These versions really confuse me. And I admit that I haven't thought about macOS.
I didn't even know that it has its own versioning rules.

My intention with libtool_soversion is that its 3 list elements shall correspond to
what libtool calls current:revison:age, so

current = libtool_soversion[0]
revision = libtool_soversion[1]
age = libtool_soversion[2]

That's not how you have interpreted it in your PR, where you set

current = libtool_soversion[0] + libtool_soversion[1]
revision = libtool_soversion[2]

It doesn't make a difference when libtool_soversion = [0, 0, 0].

You're right, I made a mistake (even though it doesn't matter in this case as you pointed out).

I will push a fix right away.

I have merged this PR and pushed a similar commit to the master branch.

Do you want a new libsigc++ release pretty soon because of this?

I suppose that darwin_versions must be added to other modules where I have added
Meson support. (cairomm, glibmm, atkmm, pangomm, gtkmm, libxml++; most of them at
https://gitlab.gnome.org/GNOME)

Thanks for merging the PR in. A new release is not urgent, as I have already added the patch to the Homebrew formula.

The other modules will indeed need similar patches.