bug#19733: disfunctional gcc binary when GCJ or gfortran is installed

[ludo@gnu.org (Ludovic Courtès)]

Ricardo Wurmus <ricardo.wurmus@mdc-berlin.de> skribis:

> Ludovic Courtès <ludo@gnu.org> writes:
>
>> ludo@gnu.org (Ludovic Courtès) skribis:
>>
>>> Ricardo Wurmus <ricardo.wurmus@mdc-berlin.de> skribis:
>>>
>>>> Commit 5f6887e8 fixes this for GCJ, but we still have this problem for
>>>> all variants of gfortran, gcc-objc, and gccgo, all of which are built
>>>> using the ‘custom-gcc’ procedure.
>>>>
>>>> It’s probably safe to add a build phase like this to ‘custom-gcc’:
>>>>
>>>>     (add-after 'install 'remove-broken-or-conflicting-files
>>>>      (lambda* (#:key outputs #:allow-other-keys)
>>>>        (for-each delete-file
>>>>          (find-files (string-append (assoc-ref outputs "out") "/bin")
>>>>                      ".*(c\\+\\+|cpp|g\\+\\+|gcc.*)"))
>>>>        #t))
>>>
>>> Sounds like it should work.  The pattern should also include ‘gcov’.
>>
>> Could you give it a try?
>>
>> Ludo’, trying to tidy up the bug database.  :-)
>
> Wouldn’t this cause a rebuild of a very large number of packages?

--8<---------------cut here---------------start------------->8---
$ GUIX_PACKAGE_PATH= guix refresh -l -e '(@@ (gnu packages gcc) gfortran)'
Building the following 57 packages would ensure 144 dependent packages are rebuilt: bless-1p02 fftw-openmpi-3.3.4 mumps-metis-5.0.1 mumps-5.0.1 r-gridbase-0.4-7 r-plotrix-3.6 r-servr-0.2 r-htmlwidgets-0.5 r-readr-0.2.2 r-lattice-0.20-33 r-data-table-1.9.6 r-dplyr-0.4.3 r-devtools-1.9.1 python2-rpy2-2.6.0 python-rpy2-2.6.0 r-qtl-1.37-11 rsem-1.2.20 python-ipython-3.2.1 python-numexpr-2.4.4 python-h5py-2.4.0 python-biopython-1.66 python-statsmodels-0.6.1 python-scikit-learn-0.16.1 python-scikit-image-0.11.3 python-seaborn-0.5.1 idr-2.0.0 python2-scikit-image-0.11.3 python2-numexpr-2.4.4 python2-statsmodels-0.6.1 python2-seaborn-0.5.1 python2-ipython-3.2.1 enblend-enfuse-4.1.3 libreoffice-5.0.3.2 macs-2.1.0.20140616 rseqc-2.6.1 crossmap-0.2.1 deeptools-1.5.11 miso-0.5.3 grit-2.0.2 seqmagick-0.6.1 clipper-0.3.0 python2-warpedlmm-0.21 pbtranscript-tofu-2.2.3.8f5467fe6 superlu-dist-3.3 gmsh-2.8.4 mumps-openmpi-5.0.1 slepc-complex-openmpi-3.6.2 plink-1.07 apl-1.5 dealii-openmpi-8.2.1 dealii-8.2.1 flann-1.8.4 slepc-complex-3.6.2 slepc-3.6.2 shogun-4.0.0 couger-1.8.2 julia-0.3.10
$ GUIX_PACKAGE_PATH= guix refresh -l gfortran gccgo gcc-objc gcc-objc++ 
No dependents other than themselves: gcc-objc++-4.8.5 gcc-objc-4.8.5 gccgo-4.8.5 gfortran-5.3.0
--8<---------------cut here---------------end--------------->8---

So that’s mostly several GCC rebuilds and a LibreOffice rebuild.  I
think it’s not unreasonable to do on ‘master’ once ‘security-updates’
has been merged (hopefully soon.)

Ludo’.