Date: Sun, 13 Jan 2013 19:53:31 -0700 (MST) From: John Hein <jhein@symmetricom.com> To: FreeBSD-gnats-submit@freebsd.org Cc: mezz@freebsd.org Subject: ports/175276: [patch] devel/py-gobject OPTIONSFILE eval order problem Message-ID: <201301140253.r0E2rVv9076794@gromit.timing.com> Resent-Message-ID: <201301140300.r0E300X8092303@freefall.freebsd.org>
next in thread | raw e-mail | index | archive | help
>Number: 175276
>Category: ports
>Synopsis: [patch] devel/py-gobject OPTIONSFILE eval order problem
>Confidential: no
>Severity: non-critical
>Priority: low
>Responsible: freebsd-ports-bugs
>State: open
>Quarter:
>Keywords:
>Date-Required:
>Class: sw-bug
>Submitter-Id: current-users
>Arrival-Date: Mon Jan 14 03:00:00 UTC 2013
>Closed-Date:
>Last-Modified:
>Originator: John Hein
>Release:
>Organization:
>Environment:
>Description:
devel/py-gobject looks for options settings in one filename and then
generates an options file of a different name.
That is, su -c 'make config' will allow the user to create an options
file, but that file is not the one included by bsd.options.mk
There is a fundamental evaluation order issue with OPTIONSFILE and
UNIQUENAME when using knobs that affect UNIQUENAME such as
PKGNAMEPREFIX. In this case, it causes bsd.options.mk to look for
gobject/options, but py<ver>-gobject/options to be generated (by
do-config in bsd.port.mk).
% cd devel/py-gobject
% make -V PORTNAME -dd | & grep /options; echo ===; make -V PORTNAME -V OPTIONSFILE
Searching for /var/db/ports/gobject/options...Looking for "/var/db/ports/gobject/options"...failed. Returning NULL
Searching for /var/db/ports/gobject/options.local...Looking for "/var/db/ports/gobject/options.local"...failed. Returning NULL
Searching for /var/db/ports/py27-gobject/options...Looking for "/var/db/ports/py27-gobject/options"...failed. Returning NULL
===
gobject
/var/db/ports/py27-gobject/options
This started happening with r309649 (OPTIONSng conversion) on 2012-12-29.
But even if that is solved (so PKGNAMEPREFIX is not ignored in the
first evaluation of OPTIONSFILE), I think we still want to override
the OPTIONSFILE value to remove the python version (e.g.,
py-gobject/options, not py27-gobject/options). So we want to override
the default OPTIONSFILE here. See attached patch which solves both
the eval order issue and the removal of <ver> in py<ver>-gobject.
>How-To-Repeat:
>Fix:
Index: Makefile
===================================================================
--- Makefile (revision 310148)
+++ Makefile (working copy)
@@ -35,6 +35,7 @@
CONFLICTS= py*-gtk-0*
PLIST_SUB= PYTHON_VER="${PYTHON_VER}"
+OPTIONSFILE= ${PORT_DBDIR}/py-${PORTNAME}/options
OPTIONS_DEFINE= LIBFFI
LIBFFI_DESC= Enable libffi support
>Release-Note:
>Audit-Trail:
>Unformatted:
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201301140253.r0E2rVv9076794>