svn commit: r381968 - head/lang/gcc5

[Sunpoet Po-Chuan Hsieh]

On Tue, Mar 24, 2015 at 3:52 AM, Bryan Drewery <bdrewery at freebsd.org> wrote:

> On 3/23/2015 2:44 PM, Gerald Pfeifer wrote:
> > On Mon, 23 Mar 2015, Bryan Drewery wrote:
> >>> USE_LDCONFIG stores the directory name to
> >>> ${LOCALBASE}/libdata/ldconfig/${UNIQUENAME}.
> >>> Since lang/gcc* now share same PORTNAME (gcc), you need to
> >>> set UNIQUENAME manually.
> >> I just hit this. I've fixed it in r382022.
> >
> > Thanks for fixing this, Bryan, but...
> >
> > ...doesn't this point to a design problem with the infrastructure?
>
> Definitely. I think UNIQUENAME is now really only used for USE_LDCONFIG.
> It may make sense to default it to the values I put into the gcc ports.
>
>
UNIQUENAME is also used by bsd.options.mk for UNIQUENAME_{SET,UNSET}
support and OPTIONSFILE.
I would vote for appending PKGNAMESUFFIX to UNIQUENAME.

Regards,
sunpoet

>
> > Mk/bsd.ports.mk says:
> >
> >   # UNIQUENAME - A name for your port that is globally unique.  By
> default,
> >   #              this is set to ${LATEST_LINK} when LATEST_LINK is set,
> >   #              and to ${PKGNAMEPREFIX}${PORTNAME} otherwise.
> >
> >   # PKGNAMESUFFIX - Suffix to specify compilation options or a version
> >   #                 designator (in case there are different versions of
> >   #                 one port as is the case for Tcl).
> >
> > If PKGNAMESUFFIX is used as Mat suggested (to carry a version number),
> > shouldn't it become part of UNIQUENAME?  Can this be improved?
> >
> > If not, I guess I should just revert PORTNAME to gcc47, gcc48, and
> > so forth.  Using PKGNAMESUFFIX is not providing the expected benefit.
> > And things just are becoming more complex. :-(
> >
> > Gerald
> >
>
>
> --
> Regards,
> Bryan Drewery
>
>