[PATCH] gnu: gdk-pixbuf: Disable pixbuf-scale test.

[PATCH] gnu: gdk-pixbuf: Disable pixbuf-scale test.

[Jookia]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1311 bytes --]

On systems with little ram (2G in my case) the pixbuf-scale test will either
freeze the system of cause excessive swapping without the test every completing.

* gnu/packages/gtk.scm (gdk-pixbuf): Don't run 'pixbuf-scale' test.
---
 gnu/packages/gtk.scm | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/gnu/packages/gtk.scm b/gnu/packages/gtk.scm
index 916873b..021ee00 100644
--- a/gnu/packages/gtk.scm
+++ b/gnu/packages/gtk.scm
@@ -8,6 +8,7 @@
 ;;; Copyright © 2015 Sou Bunnbu <iyzsong@gmail.com>
 ;;; Copyright © 2015 Andy Wingo <wingo@igalia.com>
 ;;; Copyright © 2015 David Hashe <david.hashe@dhashe.com>
+;;; Copyright © 2016 Jookia <166291@gmail.com>
 ;;;
 ;;; This file is part of GNU Guix.
 ;;;
@@ -407,6 +408,8 @@ highlighting and other features typical of a source code editor.")
              ;; SKIP Not enough memory to load bitmap image
              ;; ERROR: cve-2015-4491 - too few tests run (expected 4, got 2)
              (("cve-2015-4491\\$\\(EXEEXT\\) ") "")
+             ;; XXX FIXME: This test can crash systems with little RAM.
+             (("pixbuf-scale\\$\\(EXEEXT\\) ") "")
              ;; XXX FIXME: This test fails with:
              ;; ERROR:pixbuf-jpeg.c:74:test_type9_rotation_exif_tag:
              ;; assertion failed (error == NULL): Data differ
-- 
2.7.0

Re: [PATCH] gnu: gdk-pixbuf: Disable pixbuf-scale test.

[Andreas Enge]

On Tue, Feb 02, 2016 at 09:25:41PM +0000, Jookia wrote:
> On systems with little ram (2G in my case) the pixbuf-scale test will either
> freeze the system of cause excessive swapping without the test every completing.

The patch looks good to me, but it would be nice to have another opinion.

In any case, it should probably be applied once we have a bit more build
power, since it causes substantial package rebuilds.

Andreas

Re: [PATCH] gnu: gdk-pixbuf: Disable pixbuf-scale test.

[Leo Famulari]

On Sat, Feb 06, 2016 at 04:17:49PM +0100, Andreas Enge wrote:
> On Tue, Feb 02, 2016 at 09:25:41PM +0000, Jookia wrote:
> > On systems with little ram (2G in my case) the pixbuf-scale test will either
> > freeze the system of cause excessive swapping without the test every completing.
> 
> The patch looks good to me, but it would be nice to have another opinion.

I would like another option besides disabling the test unconditionally.
We already have several packages that can simply not be built on
hardware like Jookia's, and although it's an unfortunate situation, I
don't think we should start disabling tests for the substitutes we build
for this reason. Running all the tests possible allows us to provide
valuable feedback to upstream.

How about a comment in the package definition describing the RAM
requirement and how to disable the test, so that users on less powerful
hardware can make the change themselves?

> 
> In any case, it should probably be applied once we have a bit more build
> power, since it causes substantial package rebuilds.
> 
> Andreas
> 
> 

Re: [PATCH] gnu: gdk-pixbuf: Disable pixbuf-scale test.

[Jookia]

On Sat, Feb 06, 2016 at 04:17:49PM +0100, Andreas Enge wrote:
> On Tue, Feb 02, 2016 at 09:25:41PM +0000, Jookia wrote:
> > On systems with little ram (2G in my case) the pixbuf-scale test will either
> > freeze the system of cause excessive swapping without the test every completing.
> 
> The patch looks good to me, but it would be nice to have another opinion.
> 
> In any case, it should probably be applied once we have a bit more build
> power, since it causes substantial package rebuilds.
> 

Perhaps substantial package rebuild patches should be grouped in to a staging
branch? This would also help those that do their own builds (I'm biased.)

> Andreas

Jookia

Re: [PATCH] gnu: gdk-pixbuf: Disable pixbuf-scale test.

