[pkg-GD-devel] Bug#709050: mscgen: fails to build with current libgd2

[Ondřej Surý]

Hi,

were you able to discover something? I don't see anything in gd, but I did
study it very hard.

Ondrej


On Mon, May 20, 2013 at 4:08 PM, Niels Thykier <niels at thykier.net> wrote:

> Control: tags -1 confirmed help
>
> On 2013-05-20 15:06, Colin Watson wrote:
> > Package: mscgen
> > Version: 0.20-2
> > Severity: serious
> > User: ubuntu-devel at lists.ubuntu.com
> > Usertags: origin-ubuntu saucy
> >
> > mscgen fails to build in unstable as follows:
> >
> >   make  check-TESTS
> >   make[3]: Entering directory `/«PKGBUILDDIR»/build/test'
> >   testinput0.msc
> >   Error: gdoTextHeight: Could not read font (GDFONTPATH=)
> >   FAIL: renderercheck.sh
> >   ====================================================
> >   1 of 1 test failed
> >   Please report to Michael.McTernan.2001 at cs.bris.ac.uk
> >   ====================================================
> >   make[3]: *** [check-TESTS] Error 1
> >
> > If it helps, you can see a corresponding Ubuntu build log here:
> >
> >
> https://launchpadlibrarian.net/140264375/buildlog_ubuntu-saucy-i386.mscgen_0.20-2build2_FAILEDTOBUILD.txt.gz
> >
> > I don't quite know what's happening here, as running the tests by hand
> > seems to succeed.
> >
> > Thanks,
> >
>
> I can confirm the failure in a sid chroot, but the failure is
> non-deterministic.  I have seen 2 successful builds out of 5 or so.
> Furthermore the test that fails is also non-deterministic.  Your log has
> testinput0.msc, but I saw 17, 5 and one starting with 2 (forgot the
> exact number for that one).
>   The frequency of the failures appears to be fairly random at first
> glance (for testinput0.msc).  I have observed 0 to 496 successful runs
> between each failure.  The average number of successes between failures
> appears to be around 36 runs (based on 90 observations).  So anyone
> wanting to replicate this may want to setup a:
>    "while mscgen -Tpng <in-file> ; do : ; done"
>
>
> Dear pkg-gd developers, do you have any hints as to why gdImageStringFT
> would be non-deterministic[1]?  As far as I can tell, it receives the
> following arguments:
>
>   NULL,
>   rect (int array[8], init'ed to 8 zeroes),
>   pen (0 or 0xff - failures tend to be 0),
>   "helvetica",
>   11.0,
>   0,
>   0,
>   0,
>   "gHELLOWt"
>
> The pen value is (AFAICT) always a value returned by
> gdImageColorAllocate.  mscgen will always allocate ADRAW_COL_BLACK and
> ADRAW_COL_WHITE (in that order) prior to loading any other colors[2].
>
> ~Niels
>
> [1] Call context:
>
> http://anonscm.debian.org/gitweb/?p=collab-maint/mscgen.git;a=blob;f=src/gd_out.c;h=64d82b27817dccae83c24ce6a450306be11fcfe4;hb=2ca7e6af2dad9f34d6d1466174b13e646d8ab630#l225
>
> [2]
> src/adraw.h:    ADRAW_COL_WHITE   = 0x00ffffff,
> src/adraw.h:    ADRAW_COL_BLACK   = 0x00000000,
>
> These are passed like:
>     gdImageColorAllocate(context->img,
>                          (col & 0xff0000) >> 16,
>                          (col & 0x00ff00) >>  8,
>                          (col & 0x0000ff) >>  0);
>
> (see the getColorRef function in same file linked to in [1])
>
>
>
> _______________________________________________
> pkg-GD-devel mailing list
> pkg-GD-devel at lists.alioth.debian.org
> http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-gd-devel
>



-- 
Ondřej Surý <ondrej at sury.org>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.alioth.debian.org/pipermail/pkg-gd-devel/attachments/20130523/5709fb71/attachment.html>