BLD: fix type safety of version numbers (fix building on some systems)

[neutrinoceros]

This fixes compilation of pyerfa on my system (MacOS AMD64) as reported in liberfa/pyerfa#141
However it's not clear to me what changed recently that made this patch (apparently) necessary. It's still possible that the change was specific to my setup, but I figure this patch is probably acceptable regardless.

[mhvk]

This fails all tests, so is not the solution...

And it's weird - why is PACKAGE_VERSION_MAJOR seen as a string in the first place? Shouldn't it just be replaced by the preprocessor with a number?

We actually have two build systems, regular make and meson. Grepping, I see that in the former some care is taken to ensure there are no quotes:

erfa/m4/erfa-numver.m4

Lines 22 to 24 in df9f5c4

	AC_DEFINE_UNQUOTED(PACKAGE_VERSION_MAJOR, $MAJOR, [Define to the major version of this package.])
	AC_DEFINE_UNQUOTED(PACKAGE_VERSION_MINOR, $MINOR, [Define to the minor version of this package.])
	AC_DEFINE_UNQUOTED(PACKAGE_VERSION_MICRO, $MICRO, [Define to the micro version of this package.])

While I'm less sure about meson.built:

erfa/meson.build

Lines 22 to 29 in df9f5c4

	# API versions
	version = meson.project_version().split('.')
	confdata.set_quoted('PACKAGE_VERSION', meson.project_version())
	confdata.set('PACKAGE_VERSION_MAJOR', version[0])
	confdata.set('PACKAGE_VERSION_MINOR', version[1])
	confdata.set('PACKAGE_VERSION_MICRO', version[2])
	confdata.set_quoted('SOFA_VERSION', '20231011')

Though the fact that sofa uses set_quoted suggests to me that the others should normally be OK...

[neutrinoceros]

I see. Let me try and figure out what compiler/build system pip is even using under the hood.

[neutrinoceros]

I replaced the patch with a completely different approach. Hopefully this is the actual fix this time ("it works on my machine" ™️).

[mhvk]

It would also be good to have a separate run with a setup more like yours (we already have MacOS, though)...

[mhvk]

Sadly, the new approach doesn't work either... I think my link to meson.built may have been a red herring.

[neutrinoceros]

I guess I messed up my local testing too, this doesn't look right. Anyway, I'll get back to this later.

[neutrinoceros]

I actually can't seem to reproduce the problem today, even starting fresh with a clean copy of the repo. Pretty odd, but no point in keeping this PR alive then. Thanks for your feedback and sorry for the noise !

[mhvk]

Very weird indeed, but no worries about the noise!