FactorCosetAction fails for non-proper subgroups

[lgoettgens]

On GAP 4.14.0 my minimal example fails as

gap> G:=SymmetricGroup(4);
Sym( [ 1 .. 4 ] )
gap> FactorCosetAction(G,G);
Error, <map> must be a general mapping at /usr/share/gap/lib/mapping.gi:100 called from
Image( hom, G ) at /usr/share/gap/lib/csetperm.gi:931 called from
DoFactorCosetActionPerm( G, U ) at /usr/share/gap/lib/csetperm.gi:943 called from
<function "FactorCosetAction by right transversal operation">( <arguments> )
 called from read-eval loop at *stdin*:2
type 'quit;' to quit to outer loop

while on 4.13.1 it returns an action epimorphism, as expected.

This is a minimal example of the error mentioned in oscar-system/GAP.jl#1114 (comment).
cc @fingolfin @ThomasBreuer

[lgoettgens]

Looking through the changes made to relevant files, I suspect this to be due to #5698 by @hulpke

[ThomasBreuer]

@lgoettgens Thanks for inspecting this problem.

Yes, #5698 introduces a new method for FactorCosetAction for perm. groups. (The generic method for groups would work in the given example.)
This new code seems to assume that ActionRefinedSeries returns a series of length at least 2,
which cannot happen if the subgroup is not proper.
A brute force fix would be to catch this situation, and to call TryNextMethod() in order to delegate to the generic method.

[lgoettgens]

Yes, #5698 introduces a new method for FactorCosetAction for perm. groups. (The generic method for groups would work in the given example.) This new code seems to assume that ActionRefinedSeries returns a series of length at least 2, which cannot happen if the subgroup is not proper. A brute force fix would be to catch this situation, and to call TryNextMethod() in order to delegate to the generic method.

Could you prepare such a PR? Thanks!

[ThomasBreuer]

@lgoettgens I make a few more experiments, and then I will create a pull request that fixes this problem.