VS and Xcode backends ignore link_depends

[jon-turney]

VS backend doesn't generate a dependency on the targets in link_depends.

See #1798 for a test

(This problem is not detected by the tests added in #1172 as they are only run on linuxlike platforms)

[nirbheek]

This is because link_depends: has always been used for static files. You must change this in order to implement #1792.

[jon-turney]

I should have referenced the PR #1651 rather than bug #1172.

This is because link_depends: has always been used for static files. You must change this in order to implement #1792.

So no, currently meson claims that link_depends can take a custom_target, and that doesn't work correctly with a VS backend.

Sorry, but I will not be contributing fixes to the VS backend.

[nirbheek]

Ordinarily, we require that changes must work across both the Ninja and VS backends, but in this case it's a preexisting bug, and it's better to fix it on just the Ninja backend than not at all.

[jon-turney]

I don't actually see any sign that link_depends is used inthe xcode backend, either. I'm not sure if that's tested in CI, currently.

[drsog]

LabelAs: macos

[TheQwertiest]

IMO, the meson reference documentation should be amended to indicate that link_depends currently works only with ninja back-end, since this bug is really old and there is no indication that it will be fixed any time soon.

[TheQwertiest]

PS: I don't know how and why my comment unassigned nioncode :\

[nirbheek]

IMO, the meson reference documentation should be amended to indicate that link_depends currently works only with ninja back-end,

This is a terrible idea. Then no one will come around to fix it. Documenting a bug as expected behaviour is the easiest way to make sure that it will never get fixed.

[TheQwertiest]

This is a terrible idea.

Hm... Then what about adding smth like the implementation is currently buggy on VS and XCode backends (link to issue #1799)? This will allow meson users to at least see the issue and to stop wasting time trying to understand why it doesn't work (and then maybe even contribute to fixing it?)

[nirbheek]

How about we just fix it? ;)

[TheQwertiest]

That would be the ideal solution, but this is a two year old bug, so forgive me for being doubtful... :\ (not that I'm blaming anybody, there are priorities and all that)

[jon-turney]

Perhaps we should just document the reality that these backends are unmaintained and liable to be removed if someone doesn't step up...

[nioncode]

I don't know about the XCode backend, but I implemented many things for the VS backends just a few weeks ago (and did some major improvements months-years ago where it was really not in a good shape).

All test cases pass for the VS backend, so please create a PR with a failing test case and ping me there and I will try to fix it soonish. (Please be aware that I can't put 20 hours a week into meson)

[jon-turney]

All test cases pass for the VS backend, so please create a PR with a failing test case and ping me there and I will try to fix it soonish

The relevant test case is common/149 link depends custom target, which is disabled for the VS backend due to this issue. See #1798

[jon-turney]

Thanks.

This may still be an issue with the xcode backend?

[eli-schwartz]

Partially-Fixes: #1799

GitHub saw that as "fixes #1799" and closed the ticket. Nice.

[eli-schwartz]

Is there a CI test currently raising MESON_SKIP_TEST on xcode for this? That might be the best way to check the applicability...

[dcbaker]

Unfortunately we still don't have xcode in CI i think

[jpakkane]

That's mostly because whenever I work on Xcode I get it like 90% passing and then get a brain aneurysm from the sheer terribleness of it all.