Fix: generalize hard-coded Cint to MPI_Comm in t8_forest_get_mpicomm.

[jmark]

This patch generalizes the return value of t8_forest_get_mpicomm. This is crucial when using OpenMPI as system library. The MPI Communicator handle is not an integer but a pointer to a struct.

Along the way, I also added other minor improvements.

[JoshuaLampert]

Looks good. Does this mean that executing julia --project generator.jl && ./fixes.sh in dev/ produces the Libt8.jl file?

[JoshuaLampert]

This reminds me of the following error: https://github.com/trixi-framework/Trixi.jl/actions/runs/12182527496/job/33981760954?pr=1844#step:7:11266. Could this be similar? It fails here:

T8code.jl/src/Libt8.jl

Lines 17842 to 17848 in 73877de

	```c
	void t8_geom_load_tree_data_vertices (t8_cmesh_t cmesh, t8_gloidx_t gtreeid, const void **user_data);
	```
	"""
	function t8_geom_load_tree_data_vertices(cmesh, gtreeid, user_data)
	@ccall libt8.t8_geom_load_tree_data_vertices(cmesh::Cint, gtreeid::Cint, user_data::Ptr{Ptr{Cvoid}})::Cvoid
	end

To me the docstring looks like cmesh should be of type t8_cmesh_t and not of type CInt = UInt32 as indicated in the function call. Could this lead to the InexactError?

[jmark]

Looks good. Does this mean that executing julia --project generator.jl && ./fixes.sh in dev/ produces the Libt8.jl file?

Yes, indeed!

[jmark]

This reminds me of the following error: https://github.com/trixi-framework/Trixi.jl/actions/runs/12182527496/job/33981760954?pr=1844#step:7:11266. Could this be similar? It fails here:

T8code.jl/src/Libt8.jl

Lines 17842 to 17848 in 73877de

	```c
	void t8_geom_load_tree_data_vertices (t8_cmesh_t cmesh, t8_gloidx_t gtreeid, const void **user_data);
	```
	"""
	function t8_geom_load_tree_data_vertices(cmesh, gtreeid, user_data)
	@ccall libt8.t8_geom_load_tree_data_vertices(cmesh::Cint, gtreeid::Cint, user_data::Ptr{Ptr{Cvoid}})::Cvoid
	end

To me the docstring looks like cmesh should be of type t8_cmesh_t and not of type CInt = UInt32 as indicated in the function call. Could this lead to the InexactError?

I actually fixed this issue a while ago on t8code side. For whatever reason this fix gone. I adapted the fixes.sh accordingly.

Thanks for pointing that out!

[JoshuaLampert]

Ok, thanks! Is this fix actually gone in t8code or only in T8code.jl? If the latter (and your fix in t8code is included in the t8code version we currently wrap), we could (should?) investigate why it's not in T8code.jl. (If the former, you could maybe investigate why it's gone in t8code as it would be nice to not need a manual fix on the T8code.jl end.)

[jmark]

Ok, thanks! Is this fix actually gone in t8code or only in T8code.jl? If the latter (and your fix in t8code is included in the t8code version we currently wrap), we could (should?) investigate why it's not in T8code.jl. (If the former, you could maybe investigate why it's gone in t8code as it would be nice to not need a manual fix on the T8code.jl end.)

Actually, my old fix was added to another file in t8code. The issue here was introduced in a new file. I pushed a fix upstream:
DLR-AMR/t8code#1319

But I recommend to merge this PR since the upstream fix won't be available soon and the fix here is idempotent to the fix upstream.

[JoshuaLampert]

Thanks!