Use make as a build system for opam files.

[jhjourdan]

There are now two opam files:

• unison.opam is the main opam file. It has an optional dependency to lablgtk3, so that the GUI is built only if the lablgtk3 package is installed.
• unison-gui.opam is a pseudo-package, with both unison and lablgtk3 as dependencies. It can be used to force the GUI to be installed.

As a side effect, remove the package information from the dune-project files, to avoid redundancy.

Fixes #842, #827

[gdt]

Thanks for submitting, I will not be able to dig into this until January sometime.

[jhjourdan]

Sure. Sorry for the delay!

[jhjourdan]

One potential subject of discussion is the status of unison-fsmonitor. Up to now, this executable is not installed by make install, but seems to be needed by some features of unison. So I considered that a bug of the Makefile, and decided to install it as part of the install Makefile target.

Is this the right thing to do?

[tleedjarv]

One potential subject of discussion is the status of unison-fsmonitor. Up to now, this executable is not installed by make install, but seems to be needed by some features of unison. So I considered that a bug of the Makefile, and decided to install it as part of the install Makefile target.

Is this the right thing to do?

Yes.

The current install target needs complete revamping anyway but this is a task for some other PR.

Also note that unison-fsmonitor is not built on macOS. There is a fsmonitor.py fallback but I think many (most?) people on macOS use a third party fsmonitor implementation instead. I don't think you have to worry about this in context of this PR.

[gdt]

unison-fsmonitor is built on some places and not others, not just !mac. Also not on netbsd.

I am not sure we want to go down this path. I do not have time to pay attention until January.

[tleedjarv]

unison-fsmonitor is built on some places and not others, not just !mac. Also not on netbsd.

True, and then there's nothing to install. For mac there is a fallback, so there is something to install.

I am not sure we want to go down this path.

Yes, probably not in this PR.

[jhjourdan]

For the record, I rebased this branch, removing my change about the installation of fsmonitor.

And then I noticed that the install target has been removed, making this change more subtle because then the opam file needs to contain the installation script.

I'm not going to continue until anyone tells he is interested.

[tleedjarv]

The install taget is in #939. Your testing and/or review would be very helpful.

[jhjourdan]

Thanks!

I rebased this PR on top of #939, and this seem to work well.

[gdt]

Can you rebase to master? I came to this intending to read what I expected to be a small number of commits. I'm guessing a rebase is trivial and leads to that. I did kick off the workflow.

[jhjourdan]

Sure, done!

[gdt]

Thanks. Now I can be 100% sure easily this only touches opam. I have kicked off a workflow run and will hit merge if it finishes green.

[jhjourdan]

Perfect, thanks!

I'll do my best to publish releases on OPAM as soon as they appear here.