CI: Restore code coverage tracking for C sources

[fingolfin]

We lost code coverage tracking for the C code in the migration from Travis CI to GitHub Actions in PR #28.

I verified in the logs that all files are compiled and linked with the --coverage option. But in the logs uploaded to Codecov, the "reduced" coverage info is zero. E.g. for standalone/lib/fsawa.c we get

# path=fsawa.c.gcov.reduced
        -:    0:Source:fsawa.c
<<<<<< EOF

yet also

./standalone/lib/fsawa.c:39,45,51,54,61,62,65,74,80,81,85,89,90,97,101,107,108,113,114,119,124,132,146,150,153,159,175,183,222,224,233,236,243,245,250,251,254,259,276,280,281,290,293,297,299,303,306,312,316,321,323,325,326,329,338,344,345,349,353,354,361,365,371,372,377,378,383,388,396,410,414,417,423,439,447,486,488,497,500,507,509,514,515,518,523,540,544,545,554,557,561,563,567,570,576,580,585,587,589

In the build logs, we see:

...
==> Running gcov in . (disable via -X gcov)
./standalone/lib/nfa.gcda:cannot open data file, assuming not executed
./standalone/lib/fsageopairs.gcda:cannot open data file, assuming not executed
./standalone/lib/diffredwl.gcda:cannot open data file, assuming not executed
./standalone/lib/fsarevmid.gcda:cannot open data file, assuming not executed
./standalone/lib/fsasubmult.gcda:cannot open data file, assuming not executed
...
Cannot open source file fsa.c
Cannot open source file nfa.c
Cannot open source file fsageopairs.c
Cannot open source file diffredwl.c
Cannot open source file fsarevmid.c
...
File 'fsawa.c'
Lines executed:40.91% of 330
Branches executed:43.57% of 280
Taken at least once:31.79% of 280
Calls executed:22.89% of 83
Creating 'fsawa.c.gcov'
...
Cannot open source file fsasubmult.c
Cannot open source file kbfns.c
Cannot open source file migmdet.c
Cannot open source file fsawa.c
...

So I am guessing it is confused about how to find the source files. But I wonder why?!? Esp. since it used to work fine when we used basically the same scripts on Travis... weird.

[olexandr-konovalov]

Strange - works for https://github.com/gap-packages/FPLSA, for example, seems approximately same time ago...