Bug#723644: scilab: FTBFS on i386 [was Re: Bug#723644: scilab: FTBFS when building architecture-independent packages]

[Sylvestre Ledru]

Le 13/10/2013 12:47, Sébastien Villemot a écrit :
> Control: retitle -1 scilab: FTBFS on i386: File "/«PKGBUILDDIR»/modules/scicos/macros/scicos_utils/with_modelica_compiler.bin" does not exist.
>
> Le mercredi 18 septembre 2013 à 13:43 +0100, Colin Watson a écrit :
>> Package: scilab
>> Version: 5.4.1-4
>> Severity: serious
>>
>> If you try to build scilab in a clean chroot with "sbuild -A" so that it
>> builds architecture-independent packages, it fails to build:
>>
>>   Adding cdbs dependencies to debian/scilab-doc.substvars
>>   dh_installdirs -pscilab-doc
>>   mkdir /«PKGBUILDDIR»/debian/scilab-doc/usr/share/scilab/modules/helptools/jar/
>>   EN_US_DOC=/«PKGBUILDDIR»/debian/tmp/usr/share/scilab/modules/helptools/jar/scilab_en_US_help.jar; \
>>           mv $EN_US_DOC /«PKGBUILDDIR»/debian/scilab-doc/usr/share/scilab/modules/helptools/jar/
>>   mv: cannot stat '/«PKGBUILDDIR»/debian/tmp/usr/share/scilab/modules/helptools/jar/scilab_en_US_help.jar': No such file or directory
>>   make: *** [install/scilab-doc] Error 1
> Actually the problem is not so much the building of arch-indep packages,
> since "debuild -A" works on amd64.
>
> The problem is rather specific to i386, where even a full build (with
> source, binary-arch and binary-indep) fails.
>
> The relevant part of the build log seems to be the following:
>
> genlib: Regenerate names and lib
> if with_modelica_compiler() then
>      !--error 241 
> File "/«PKGBUILDDIR»/modules/scicos/macros/scicos_utils/with_modelica_compiler.bin" does not exist.
> at line      28 of exec file called by :    
>   exec('buildmacros.sce');
> at line      51 of exec file called by :    
>     exec('buildmacros.sce');
> at line      46 of exec file called by :    
> exec('modules/functions/scripts/buildmacros/buildmacros.sce',-1)
>
I think it is an optimisation issue created by gfortran 4.8.
I reported it here:
http://gcc.gnu.org/ml/fortran/2013-08/msg00077.html
(but I haven't been able to produce a tst case).

I am going to try to build in experimental forcing the usage of gfortran
4.7.

Sylvestre