fpm-deployment: update VERSION field in fpm.toml with actual version

[perazz]

See #701, this update in the fpm deployment script updates the VERSION placeholder in fpm.toml so an actual version is printed.

I don't know how to test the deployment on my fork.

[jvdp1]

LGTM. Thanks @perazz for this PR.

[arteevraina]

I don't know how to test the deployment on my fork.

You can try running this script locally and try to generate the fpm compatible stdlib ? and look if the changes are reflected in fpm.toml.

[awvwgk]

Why is this change necessary? We already deploy the VERSION file correctly for the stdlib-fpm branch.

[perazz]

Currently, fpm.toml in the deployed branch reports version = "VERSION"; this proposed change updates that field with the actual version, i.e., version = "0.2.1"

[awvwgk]

Currently, fpm.toml in the deployed branch reports version = "VERSION"; this proposed change updates that field with the actual version, i.e., version = "0.2.1"

Maybe, I misunderstand something here, but this is exactly what is expected. The version is already deployed correctly in VERSION on the stdlib-fpm branch. It will be read dynamically by the fpm binary. So why do we need to change the manifest at all?

If I recall correctly, I actually implemented this feature of reading the version from a file in fpm for stdlib to have a single place to specify the version for both build systems (fpm and CMake).

[perazz]

Oh I see, version can be an external file. That .toml can only be read proprely by fpm then. If we were going to support stdlib as a metapackage, I thought it was intuitive for the fpm-specific deployment to have a manifest file that contains all relevant pieces of information.

[awvwgk]

That was unfortunately never true, especially due to the default names for the source directories as well as the automatic discovery of executables, the TOML manifest alone is usually incomplete. Running fpm will generate an internal representation which is complete, the two options are to have fpm dump the internal model or reimplements the whole logic supported by fpm.

[jvdp1]

@perazz @awvwgk Is this PR still needed? Or could it be closed without merging it?

[perazz]

Please close without merging @jvdp1, thank you.

[jvdp1]

Ok. Thank you.