[Leo Famulari]

On Sun, Feb 07, 2016 at 08:04:39AM +1100, Jookia wrote:
> On Sat, Feb 06, 2016 at 04:17:49PM +0100, Andreas Enge wrote:
> > On Tue, Feb 02, 2016 at 09:25:41PM +0000, Jookia wrote:
> > > On systems with little ram (2G in my case) the pixbuf-scale test will either
> > > freeze the system of cause excessive swapping without the test every completing.
> > 
> > The patch looks good to me, but it would be nice to have another opinion.
> > 
> > In any case, it should probably be applied once we have a bit more build
> > power, since it causes substantial package rebuilds.
> > 
> 
> Perhaps substantial package rebuild patches should be grouped in to a staging
> branch? This would also help those that do their own builds (I'm biased.)

If I understand your suggestion correctly, we already do this.

When we need to update a core package like glibc that will cause mass
rebuilds, we put the change in another branch, and build that branch.
Once it is built, we merge the branch into master. That way, the binary
substitutes are already built when users start requesting them.

Sorry if that is not totally accurate. I am sure someone will correct me
in that case :)

> 
> > Andreas
> 
> Jookia
> 

Re: [PATCH] gnu: gdk-pixbuf: Disable pixbuf-scale test.

[Andreas Enge]

On Sat, Feb 06, 2016 at 03:56:24PM -0500, Leo Famulari wrote:
> I would like another option besides disabling the test unconditionally.
> We already have several packages that can simply not be built on
> hardware like Jookia's, and although it's an unfortunate situation, I
> don't think we should start disabling tests for the substitutes we build
> for this reason. Running all the tests possible allows us to provide
> valuable feedback to upstream.

I basically agree with your comment, which is also in line with our policy
of following upstream as closely as possible when building packages.
On the other hand, a simple unit test that requires 2GB of memory seems a bit
excessive to me and could be considered a bug in the upstream package.
How about reporting a bug upstream, Jookia? And maybe disable the test
in the package with a comment linking to the bug report?

Andreas

Re: [PATCH] gnu: gdk-pixbuf: Disable pixbuf-scale test.

[Jookia]

On Sun, Feb 07, 2016 at 12:15:37PM +0100, Andreas Enge wrote:
> I basically agree with your comment, which is also in line with our policy
> of following upstream as closely as possible when building packages.
> On the other hand, a simple unit test that requires 2GB of memory seems a bit
> excessive to me and could be considered a bug in the upstream package.
> How about reporting a bug upstream, Jookia? And maybe disable the test
> in the package with a comment linking to the bug report?

Sure, I'm down for this. I'll see what I can do this week if not tomorrow.

> Andreas

Jookia.

Re: [PATCH] gnu: gdk-pixbuf: Disable pixbuf-scale test.

[Ludovic Courtès]

Jookia <166291@gmail.com> skribis:

> On systems with little ram (2G in my case) the pixbuf-scale test will either
> freeze the system of cause excessive swapping without the test every completing.

Out of curiosity, did you check how the test manages to achieve this
level or memory greediness?

Ludo’.

Re: [PATCH] gnu: gdk-pixbuf: Disable pixbuf-scale test.

[Leo Famulari]

On Sun, Feb 07, 2016 at 12:15:37PM +0100, Andreas Enge wrote:
> On Sat, Feb 06, 2016 at 03:56:24PM -0500, Leo Famulari wrote:
> > I would like another option besides disabling the test unconditionally.
> > We already have several packages that can simply not be built on
> > hardware like Jookia's, and although it's an unfortunate situation, I
> > don't think we should start disabling tests for the substitutes we build
> > for this reason. Running all the tests possible allows us to provide
> > valuable feedback to upstream.
> 
> I basically agree with your comment, which is also in line with our policy
> of following upstream as closely as possible when building packages.
> On the other hand, a simple unit test that requires 2GB of memory seems a bit
> excessive to me and could be considered a bug in the upstream package.
> How about reporting a bug upstream, Jookia? And maybe disable the test
> in the package with a comment linking to the bug report?

True, depending on the nature of the software it could be considered a
bug. There are some programs, like compression algorithm
implementations, where I think such tests can be appropriate, however.

> 
> Andreas
> 

Re: [PATCH] gnu: gdk-pixbuf: Disable pixbuf-scale test.

[Jookia]

On Sun, Feb 07, 2016 at 09:06:20PM +0100, Ludovic Courtès wrote:
> Jookia <166291@gmail.com> skribis:
> 
> > On systems with little ram (2G in my case) the pixbuf-scale test will either
> > freeze the system of cause excessive swapping without the test every completing.
> 
> Out of curiosity, did you check how the test manages to achieve this
> level or memory greediness?

I hadn't dug in to the code any further than I should've as I was unable to run
the test suite outside the automatic guix builder, meaning any change I made to
test the tests would require me rebuilding the entire package.

> Ludo’.

Jookia.