hdf5: 2.2.0; shouldn't need revbumps

[eborisch]

Maintainer update. Shouldn't need rev-bumps, but if we find some are needed, let me know.

[barracuda156]

Destroot fails, glob pattern matches nothing.

[Dave-Allured]

@eborisch Sorry to disappoint, but I think full rev-bumps will be needed for all HDF5 releases, possibly excepting patch releases. HDF5 continues to use H5check_version. For 2.1.1 --> 2.2.0, this will mean fatal errors for applications that do not rebuild their HDF5 direct dependencies. Please correct me if I am mistaken.

[barracuda156]

@Dave-Allured Oh, this is painful. I will check soon, since I already update HDF5, believing the statement re no need of revbumps.

[Dave-Allured]

Furthermore, it looks like the upstream version bump for 2.1.1 --> 2.2.0 was completely unnecessary. Their compatibility reports show 100% backward compatibility for all components used by McPorts:

https://support.hdfgroup.org/releases/hdf5/2.2.0/documentation/compat_reports/index.html

[Dave-Allured]

Okay, I might be wrong about the rev-bumps. It looks like HDF5 recently adopted a new compatibility policy. According to that, the jump 2.1.1 --> 2.2.0 should NOT require any rev-bumps.

Unfortunately the documentation for H5check_version is not consistent with the new policy, and does not mention it. I can not tell from the docs or code what the actual requirements and side effects (warning messages) are right now. Some simple testing may be in order.

[eborisch]

They’ve changed the logic for when it complains to be more sane; I’ll have to dig up the code if you want specifics. But a quick check of a compiled tool worked when swapping out for this hdf5 version.

I haven’t had a chance to fix the build failures yet; it looks like they changed how the Fortran modules (which I don’t use) are placed.

[Dave-Allured]

Thank you, I don't need more specifics. I just wanted you to be aware of this possible problem. It looks like your quick check verified that rev-bumps will NOT be needed. That's good.