[PATCH] Allow Samba/WAF to have pkg-config path overriden

[Zen]

Yes - this would be good enough. We'll just assign PKG_CONFIG to PKGCONFIG.

Do you want me to send an updated patch for just the backporting part of
that function?

On Tue, Jan 17, 2017 at 10:16 PM, Uri Simchoni <uri at samba.org> wrote:

> On 01/11/2017 01:16 PM, Uri Simchoni wrote:
> > On 01/11/2017 03:36 AM, Zen wrote:
> >> I'm cross-compiling for a configuration where the target sysroot is
> >> different from the build system.
> >>
> >> In order for several parts of configure that use pkg-config to verify
> >> dependencies to work they need to use the pkg-config from the targets
> root
> >> not the build system.
> >>
> >> I looked into it and the call to pkg-config is just in a string (which
> will
> >> resolve in the context of the build machine). I looked at the latest
> >> version of WAF [1] and it does something a little different and calls
> >> find_program(). This will also not work correctly for this configuration
> >> because it will still just look on the build systems path.
> >>
> >> I attached a patch to check PKG_CONFIG environment variable (which is
> what
> >> autoconf uses) first then fallback to using find_program() like the more
> >> recent version of WAF does.
> >>
> >> Thanks!
> >>
> >> [1] -
> >> https://github.com/waf-project/waf/blame/master/
> waflib/Tools/c_config.py#L236
> >>
> >
> > I'm OK with the idea. You can call it "honor PKG_CONFIG environment
> > variable" which is what autoconf does. However, upstream waf doesn't
> > honor PKG_CONFIG - as you show in your link. Can you send it upstream
> > first so that it doesn't become an obstacle to upgrading waf? (If
> > someone on the Team wants to overrule this judgment please do - it's a
> > balance between making Samba useful without private patches and making
> > its long-term maintenance possible).
> >
> > Beyond that, the patch itself does not apply (seems to have different
> > indentation than the source), is not in git-am format, and doesn't have
> > a signed-off-by attribution, but we can take care of that once we
> > resolve the upstream waf business.
> >
> > Thanks,
> > Uri.
> >
> Hi,
>
> I've seen your pull request to waf and Thomas' response (namely that waf
> honors PKGCONFIG but not PKG_CONFIG). I've also verified that
> back-porting the upstream behavior would make Samba honor PKGCONFIG
> variable. Would such back-porting be sufficient?
>
> Thanks,
> Uri.
>
>