HACK: enable backtraces in tests to find out what's going wrong

[fingolfin]

I've recently update master with 4e38aef to fix the Travis CI tests. This fixed the tests against GAP's stable-4.11; there are still errors against GAP's master branch, due to gap-system/gap#3889, but that'll be resolved elsewhere soon.

But there are different errors when using GAP 4.10 and 4.9, namely a recursion depth trap. So this HACK PR enables error backtraces to help narrow down the issue.

[codecov]

Codecov Report

Merging #7 into master will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master       #7   +/-   ##
=======================================
  Coverage   45.39%   45.39%           
=======================================
  Files          27       27           
  Lines        3049     3049           
=======================================
  Hits         1384     1384           
  Misses       1665     1665

[fingolfin]

Hmm, I am seeing this backtrace, but it doesn't really ring a bell:

... at /home/travis/gap/lib/ffe.gi:467 called from 
GF( Characteristic( gens ), DegreeFFE( gens ) ) at /home/travis/gap/lib/ffe.gi:1072 called from
DefaultRingByGenerators( arg[1] ) at /home/travis/gap/lib/ring.gi:412 called from
DefaultRing( coeff ) at /home/travis/gap/lib/ratfun1.gi:47 called from
LaurentPolynomialByExtRepNC( fam, fc, val, indn ) at /home/travis/gap/lib/ratfunul.gi:591 called from
QuotRemLaurpols( a, b, 2 ) at /home/travis/gap/lib/ratfunul.gi:946 called from
g mod m at /home/travis/gap/lib/ratfunul.gi:738 called from
PowerMod( g, (Characteristic( br ) ^ (DegreeOverPrimeField( br ) * d) - 1) / 2, f ) at /home/travis/gap/lib/polyfinf.gi:67 called from
FactorsCommonDegreePol( R, Quotient( R, f, g ), d ) at /home/travis/gap/lib/polyfinf.gi:73 called from
FactorsCommonDegreePol( R, Quotient( R, f, g ), d ) at /home/travis/gap/lib/polyfinf.gi:73 called from
FactorsCommonDegreePol( R, g, d ) at /home/travis/gap/lib/polyfinf.gi:74 called from
FactorsCommonDegreePol( R, Quotient( R, f, g ), d ) at /home/travis/gap/lib/polyfinf.gi:73 called from
FactorsCommonDegreePol( R, Quotient( R, f, g ), d ) at /home/travis/gap/lib/polyfinf.gi:73 called from
FactorsCommonDegreePol( R, g, d ) at /home/travis/gap/lib/polyfinf.gi:74 called from
FactorsCommonDegreePol( R, g, d ) at /home/travis/gap/lib/polyfinf.gi:74 called from
FactorsCommonDegreePol( R, Quotient( R, f, g ), d ) at /home/travis/gap/lib/polyfinf.gi:73 called from
FactorsCommonDegreePol( R, Quotient( R, f, g ), d ) at /home/travis/gap/lib/polyfinf.gi:73 called from
FactorsCommonDegreePol( R, Quotient( R, f, g ), d ) at /home/travis/gap/lib/polyfinf.gi:73 called from
FactorsCommonDegreePol( R, g, d ) at /home/travis/gap/lib/polyfinf.gi:74 called from
FactorsCommonDegreePol( R, gcd, deg ) at /home/travis/gap/lib/polyfinf.gi:131 called from
FactorsSquarefree( R, k, opt ) at /home/travis/gap/lib/polyfinf.gi:256 called from
...  at stream:1

[olexandr-konovalov]

At least not in master and stable-4.11, so the problem may simply disappear with GAP 4.11 release...

[fingolfin]

Forced a deeper backtrace, and now am seeing this, repeated over and over; so I am now suspecting that this is due to a bug in cryst that was fixed between the version in GAP 4.10 (Cryst 4.1.20) and that in GAP 4.11 (Cryst 4.1.23).

...
PointGroup( S ) at /home/travis/gap/pkg/cryst/gap/cryst.gi:499 called from
x in H at /home/travis/gap/lib/ghom.gi:309 called from
func( elm ) at /home/travis/gap/lib/coll.gi:1530 called from
ForAll( imgs, function ( x )
      return x in H;
  end ) at /home/travis/gap/lib/ghom.gi:309 called from
GroupGeneralMappingByImagesNC( G, H, gens, imgs  ) at /home/travis/gap/lib/ghom.gi:439 called from
GroupHomomorphismByImagesNC( N, S, perms, Sgens  ) at /home/travis/gap/pkg/cryst/gap/cryst.gi:592 called from
PointGroupHomomorphism( S  ) at /home/travis/gap/pkg/cryst/gap/cryst.gi:609 called from
PointGroup( S ) at /home/travis/gap/pkg/cryst/gap/cryst.gi:499 called from
x in H at /home/travis/gap/lib/ghom.gi:309 called from
...

I suspect that this might be the fix in cryst that matters: gap-packages/cryst@c5e97a7. If my guess is right, then one way to resolve the issue is to reduce the assertion level before the affected tests. Another is to install a newer cryst version.

[fingolfin]

I guess the easiest fix is to not test against 4.9 and 4.10... shrug