.github/workflows/ci.yml: run 'cargo valgrind test --lib'

[rouault]

• I agree to follow the project's code of conduct.
• I added an entry to CHANGES.md if knowledge of this change could be valuable to users.

---

[lnicola]

I expected this to be slower, but 30 seconds isn't bad at all.

[lnicola]

I'm thinking it might be a good idea to run this in the 3.8 container, since LTS might not have some of the APIs (like MD) we've implemented.

[rouault]

I'm thinking it might be a good idea to run this in the 3.8 container, since LTS might not have some of the APIs (like MD) we've implemented.

I've added a commit for that. Ok, it detects a leak in gdal::raster::mdarray::MDArray::read_into_slice (). I need to figure out how to build georust/gdal against my custom GDAL build rather than the 3.0.4 that comes with my Ubuntu version

[lnicola]

Great, we can disable it for now.

https://github.com/georust/gdal/tree/master/gdal-sys#build might help with linking.

[lnicola]

Not sure if that, but we might be leaking the data type in the error case at https://github.com/georust/gdal/blob/master/src/raster/mdarray.rs#L166.

[rouault]

ok, the leak in read_into_slice() is fixed now, but there's a "system leak" related to threads & TLS, that is IMHO a false positive. But I can't find an option to specify a Valgrind suppression file to cargo valgrind. I guess we need to skip 3.8.1 testing for now

[lnicola]

Yeah, let's disable it. I'll merge this in the morning.

[rouault]

Yeah, let's disable it. I'll merge this in the morning.

I've dropped the 3.8.1 related commit

[lnicola]

Funny, I don't get the TLS leak on my system (Gedora 39, GDAL 3.7.3).

I think we can pass in a suppressions file with VALGRINDFLAGS="--suppressions=gdal.supp" cargo valgrind test.