Codecov Report

Attention: Patch coverage is 0% with 32 lines in your changes missing coverage. Please review.

Project coverage is 74.16%. Comparing base (fda88b3) to head (e3e5907).
Report is 1 commits behind head on master.

@@            Coverage Diff             @@
##           master    #1067      +/-   ##
==========================================
- Coverage   74.69%   74.16%   -0.54%     
==========================================
  Files          55       55              
  Lines        4636     4668      +32     
==========================================
- Hits         3463     3462       -1     
- Misses       1173     1206      +33     

Thank you for this @lgoettgens. Could you please rebase it? I'll try to have a closer in the coming week(s)

Good: gapdoc and wpe pass now.

Bad: most other packages now fail despite reporting success:

#I  No errors detected while testing

Test Failed at none:1
  Expression: GAP.Packages.test("automata")

ERROR: There was an error during testing

This is probably due to the return value handling. I'll look into it tomorrow

Ok, so the problem is how gap packages try to exit after running their tests.

Most packages call TestDirectory with exitGAP:=true (e.g. https://github.com/gap-packages/datastructures/blob/24f8ea8d4265ffb64a6216a57fabf3aee739cc15/tst/testall.g#L4) (which internally calls QuitGap with an exit code. Up to 5587398 (#1067) we used this exit code to determine if the tests succeeded. Afterwards, these packages call ForceQuitGap(1) as they think this is an unreachable statement and thus an error if reached (cf https://github.com/gap-packages/datastructures/blob/24f8ea8d4265ffb64a6216a57fabf3aee739cc15/tst/testall.g#L7). Due to 5587398 (#1067), we now fail due to this exit code of 1.

wpe does not use TestDirectory, but instead calls ForceQuitGap with an exit code indicating if the tests succeeded (cf. https://github.com/gap-packages/WPE/blob/adc54d808f8b1634c227719f2b41a6fe81373c41/tst/testall.g#L27-L31).

Ways to solve this:

1. Patch wpe to use QuitGap instead of ForceQuitGap.
2. Only consider the very first call to any of QuitGap and ForceQuitGap as the try to convey an exit code and ignore all further calls (as they are unreachable in usual GAP execution anyway).

I will try to implement 2.

Option 2 to me sounds like the way to go as it models more accurately what GAP does.

The help package still has some failures. Could you have a look at it?