Tests fail in stable-4.10 and master GAP branches

olexandr-konovalov commented 6 years ago

Jenkins build of gap-stable-test #162 (Sep 27, 2018 5:55:48 PM) passed, and the next day it failed. There was the only new package (Forms) picked up, and the failed test was based on revision https://github.com/gap-system/gap/commit/c94a3641a80a8a1ad2b1759763aa84d10d9b3763 with the following new commits:

test: tweak tst/testbugfix/2005-08-29-t00318.tst
Fix IsNilpotentGroup: infinite does not imply "not p-group"
FIX: Bad Memory inefficiency in new MTC.
ENHANCE: Better strategy for cyclic subgroup for fp group order
Add IsPackageLoaded
Improve IsPackageLoaded

With no packages:

########> Diff in /circa/scratch/gap-jenkins/workspace/GAP-stable-test/GAPCOPT\
S/64build/GAPTARGET/packages/label/kovacs/GAP-stable-snapshot/pkg/wedderga-4.9\
.4/tst/wedderga07.tst:167
# Input is:
LocalIndicesOfCyclotomicAlgebra(W[27]);
# Expected output:
[ [ infinity, 2 ] ]

# But found:
Error, the coset enumeration has defined more than 4096001 cosets

########

With all and with default packages:

########> Diff in /circa/scratch/gap-jenkins/workspace/GAP-stable-test/GAPCOPT\
S/64build/GAPTARGET/packages/label/kovacs/GAP-stable-snapshot/pkg/wedderga-4.9\
.4/tst/wedderga02.tst:106
# Input is:
WedderburnDecompositionInfo(QG);
# Expected output:
Wedderga: Warning!!! 
Some of the Wedderburn components displayed are FRACTIONAL MATRIX ALGEBRAS!!!

[ [ 1, Rationals ], [ 1, Rationals ], [ 1, Rationals, 10, [ 4, 3, 5 ] ],
  [ 4, Rationals ], [ 4, Rationals ], [ 5, Rationals ], [ 5, Rationals ],
  [ 6, Rationals ], [ 1, NF(12,[ 1, 11 ]), 10, [ 4, 3, 5 ] ],
  [ 3/2, NF(8,[ 1, 7 ]), 10, [ 4, 3, 5 ] ] ]

# But found:
Error, Assertion failure
########
    5396 ms (699 ms GC) and 810MB allocated for wedderga02.tst
testing: /circa/scratch/gap-jenkins/workspace/GAP-stable-test/GAPCOPTS/64build\
/GAPTARGET/packages/label/kovacs/GAP-stable-snapshot/pkg/wedderga-4.9.4/tst/we\
dderga03.tst
    2407 ms (408 ms GC) and 383MB allocated for wedderga03.tst
testing: /circa/scratch/gap-jenkins/workspace/GAP-stable-test/GAPCOPTS/64build\
/GAPTARGET/packages/label/kovacs/GAP-stable-snapshot/pkg/wedderga-4.9.4/tst/we\
dderga04.tst
   20468 ms (3282 ms GC) and 1.10GB allocated for wedderga04.tst
testing: /circa/scratch/gap-jenkins/workspace/GAP-stable-test/GAPCOPTS/64build\
/GAPTARGET/packages/label/kovacs/GAP-stable-snapshot/pkg/wedderga-4.9.4/tst/we\
dderga05.tst
     928 ms (246 ms GC) and 137MB allocated for wedderga05.tst
testing: /circa/scratch/gap-jenkins/workspace/GAP-stable-test/GAPCOPTS/64build\
/GAPTARGET/packages/label/kovacs/GAP-stable-snapshot/pkg/wedderga-4.9.4/tst/we\
dderga06.tst
     389 ms (203 ms GC) and 34.9MB allocated for wedderga06.tst
testing: /circa/scratch/gap-jenkins/workspace/GAP-stable-test/GAPCOPTS/64build\
/GAPTARGET/packages/label/kovacs/GAP-stable-snapshot/pkg/wedderga-4.9.4/tst/we\
dderga07.tst
########> Diff in /circa/scratch/gap-jenkins/workspace/GAP-stable-test/GAPCOPT\
S/64build/GAPTARGET/packages/label/kovacs/GAP-stable-snapshot/pkg/wedderga-4.9\
.4/tst/wedderga07.tst:37
# Input is:
W:=WedderburnDecompositionInfo(R);
# Expected output:
Wedderga: Warning!!! 
Some of the Wedderburn components displayed are FRACTIONAL MATRIX ALGEBRAS!!!

