Re: [lmi] Ignore "harmless" 'configure' warnings?

[Vadim Zeitlin]

On Thu, 31 Aug 2017 00:14:45 +0000 Greg Chicares <address@hidden> wrote:

GC> On 2017-08-30 22:51, Vadim Zeitlin wrote:
GC> > On Wed, 30 Aug 2017 22:28:47 +0000 Greg Chicares <address@hidden> wrote:
GC> [...]
GC> > GC> Or is it that I'm missing something important, and it's actually
GC> > GC> imperative for some unseen reason to specify $(prefix) in both
GC> > GC> PKG_CONFIG_PATH and a '--with-lzma' argument?
GC> > 
GC> >  AFAICS adding /opt/lmi/local/lib/pkgconfig to PKG_CONFIG_PATH should be
GC> > sufficient and would be preferable to explicitly specifying --with-lzma
GC> > argument, if only because it would work for all libraries (if another one
GC> > is ever needed) and not just this one, but also because using pkg-config 
is
GC> > safer as it could check for the appropriate library version, for example
GC> > (even if currently it doesn't).
GC> 
GC> That turned out not to work for me.

 I should have tested this myself before posting... It turns out there is
another bug in the same code, it doesn't use the result of pkg-config-based
detection when checking for lzma.h. Leaving aside the fact that it seems
perfectly useless to check for the header at all (it should just trust
pkg-config), this can never work because it looks for it in the standard
include paths, not adding the one detected by pkg-config.

GC> I'll just do both. We can always change it later if we don't like that.

 Yes, this doesn't really hurt, but right now setting PKG_CONFIG_PATH
simply doesn't work at all. What a mess...

VZ