meson: add patch to produce properly versioned dylibs and gobject-introspection data

[tschoonj]

Long story as to why can be read at: mesonbuild/meson#2577
Note: this patch will not be merged in anytime soon as it breaks meson's testsuite, and I am waiting for help to fix this...

When this one is merged in, I will start updating formulas that switched to meson, as well as add some new ones...

[tschoonj]

@ilovezfs Assuming you don't mind the libepoxy warning, this is ready to merge.

[tschoonj]

I am thinking that perhaps meson should become keg-only now, as it becomes hard to use it with this patch for development purposes. Perhaps adding a caveat, asking people to install it with pip3, if necessary? Something like is done in the Cython formula.

[ilovezfs]

as it becomes hard to use it with this patch for development purposes.

can you elaborate?

Josephs-MacBook-Pro:~ joe$ brew formula-analytics meson
install events in the last 30 days for meson
====================================================================
1 | meson                                          | 3,162 |  98.87%
2 | meson --HEAD                                   |    36 |   1.13%
====================================================================
Total                                              | 3,198 |    100%
====================================================================

install_on_request events
====================================================================
1 | meson                                          | 2,185 |  98.42%
2 | meson --HEAD                                   |    35 |   1.58%
====================================================================
Total                                              | 2,220 |    100%
====================================================================

BuildError events
====================================================================
1 | meson --HEAD                                      | 21 |  87.50%
2 | meson                                             |  3 |  12.50%
====================================================================
Total                                                 | 24 |    100%
====================================================================
Josephs-MacBook-Pro:~ joe$ 

The analytics suggest we really can't do that given the number of install-on-request events.

[tschoonj]

Assuming you are developing some software project with meson, you wouldn't be able to run the compiled executables anymore before installing them (actually the shared libraries in the project must be installed). Also, ninja test will fail for the same reason.

[ilovezfs]

And what happens if we don't patch meson?

[tschoonj]

Then we cannot build any of the GNOME projects containing shared libraries and GIR bindings that use meson 😄

[ilovezfs]

Does it not work to do two stage installs, first building without the bindings, installing, and then building with the bindings, and installing?

[tschoonj]

You would still need the patch. As I explain in mesonbuild/meson#2577, meson is putting the wrong path to the dylib into the typelib files.

Also the patch adds versioning information to the dylibs, which is currently not done...

[ilovezfs]

How do you feel about using a separate meson-internal keg-only formula?

[tschoonj]

sure, I could live with that. Is there a precedent for that?

[ilovezfs]

Not that I know of, though the ghc @ formulae that are "for our own purposes" aren't conceptually much different from that.

[tschoonj]

Ok, will look into that. Thanks for the suggestion

[tschoonj]

@ilovezfs If you can live with the audit warnings, than this one is ready to merge...

[ilovezfs]

What's the status of the gom patch?

[tschoonj]

I removed all bits that assumed pygobject3 is present.

They were redundant anyway as we reinplace the install location of the override files.

[ilovezfs]

🍪

[tschoonj]

Obliged! Feel free to reopen #25428. There shouldn't be any rev bumps necessary now.