There's a couple of tests under the check folder that fail when
--enable-indirect-deps are passed to the configure script. Namely, check-libs
I'll attach a short patch I created against CVS HEAD to correct this, in case
you want to review it.
Created attachment 3104 [details]
Xorg.0.log with Option "MonitorLayout" "CRT,NONE"
Present on OS X also (pkg-config 0.18.1)
(glad I finally found this bugzilla!)
I think it would be better to adjust the tests to what *should* happen instead
of just skipping those tests. Otherwise we'll never know if what those tests are
supposed to test is working. Perhaps have alternate check-libs and
check-requires-private test cases.
I don't think I created this bugzilla item...why isn't the original comment (and
the patch I was discussing) here any more? Why is the "Opened by" name missing
in the 0th comment? Am I going crazy here?
Still broken in 0.20. Could we get some feedback here about how to proceed?
Bugzilla on fd.o has a small breakdown some time ago.
I haven't gotten around to investigating it, if you provide a patch, I can apply and test it, else I'll
have to develop one myself, which takes more time.
The issue with FAIL:check-libs is what *should* happen with *.private fields
when --enable-indirect-deps is used. That is, for simple.pc:
should --libs be seeing -lm or not? I think not, since the whole idea of
.private (as I understand it) is that it is for things used when libsimple was
built but that are not exposed to things built against libsimple. It's something
even less than an indirect dependency, which libsimple exposes to callers and so
callers might (depending on linker features) need to link themselves as well.
I see that Cflags/Cflags.private is implemented the other way:
pkg-config --cflags gives -Dpublic only (there does not appear to be a test of
*duh*, there's no such thing as Cflags.private. Please disregard my
Okay, looking at main.c, --enable-indirect-deps from ./configure merely sets
pkg-config to assume --static at runtime. So the test says that this is behaving
"correctly as implemented". However, this is not what the test expects, which
appears to be that indirect means "recurse down the Requires: web only", not
"and also merge every foo.private: into its associated foo:".
I believe this bug is fixed in 0.25 at least, so closing this. Please reopen if you still see it.