[Debian-med-packaging] Bug#919189: Fwd: probabel FTBFS on arm64, likely due to Eigen 3 NEON code

[Andreas Tille]

Hi,

I think these are just rounding errors due to different representation of
floating point numbers.  Relaxing the precision of the test would solve
the problem (but I have no idea about the test suite and so do not know
how to implement this).

What I do also not understand why a fix that only occures on arm64 now
has an influence on ppc64el.  Did ppc64el compiled fine for version
0.5.0+dfsg-1 (without that fix)?

Kind regards

        Andreas.

On Fri, Feb 01, 2019 at 04:41:04PM +0100, Thierry Fauck at linux.ibm.com wrote:
> 
> Unfortunately the workaround mentioned for arm64 breaks ppc64el arch as
> compiling on buster fails with following error - what do we do ?
> Thanks
> 
> ------------------------------------------------------------------------
> BT check (recess model): prob vs. prob_fv                              OK
> 2c2
> < rs7247199 GAC A 0.5847 0.415 0.9299 0.8666 200 0.333517 19 204938
> 544.518 437.583 -567.388 435.04 1.75239
> ---
> > rs7247199 GAC A 0.5847 0.415 0.9299 0.8666 200 0.333517 19 204938
> 544.518 437.583 -567.387 435.04 1.75239
> BT check (2df model): prob vs. prob_fv
> FAILED
> FAIL test_bt.sh (exit status: 1)
> -------------------------------------------------------------------------
> 
> 
> 
> On Sun, 13 Jan 2019 16:37:02 +0200 Adrian Bunk <bunk at debian.org> wrote:
> > Source: probabel
> > Version: 0.5.0+dfsg-1
> > Severity: serious
> > Tags: ftbfs
> > 
> > https://buildd.debian.org/status/fetch.php?pkg=probabel&arch=arm64&ver=0.5.0%2Bdfsg-1&stamp=1538603399&raw=0
> > 
> > ...
> > FAIL: test_bt
> > =============
> > 
> > Analysing BT...
> > BT check: dose vs. dose_fv                                             OK
> > BT check (add model): prob vs. prob_fv                                 OK
> > BT check (domin model): prob vs. prob_fv                               OK
> > BT check (over_domin model): prob vs. prob_fv                          OK
> > BT check (recess model): prob vs. prob_fv                              OK
> > 2c2
> > < rs7247199 GAC A 0.5847 0.415 0.9299 0.8666 200 0.333517 19 204938 544.518 437.583 -567.387 435.04 1.75239
> > ---
> > > rs7247199 GAC A 0.5847 0.415 0.9299 0.8666 200 0.333517 19 204938 544.518 437.583 -567.388 435.04 1.75239
> > BT check (2df model): prob vs. prob_fv                                 FAILED
> > FAIL test_bt.sh (exit status: 1)
> > 
> > ============================================================================
> > Testsuite summary for ProbABEL 0.5.0
> > ============================================================================
> > # TOTAL: 11
> > # PASS:  10
> > # SKIP:  0
> > # XFAIL: 0
> > # FAIL:  1
> > # XPASS: 0
> > # ERROR: 0
> > ============================================================================
> > See checks/test-suite.log
> > Please report to genabel-devel at r-forge.wu-wien.ac.at
> > ============================================================================
> > make[4]: *** [Makefile:713: test-suite.log] Error 1
> > 
> > 
> > I suspect this might be a problem with the NEON code in Eigen 3,
> > and the fact that this can be fixed with the following workaround
> > makes that even more probable:
> > 
> > --- debian/rules.old	2019-01-09 15:07:11.950823327 +0000
> > +++ debian/rules	2019-01-09 15:17:23.280539016 +0000
> > @@ -6,6 +6,10 @@
> >  
> >  export DEB_BUILD_MAINT_OPTIONS=hardening=+all
> >  
> > +ifeq ($(DEB_HOST_ARCH),arm64)
> > +  export DEB_CXXFLAGS_MAINT_APPEND = -march=armv8-a+nosimd
> > +endif
> > +
> >  include /usr/share/dpkg/default.mk
> >  
> >  # Location of the example files, will be converted to the
> > 
> > 
> 
> _______________________________________________
> Debian-med-packaging mailing list
> Debian-med-packaging at alioth-lists.debian.net
> https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/debian-med-packaging

-- 
http://fam-tille.de