Rare no-method-found error in sanity.g in stable-4.10 branch

[olexandr-konovalov]

Happened yesterday in https://gap-ci.cs.st-andrews.ac.uk/view/GAP-weekly/job/GAP-stable-sanity/GAPCOPTS=32build,label=kovacs/11/console (link only accessible from St Andrews):

testing group t20n971
conjugacy classes
rational classes
automorphism group
testing group Group( [ f1, f2, f3, f4, f5, f6, f7 ] )
subgroups
normal subgroups
conjugacy classes
rational classes
automorphism group
20,971,192,136
testing group t20n972
conjugacy classes
Error, no method found! For debugging hints type ?Recovery from NoMethodFound
Error, no 1st choice method found for `+' on 2 arguments
The 1st argument is 'fail' which might point to an earlier problem
 at /circa/scratch/gap-jenkins/workspace/GAP-stable-sanity/GAPCOPTS/32build/la\
bel/kovacs/GAP-stable-snapshot/lib/methsel2.g:250 called from
Build step 'Execute shell' marked build as failure
Sending e-mails to: alexander.konovalov@st-andrews.ac.uk
Finished: FAILURE

Was able to reproduce this locally in stable-4.10 branch, starting GAP with -r -A and calling

Read("tst/sanity.g");
for i in [1..100] do Print(i,"\n");TestsForGroup(TransitiveGroup(20,972));od;

but it took several minutes and 59 iterations:

59
testing group t20n972
conjugacy classes
Error, no method found! For debugging hints type ?Recovery from NoMethodFound
Error, no 1st choice method found for `+' on 2 arguments
The 1st argument is 'fail' which might point to an earlier problem
 at /Users/alexk/GITREPS/gap-stable-4.10/lib/methsel2.g:250 called from
First( dsz, function ( i )
        return (i >= minlen);
    end ) - minlen at /Users/alexk/GITREPS/gap-stable-4.10/lib/clashom.gi:742 called from
ConjugacyClassesSubwreath( F, FM, n, S1, Action( FM, components[1] ), T1, 
 components, emb, proj 
 ) at /Users/alexk/GITREPS/gap-stable-4.10/lib/clashom.gi:1339 called from
ConjugacyClassesFittingFreeGroup( G 
 ) at /Users/alexk/GITREPS/gap-stable-4.10/lib/clashom.gi:2363 called from
ConjugacyClassesViaRadical( G 
 ) at /Users/alexk/GITREPS/gap-stable-4.10/lib/clashom.gi:3071 called from
ConjugacyClasses( G ) at tst/sanity.g:61 called from
...  at *stdin*:2
type 'quit;' to quit to outer loop
brk> ShowArguments();
[ fail, -722 ]
brk> dsz;
[ 1, 2, 3, 4, 5, 6, 8, 9, 10, 12, 15, 16, 18, 20, 24, 30, 36, 40, 45, 48, 60, 72, 80, 90, 120, 144, 
  180, 240, 360, 720 ]
brk> minlen;
722

Indeed, all elements of dsz are smaller than 722, so First returns fail.

[ChrisJefferson]

I tried running various memory checkers on this, and they didn't fail in the first past (they might fail on iteration 59 of course...)

[fingolfin]

@alex-konovalov I assume this was run with ReproducibleBehaviour set to false (i.e., the default), right?

@hulpke does this error ring any bells, by chance?

[olexandr-konovalov]

@fingolfin yes, GAP started as above, so no modifications of ReproducibleBehaviour. Only SetAssertionLevel(2); because this is a line in tst/sanity.g.

[ChrisJefferson]

I might not have set the assertion level. I will check that in the memory checker. As I say, I only did one loop, hoping the bug would turn up. If noone has tracked anything down in a couple of days, I'll try running the full 59 iterations over the weekend and see if anything crops up.

[olexandr-konovalov]

@ChrisJefferson if you read tst/sanity.g, then the assertion level would be 2.

[olexandr-konovalov]

Closing this as #3110 claims that it fixes this.