[ [ 1, Rationals ], [ 1, Rationals ], [ 1, Rationals, 10, [ 4, 3, 5 ] ], 
  [ 4, Rationals ], [ 4, Rationals ], [ 5, Rationals ], [ 5, Rationals ], 
  [ 6, Rationals ], [ 1, NF(12,[ 1, 11 ]), 10, [ 4, 3, 5 ] ], 
  [ 3/2, NF(8,[ 1, 7 ]), 10, [ 4, 3, 5 ] ] ]
# But found:
Error, Assertion failure
########
########> Diff in /circa/scratch/gap-jenkins/workspace/GAP-stable-test/GAPCOPT\
S/64build/GAPTARGET/packages/label/kovacs/GAP-stable-snapshot/pkg/wedderga-4.9\
.4/tst/wedderga07.tst:45
# Input is:
CyclotomicAlgebraWithDivAlgPart(W[3]);
# Expected output:
[ 2, rec( Center := Rationals, DivAlg := true, 
      LocalIndices := [ [ 5, 2 ], [ infinity, 2 ] ], SchurIndex := 2 ) ]
# But found:
[ 1, NF(39,[ 1, 2, 4, 5, 8, 10, 11, 16, 20, 22, 25, 32 ]) ]
########
########> Diff in /circa/scratch/gap-jenkins/workspace/GAP-stable-test/GAPCOPT\
S/64build/GAPTARGET/packages/label/kovacs/GAP-stable-snapshot/pkg/wedderga-4.9\
.4/tst/wedderga07.tst:48
# Input is:
CyclotomicAlgebraWithDivAlgPart(W[9]);
# Expected output:
[ 2, rec( Center := NF(12,[ 1, 11 ]), DivAlg := true, 
      LocalIndices := [ [ infinity, 2 ] ], SchurIndex := 2 ) ]
# But found:
Error, List Element: <list>[9] must have an assigned value
########
########> Diff in /circa/scratch/gap-jenkins/workspace/GAP-stable-test/GAPCOPT\
S/64build/GAPTARGET/packages/label/kovacs/GAP-stable-snapshot/pkg/wedderga-4.9\
.4/tst/wedderga07.tst:51
# Input is:
CyclotomicAlgebraWithDivAlgPart(W[10]);
# Expected output:
[ 3, rec( Center := NF(8,[ 1, 7 ]), DivAlg := true, 
      LocalIndices := [ [ infinity, 2 ] ], SchurIndex := 2 ) ]

# But found:
Error, List Element: <list>[10] must have an assigned value
########
########> Diff in /circa/scratch/gap-jenkins/workspace/GAP-stable-test/GAPCOPT\
S/64build/GAPTARGET/packages/label/kovacs/GAP-stable-snapshot/pkg/wedderga-4.9\
.4/tst/wedderga07.tst:167
# Input is:
LocalIndicesOfCyclotomicAlgebra(W[27]);
# Expected output:
[ [ infinity, 2 ] ]

# But found:
Error, reached the pre-set memory limit
(change it with the -o command line option)
########
  117746 ms (8024 ms GC) and 4.01GB allocated for wedderga07.tst
testing: /circa/scratch/gap-jenkins/workspace/GAP-stable-test/GAPCOPTS/64build\
/GAPTARGET/packages/label/kovacs/GAP-stable-snapshot/pkg/wedderga-4.9.4/tst/we\
dderga08.tst
     460 ms (278 ms GC) and 19.0MB allocated for wedderga08.tst
-----------------------------------
total    253973 ms (25981 ms GC) and 16.4GB allocated

#I  Errors detected while testing

============================OUTPUT END================================

With default packages:

olexandr-konovalov commented 6 years ago

This is caused by some changes in the stable-4.10 branch: in the Travis tests, with the same set of packages, it failed first time at https://travis-ci.org/gap-packages/gap-docker-pkg-tests-stable-4.10-devel/jobs/434965129 but passed at https://travis-ci.org/gap-packages/gap-docker-pkg-tests-stable-4.10-devel/jobs/434578488 a day before.

