On 2010/07/20 12:12, Sebastian Reitenbach wrote:
> Hi,
> On Tuesday 20 July 2010 09:21:53 am Landry Breuil wrote:
> > On Tue, Jul 20, 2010 at 07:42:50AM +0200, Sebastian Reitenbach wrote:
> > > Hi,
> > >
> > > On Monday 19 July 2010 09:35:14 pm Landry Breuil wrote:
> > > > On Sun, Jul 18, 2010 at 07:34:54PM +0200, Sebastian Reitenbach wrote:
> > > > > Hi,
> > > > >
> > > > > appended three new tcl/tk related ports. Since the last submission a
> > > > > couple of weeks ago, more or less completely reworked.
> > > > >
> > > > > graphics/tkimg:
> > > > > This package enhances Tk, adding support for many
> > > > > other Image formats: BMP, XBM, XPM, GIF (with
> > > > > transparency, but without LZW), PNG, JPEG, TIFF.
> > > >
> > > > patch-Makefile_in looks wrong to me. what's that fucking up with fake
> > > > dirs ? And what rm -r ${DESTDIR}/Img is doing in post-install ? that's
> > > > a two-step install target ?
> > >
> > > I'll take a look at it again.
> > >
> > > > patch-XXX_configure should contain LIBfoo_VERSION (+ added to
> > > > SUBST_VARS) to be cleaner... oh well. whatever. If there's only one
> > > > SHLIB_VERSION.. forget about it :)
> > > >
> > > > -x-includes should use ${X11BASE}
> > >
> > > oops, I missed that.
> > >
> > > > replace that ugly sed+mv in post-install by a perl -pi construct
> > > >
> > > > why not trying to fix coccinella to work with png package instead of
> > > > just disabling it ?
> > >
> > > Coccinella requires graphics/tkpng to start up. The graphics/tkimg is for
> > > additional image formats that can be used with the whiteboard in
> > > coccinella. the regression test in tkimg segfaults when it comes to test
> > > the png stuff. Also when I try to startup coccinella with the png package
> > > enabled it segfaults. I'll need to take a look again to try
> > > debugging/figuring out why it segfaults, I failed so far.
> >
> > ok, makes sense then - as long as there's a comment in Makefile :)
>
> appended a reworked tkimg tarball.
> changes: * now using ${X11BASE} in the right place.
> * replaced sed/mv with perl -pi
> * comment in the makefile regarding the broken img::png package
> * added a comment to the header of the patch-Makefile_in, why I
>
> do the strange stuff in there (its indeed a two stage install)
> * remove the ${DESTDIR}/Img not in the ports Makefile anymore,
> updated the patch-Makefile_in to remove the dir in the
> install: stage
>
> I think that was all I did. Hope its better now. If there is a better way to
> handle that two stage install, please let me know about it, or point me to a
> port with a similar "problem"
>
> cheers,
> Sebastian
>
>
A couple of nits:
This no longer pulls in pthread, so that can be removed
from WANTLIB.
It wants something like REGRESS_DEPENDS=::${PKGPATH} as the tests
don't run unless it's installed.
And one other thing I'm not sure about, but I don't know how it
should be handled, it may be that it's totally ok, or it may be
that it's an unnecessary change..:
-SHARED_LIB_SUFFIX='${TCL_TRIM_DOTS}\$\{DBGX\}.so'
+SHARED_LIB_SUFFIX='${TCL_TRIM_DOTS}\$\{DBGX\}.so.${SHLIB_VERSION}'
I don't have a good handle on the Tcl dlopen()d modules, is there
anyone around who does who can say if this makes sense? For many
ports, this sort of module doesn't get a version number..
