Framework: Fix `make describe` for SUBPACKAGES case

[arrowd]

Summary

The main idea behind the change is to remove ${deptype}_DEPENDS_ALL variables. Instead, all deps up to BUILD ones are essentially shared, while LIB, RUN and TEST deps are directly taken from the corresponding .${sp} ones. It is up to port writer to add ${LIB_DEPENDS} to LIB_DEPENDS.${sp}.

Test Plan

poud testport on audio/alsa-plugins (a port with subpackages) and sysutils/44bsd-more (a port without subpackages). The make describe output for the latter did not change.

[pizzamig]

In general, it should works, but the *DEPENDS_ALL needs to stay (even if not used by the framework) as they are used externally (by poudriere).

Mk/bsd.port.mk

2668

the *_DEPENDS_ALL you are removing are the one used by poudriere to fetch dependency of the port before building it.
For this reason, they cannot be removed until poudriere is not patched to use something else

4235

I would add, as a comment, that _DEPENDS_ALL_SUBPACKAGES is only defined for RUN.
Maybe it makes sense to check for RUN_DEPENDS_ALL_SUBPACKAGES separately

4479

While equivalent, I would keep the for loop in the else branch of empty(FLAVORS)

[arrowd]

Mk/bsd.port.mk

2668

The ALL variables were introduced just recently with your https://reviews.freebsd.org/D40549
You mean Poudriere picked these up already?

4235

_DEPENDS_ALL_SUBPACKAGES are defined for RUN, TEST and LIB, but only RUN is used here.

Maybe it makes sense to check for RUN_DEPENDS_ALL_SUBPACKAGES separately

That'd mean more lines of code. But I will do that if you insist.

[pizzamig]

Mk/bsd.port.mk

2668

The ALL variables are used by poudriere to support subpackages.

4235

No strong opinion, but it seems uneven to me.
In any case, I won't block the merge because of this, tho.

[arrowd]

The follow-up commit for Poudriere, if I get it right: https://github.com/freebsd/poudriere/commit/ee34405ae8e7d1763f2cc70e1f179b8c643de1e9

Didn't test it runtime, though.

[arrowd]

• Address some comments

[arrowd]

Mk/bsd.port.mk

4235

Maybe it makes sense to check for RUN_DEPENDS_ALL_SUBPACKAGES separately

I took another look at that and don't quite understand what do you mean by checking separately? You mean plugging in another .if defined(${deptype}_DEPENDS_ALL_SUBPACKAGES) check?

[arrowd]

• Rebase
• Add compatibility shim for Poudriere

To test this one can run git revert 7dd904d9178aef2ac7369c77551e6c7774997564
to bring audio/alsa-plugins into the subpackagized state.

Running make -V {BUILD,LIB,RUN}_DEPENDS_ALL now produce the same list as before
modulo some extra duplicated entries, which were present in the previous version anyways.