The same likely happened in the master branch, but was not noticed earlier because tests there are known to massively fail due to the method reordering.

I suspect Wedderga may have wrongly assuming something, now revealed by Fix IsNilpotentGroup: infinite does not imply "not p-group". Need to investigate.

olexandr-konovalov commented 6 years ago

Ah, the same pattern in master branch of GAP: passed at https://travis-ci.org/gap-packages/gap-docker-pkg-tests-master-devel/jobs/434372146, fails now at https://travis-ci.org/gap-packages/gap-docker-pkg-tests-master-devel/jobs/435114569. So this is new test failure, it was not sitting there for a while already.

olexandr-konovalov commented 6 years ago

The first of the problems traced and reported in https://github.com/gap-system/gap/issues/2892.

The second one may not happen each run:

gap> WedderburnDecompositionInfo(GroupRing(Rationals,SmallGroup(240,89)));
Wedderga: Warning!!! 
Some of the Wedderburn components displayed are FRACTIONAL MATRIX ALGEBRAS!!!

[ [ 1, Rationals ], [ 1, Rationals ], [ 1, Rationals, 10, [ 4, 3, 5 ] ], [ 4, Rationals ], [ 4, Rationals ], [ 5, Rationals ], 
  [ 5, Rationals ], [ 6, Rationals ], [ 1, NF(12,[ 1, 11 ]), 10, [ 4, 3, 5 ] ], [ 3/2, NF(8,[ 1, 7 ]), 10, [ 4, 3, 5 ] ] ]
gap> WedderburnDecompositionInfo(GroupRing(Rationals,SmallGroup(240,89)));
Error, MaximumList: <list> must contain at least one element at /Users/alexk/GITREPS/gap-stable-4.10/lib/list.gi:2613 called from
MaximumList( arg[1] ) at /Users/alexk/GITREPS/gap-stable-4.10/lib/list.gi:2527 called from
Maximum( mpE ) at /Users/alexk/GITREPS/gap-stable-4.10/lib/stbcbckt.gi:2739 called from
oper( super, sub ) at /Users/alexk/GITREPS/gap-stable-4.10/lib/domain.gd:439 called from
Normalizer( G, r.representative ) at /Users/alexk/GITREPS/gap-stable-4.10/lib/oprtglat.gi:224 called from
SubgroupsOrbitsAndNormalizers( n, com, false ) at /Users/alexk/GITREPS/gap-stable-4.10/lib/grplatt.gi:1172 called from
...  at *stdin*:115
you can 'quit;' to quit to outer loop, or
you can 'return;' to continue
brk>

what, as a call with assertion level 1 show, may be pointing to an earlier problem:

gap> SetAssertionLevel(1);
gap> WedderburnDecompositionInfo(GroupRing(Rationals,SmallGroup(240,89)));
Error, Assertion failure in
  Assert( 1, Size( c ) = Index( a, ser[(i - 1)] ) * Size( nts[j] ) ); at /Users/alexk/GITREPS/gap-stable-4.10/lib/grplatt.gi:1163 called from 
LatticeSubgroups( G ) at /Users/alexk/GITREPS/gap-stable-4.10/lib/grplatt.gi:208 called from
ConjugacyClassesSubgroups( G ) at /Users/alexk/GITREPS/gap-stable-4.10/pkg/wedderga-4.9.4/lib/ExtremeSSPs.gi:363 called from
SSPNonESSPAndTheirIdempotents( QG ) at /Users/alexk/GITREPS/gap-stable-4.10/pkg/wedderga-4.9.4/lib/main.gi:1932 called from
StrongShodaPairs( G ) at /Users/alexk/GITREPS/gap-stable-4.10/pkg/wedderga-4.9.4/lib/main.gi:414 called from
WeddDecompData( G ) at /Users/alexk/GITREPS/gap-stable-4.10/pkg/wedderga-4.9.4/lib/main.gi:326 called from
...  at *stdin*:116
you may 'return;'

olexandr-konovalov commented 6 years ago

Tests now pass both in stable-4.10 and master.

gap-packages / wedderga

Tests fail in stable-4.10 and master GAP branches #43