Use robust rank for generic SemidirectProduct method; function fields are fields; tweak a LocationFunc test

[fingolfin]

This fixes an infinite recursion if the rank of IsGroup and IsFinite
becomes very large (e.g. due to many packages are loaded which add
implications from that).

Resolves part of #2818

[codecov]

Codecov Report

Merging #2977 into master will decrease coverage by 0.97%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master    #2977      +/-   ##
==========================================
- Coverage   83.81%   82.83%   -0.98%     
==========================================
  Files         682      679       -3     
  Lines      345976   341571    -4405     
==========================================
- Hits       289970   282932    -7038     
- Misses      56006    58639    +2633

Impacted Files	Coverage Δ
lib/gprd.gi	56.29% <100%> (-15.61%)	⬇️
src/modules.h	33.33% <0%> (-66.67%)	⬇️
src/libgap-api.c	0% <0%> (-66.67%)	⬇️
src/gvars.h	66.66% <0%> (-33.34%)	⬇️
src/fibhash.h	66.66% <0%> (-33.34%)	⬇️
lib/grpreps.gi	68.62% <0%> (-29.42%)	⬇️
src/gapstate.h	42.85% <0%> (-28.58%)	⬇️
src/gasman.h	69.23% <0%> (-25.9%)	⬇️
src/compiler.c	67.54% <0%> (-21.01%)	⬇️
src/permutat.h	48.27% <0%> (-20.69%)	⬇️
... and 196 more

[olexandr-konovalov]

Excellent - this seem to fix it! I am running a test of this PR on Jenkins overnight, hope we can merge this tomorrow!

[olexandr-konovalov]

The tweak of the LocationFunc test will certainly help, but after #2898 when I run it on Jenkins, I observe this:

########> Diff in /circa/scratch/gap-jenkins/workspace/GAP-pull-request-quickt\
est/GAPCOPTS/64build/GAPGMP/gmp/GAPTARGET/install/label/kovacs/GAP-pull-reques\
t-snapshot/tst/testinstall/opers/LocationFunc.tst:9
# Input is:
LocationFunc(Where);
# Expected output:
"GAPROOT/lib/error.g:79"
# But found:
"/circa/scratch/gap-jenkins/workspace/GAP-pull-request-quicktest/GAPCOPTS/64bu\
\
ild/GAPGMP/gmp/GAPTARGET/install/label/kovacs/GAP-pull-request-snapshot/lib/er\
\
ror.g:79"
######## 

and I haven't yet had a look why. Anyhow, that's not related to method reordering.

[ChrisJefferson]

I've tried reproducing this GAPROOT problem and can't. If you can find a machine you can reproduction it on, can you print out the gaproots?

[fingolfin]

Since the tests pass here, and it resolves several (hopefully: all except for the LocationFunc one) of the issues @alex-konovalov reported on #2818, can we just merge this, and debug the LocationFunc issue somewhere else?

[olexandr-konovalov]

I've checked those nightly Jenkins builds, and this PR heals all problems arising in the tests that load all packages. That's exactly what I was going to suggest, thanks.

[olexandr-konovalov]

See #2981 for LocationFunc.