[Pkg-gmagick-im-team] Bug#791077: transition: libmagick++-6.q16-5v5

[Simon McVittie]

On 14/08/15 18:02, Julien Cristau wrote:
> On Fri, Aug 14, 2015 at 15:58:30 +0100, Simon McVittie wrote:
>> This might also be a good time to check whether the mips buildd has
>> got stuck.
>>
>
> $ ps u 24745  | cat
> USER       PID %CPU %MEM    VSZ   RSS TTY      STAT START   TIME COMMAND
> buildd   24745  184  0.2  66728 20748 ?        RNl  16:37  45:38 /build/imagemagick-4he8Q1/imagemagick-6.8.9.9/debian/build-quantum-Q16/utilities/.libs/lt-convert /build/imagemagick-4he8Q1/imagemagick-6.8.9.9/debian/display-im6.svg -background none -define filter:blur=0.75 -filter Gaussian -resize 32x32 -gravity center -extent 32x32 /build/imagemagick-4he8Q1/imagemagick-6.8.9.9/debian/tmp-Q16/usr/share/icons/hicolor/32x32/apps/display-im6.q16.png

This looks like <https://bugs.debian.org/770009>. It has been "fixed" in
experimental by printing progress messages more often so the buildd
won't kill it after 5 hours:
<http://anonscm.debian.org/cgit/collab-maint/imagemagick.git/commit/?h=debian/6.9.1.2-1&id=a81c49cee902587ba232aa2e11c71a8d1441e1a6>

This results in using 100% of 2 cores for more than 5 hours, which is
clearly not how things should be. I don't think that change is a valid
solution to this problem.

I'll see whether I can reproduce a similar thing on minkus.d.o with
pre-built imagemagick binaries; if I can, I'll try the possible
workarounds that were suggested on #770009.

debian/rules says:

# build icons cache (build for each arch in order to get more testing)

Well... yes, this has indeed obtained more testing, and the result of
that testing is "it's broken on mips". Where do we go from here? I don't
think we should let mips hold up the libstdc++ transition, and the test
suite has already been disabled on that architecture because it was
breaking security builds.

ImageMagick maintainers: is it feasible to skip building those icons on
mips? They seem to end up in an Architecture:all package.

Regards,
    S