[Debian-med-packaging] Bug#907905: garli: FTBFS randomly (failing tests)

[Santiago Vila]

Package: src:garli
Version: 2.1-2
Tags: ftbfs

Hello Andreas et al.

I tried to build this package in buster but it failed:

--------------------------------------------------------------------------------
[...]
 debian/rules build-indep
dh build-indep
   dh_update_autotools_config -i
   dh_autoreconf -i
aclocal: warning: couldn't open directory 'config/m4': No such file or directory
configure.ac:355: warning: AC_LANG_CONFTEST: no AC_LANG_SOURCE call detected in body
../../lib/autoconf/lang.m4:193: AC_LANG_CONFTEST is expanded from...
../../lib/autoconf/general.m4:2601: _AC_COMPILE_IFELSE is expanded from...
../../lib/autoconf/general.m4:2617: AC_COMPILE_IFELSE is expanded from...
configure.ac:355: the top level
configure.ac:355: warning: AC_LANG_CONFTEST: no AC_LANG_SOURCE call detected in body
../../lib/autoconf/lang.m4:193: AC_LANG_CONFTEST is expanded from...
../../lib/autoconf/general.m4:2601: _AC_COMPILE_IFELSE is expanded from...
../../lib/autoconf/general.m4:2617: AC_COMPILE_IFELSE is expanded from...

[... snipped ...]

pass 14: -480.5087   (branch= 0.0000  subset rates= 0.0000)
Looking for minimum length branches...
Final score = -480.5086
Time used = 0 hours, 0 minutes and 0 seconds
Writing site likelihoods for tree 1 ...

#######################################################

Completed 1 replicate search(es) (of 1).
Results:
Replicate 1 : -480.5086       

Parameter estimates:

Partition model subset 1:
	 Model contains no estimated parameters

Partition model subset 2:
	 Model contains no estimated parameters

Partition model subset 3:
	 Model contains no estimated parameters

Treelengths and subset rate multipliers:
          TL   R(1)  R(2)  R(3)
rep 1:  16.376 0.637 1.899 2.724

NOTE: collapsebranches setting ignored when writing and comparing optimized trees...

Writing optimized trees and models to scr.p.mkO.ssr.best.all.tre
***********TEST**************
***Score is 480.50865
***Expected is 477.6848
***SCORE DIFFERENCE IS 2.82385
***ALLOWED ERROR IS 0.05
***Scoring test failed for ./scoring/p.mkO.ssr.conf ***
make[3]: *** [Makefile:425: check-local] Error 1
make[3]: Leaving directory '/<<PKGBUILDDIR>>/tests'
make[2]: *** [Makefile:308: check-am] Error 2
make[2]: Leaving directory '/<<PKGBUILDDIR>>/tests'
make[1]: *** [Makefile:394: check-recursive] Error 1
make[1]: Leaving directory '/<<PKGBUILDDIR>>'
dh_auto_test: make -j1 check VERBOSE=1 returned exit code 2
make: *** [debian/rules:6: build-indep] Error 2
dpkg-buildpackage: error: debian/rules build-indep subprocess returned exit status 2
--------------------------------------------------------------------------------

That's how the build ends and not necessarily the relevant part, but
I've put several build logs here:

https://people.debian.org/~sanvila/build-logs/garli/

I estimate that the failure rate might be around 10% so I leave the severity to you.

Please note, however, that there has been a failure on mips for this
version:

https://buildd.debian.org/status/fetch.php?pkg=garli&arch=mips&ver=2.1-2%2Bb1&stamp=1534400116&raw=0

(but I've not checked if it's the same type of failure).

The tests seem to perform simulations based on random numbers, this is
the perfect recipe to make the failure rate to be a number strictly
greater than zero, which is a pity.

A simple fix would be to use a fixed random seed for those random numbers,
that way at least the random numbers would always be the same.

Thanks.