Mac OS X and pkg-config

[David Tardon]

Hi,

On Wed, Jun 10, 2015 at 01:17:07PM +0200, David Tardon wrote:
> Hi,
> 
> On Tue, Jun 09, 2015 at 04:30:45PM +0200, Michael Stahl wrote:
> > On 09.06.2015 16:20, Thorsten Behrens wrote:
> > > Tor Lillqvist wrote:
> > >> But neither is it very useful to have the already small set of OS X -based
> > >> developers split into those who use a pure upstream way to build, those who
> > >> use HomeBrew libjpeg but bundled Python, those who use MacPorts Python and
> > >> libjpeg, etc.
> > >>
> > > Maybe. But is that really so different on Linux (where we seem to
> > > cope)? And I guess Khaled's intention is to rather grow the pool of
> > > Mac hackers, by removing one very early point of frustration ...
> > 
> > we cope with this on Linux mainly because we have distribution package
> > maintainers who actually work on having our build system pick up
> > $random_distro_of_the_week's inconsistently packaged system libraries
> > properly.
> 
> ... and even so we fail to ensure that system headers/libs are not used
> if one configures without them. E.g., as soon as one installs system
> boost (or mdds, or glm, or...), it will be used no matter what value
> SYSTEM_BOOST contains. Simply because -I/usr/include is in include path
> before -I$W/UnpackedTarball/boost. But we do not seem to care overly
> about this...

Actually, I just found out that this can be a problem on OS X as well. A
recent build of master fails in liborcus if there is boost installed
from Homebrew. Because, unfortunately, Homebrew installs stuff into
/usr/local and the boost.m4 script used by liborcus looks there... (This
would not cause build breakage by itself. But the check does not add
-L/usr/local/lib into LDFLAGS, so the detected boost libraries are not
found by linker.)

D.