gh-128354: Use LIBS consistently over LDFLAGS in library build checks

[zanieb]

As noted in #128354, I audited all uses of LIBS and LDFLAGS and adjusted checks using $<LIB>_LIBS to set LIBS instead of LDFLAGS and ensured we consistently ordered $LIBS before $<LIB>_LIBS. There are some other cases where a constant is added to LIBS that I did not change here since it's a different pattern — we can consider those separately.

I don't believe this needs a news entry, but defer to whatever the reviewer prefers.

I tested this locally on macOS and in a Linux container. It seems nice to get more test coverage too, perhaps via the build-bots?

In #95288, the ordering of CFLAGS and CPPFLAGS was fixed in a similar manner. I also noticed that some of these uses were introduced in #94802.

[erlend-aasland]

Great, thanks!

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @erlend-aasland for commit a07b043 🤖

If you want to schedule another build, you need to add the 🔨 test-with-buildbots label again.

[erlend-aasland]

In #95288, the ordering of CFLAGS and CPPFLAGS was fixed in a similar manner. I also noticed that some of these uses were introduced in #94802.

IMO, the safest thing to do is to consistently use CPPFLAGS iso. CFLAGS; CPPFLAGS will work with both pre-processor and compiler checks; CFLAGS may not work with pre-processor checks.

[erlend-aasland]

IMO, the safest thing to do is to consistently use CPPFLAGS iso. CFLAGS; CPPFLAGS will work with both pre-processor and compiler checks; CFLAGS may not work with pre-processor checks.

I remember having issues with sqlite3 dependency detection on a FreeBSD buildbot, because we initially used CFLAGS instead of CPPFLAGS.

[zanieb]

In #95288, the ordering of CFLAGS and CPPFLAGS was fixed in a similar manner. I also noticed that some of these uses were introduced in #94802.

IMO, the safest thing to do is to consistently use CPPFLAGS iso. CFLAGS; CPPFLAGS will work with both pre-processor and compiler checks; CFLAGS may not work with pre-processor checks.

That sounds good to me, though I think it's an independent change from this one. I'm just referring to the ordering when extending flags, e.g., CPPFLAGS = $CPPFLAGS $FOO vs CPPFLAGS = $FOO $CPPFLAGS. I can look into that too though.

[erlend-aasland]

That sounds good to me, though I think it's an independent change from this one.

Yes, I agree.

[erlend-aasland]

I think it would be safe to backport this, at least to 3.13, and possibly also 3.12. WDYT, @corona10 & @zanieb?

[corona10]

Yeah I think it is fine to backport the patch!!

[erlend-aasland]

Well, it does not apply cleanly; I don't think it is worth the effort to manually backport this, but if someone wants to take it on, feel free to do so :)

[bedevere-app]

GH-128465 is a backport of this pull request to the 3.13 branch.

[bedevere-app]

GH-128466 is a backport of this pull request to the 3.12 branch.