[47969] trunk/dports/graphics/jpeg/Portfile

[Marcus Calhoun-Lopez]

Ryan Schmidt <ryandesign at ...> writes:

>
> > +
> > +    # Reorder link flags so that so that local -L options come  
> > first (especially before -L${prefix}/lib)
> > +    #     (see http://trac.macports.org/ticket/16411).
> > +    reinplace "s|\\(.*\\)\\(\$(LDFLAGS)\\)\\(.*\\)\\(\$(LDLIBS)\\)\ 
> > \(.*\\)|\\1\\4\\3\\2\\5|" ${worksrcpath}/makefile.cfg
> 
> This is hard to decipher. Could it be implemented as a patchfile  
> instead, or would that be huge?
As I recall there were four or five instances which had to be changed,
so a patchfile would not be too bad.
I was hoping the comment would make it more palatable.

> 
> > +    # As in the case of -L, CPPFLAGS come before -I. during  
> > compilation.
> > +    configure.cppflags-append -isystem${prefix}/include
> > +    configure.cppflags-delete -I${prefix}/include
> 
> What is the story behind using -isystem instead of -I? Where does  
> that come from?
> 
> Is this a change that would be useful generally? Should MacPorts base  
> do this?

I saw in the output that -I${prefix}/include was first during compilation.

For me, this has been an issue with ghostscript and qt4-mac:
http://trac.macports.org/changeset/42207/trunk/dports/print/ghostscript/Portfile
http://trac.macports.org/changeset/47699/trunk/dports/aqua/qt4-mac/Portfile

It seems to have been an issue for others.
There are a number of ports with -I${worksrcpath}.
I can not speak for all of them, but mono at least seems to have picked up the
wrong header file:
http://trac.macports.org/changeset/46637

-Marcus