>>
>> gnu: guile-cairo: Switch to Guile 2.2.
>>
>> This patch adjusts dependent packages accordingly.
>>
>> * gnu/packages/gtk.scm (guile-cairo)[inputs]: Switch to GUILE-2.2 and
>> GUILE-LIB.
>> (guile-rsvg)[inputs]: Likewise.
>> (guile-present)[inputs]: Likewise.
>> (guile-gnome)[inputs]: Likewise.
>> * gnu/packages/plotutils.scm (guile-charting)[inputs]: Switch go
>> GUILE-2.2.
>
> This commit introduced some problems with (gnu build svg), which
> prevents reconfiguring GuixSD:
[...]
> In ./gnu/build/svg.scm:
> 51: 0 [svg->png
> "/gnu/store/w893227m8wllganabxqx7sn2mgga8083-guix-artwork-6998d30-checkout/grub/GuixSD-fully-black-4-3.svg"
> ...]
>
> ./gnu/build/svg.scm:51:4: In procedure svg->png:
> ./gnu/build/svg.scm:51:4: In procedure module-lookup: Unbound variable:
> rsvg-handle-new-from-file
> builder for `/gnu/store/nldcbaghx031b799mq9bdh8l00hg4fz6-grub-image.png.drv'
> failed with exit code 1
Apologies for not noticing this earlier!
I fixed it in a single commit, 1b0f266e40aead09be95a984bd9c6cec3dff397e,
that reinstates the original commits and fixes the problem. I
considered re-reverting the commits individually but that didn’t seem
very nice either.
The problem was that Guile-RSVG was switched to Guile 2.2 but
derivations are still built with Guile 2.0 by default. Thus, the
derivation that creates “grub-image.png” would not find Guile-RSVG.
The fix I committed is to explicitly build this derivation with 2.2.
I agree with what you wrote: CI could in theory help avoid such
problems, but we lack the resources right now. We’d also need to change
our workflow to commit to a staging branch and have a robot that
automatically merges branches that pass the tests. Not trivial.
For now, reverting the changes promptly was the right thing to do, so
thanks for doing it.
Ludo’.