After an evaluation, GNOME has moved from Bugzilla to GitLab. Learn more about GitLab.
No new issues can be reported in GNOME Bugzilla anymore.
To report an issue in a GNOME project, go to GNOME GitLab.
Do not go to GNOME Gitlab for: Bluefish, Doxygen, GnuCash, GStreamer, java-gnome, LDTP, NetworkManager, Tomboy.
Bug 726518 - g-ir-scanner: make --libtool= a no-op on OpenBSD
g-ir-scanner: make --libtool= a no-op on OpenBSD
Status: RESOLVED NOTGNOME
Product: gobject-introspection
Classification: Platform
Component: general
2.39.x
Other OpenBSD
: Normal normal
: ---
Assigned To: gobject-introspection Maintainer(s)
gobject-introspection Maintainer(s)
Depends on:
Blocks:
 
 
Reported: 2014-03-17 12:21 UTC by Antoine Jacoutot
Modified: 2015-02-07 17:01 UTC
See Also:
GNOME target: ---
GNOME version: ---


Attachments
force libtool to /usr/bin/libtool on OpenBSD (1.07 KB, patch)
2014-03-17 12:21 UTC, Antoine Jacoutot
none Details | Review

Description Antoine Jacoutot 2014-03-17 12:21:57 UTC
Created attachment 272137 [details] [review]
force libtool to /usr/bin/libtool on OpenBSD

Hi.

I would like to get some comments on the following patch. We've been using some variations of this in the last 3 years in OpenBSD ports tree.

libtool(1) on OpenBSD is a completely different program and is part of
the OS; GNU libtool does not properly work there (which is what softwares usually
bundle in their release tarballs) -- so make it a no-op.

Any opinion or better way to handle this?
Thank you.
Comment 1 Allison Karlitskaya (desrt) 2014-03-17 13:22:31 UTC
We have some of the same issues on FreeBSD mostly due to the fact that libtool on FreeBSD is a disaster.  We solve this by passing a LIBTOOL= argument to make, which works fine for now.

I think you should do that if you want to do this kind of overriding because otherwise you're going to have to petition a lot of people to make changes like this...

Meanwhile, maybe you could contribute your OpenBSD-specific improvements to upstream libtool?  This would benefit people who just casually download stuff and build it on OpenBSD (which would happen using the in-tree libtool, by default).
Comment 2 Antoine Jacoutot 2014-03-17 13:29:34 UTC
(In reply to comment #1)
> We have some of the same issues on FreeBSD mostly due to the fact that libtool
> on FreeBSD is a disaster.  We solve this by passing a LIBTOOL= argument to
> make, which works fine for now.

Hi Ryan.

Yes but this does not work. Passing LIBTOOL= works for autotools, not for things that use stuffs like:
g-ir-scanner --libtool=../../libtool

> I think you should do that if you want to do this kind of overriding because
> otherwise you're going to have to petition a lot of people to make changes like
> this...
> 
> Meanwhile, maybe you could contribute your OpenBSD-specific improvements to
> upstream libtool?  This would benefit people who just casually download stuff
> and build it on OpenBSD (which would happen using the in-tree libtool, by
> default).

We tried that a countless times, it ended up being a lost cause. Which is why we started our own libtool. Not to mention that patches that "may" end up in upstream libtool make take a while to be included in the libtool script some projects are bundling.
Comment 3 Allison Karlitskaya (desrt) 2014-03-17 13:36:55 UTC
(In reply to comment #2)
> Yes but this does not work. Passing LIBTOOL= works for autotools, not for
> things that use stuffs like:
> g-ir-scanner --libtool=../../libtool

If projects are hardcoding this in their use of g-ir-scanner then that is their fault.  The common use here is to use $(LIBTOOL) (and the real common use it to use the .m4 macros -- and they do that).

I've been dealing with those cases by filing bugs against individual projects with patches to move them over to using the macros and everyone has been happy to accept those patches (after some light prodding).

> We tried that a countless times, it ended up being a lost cause. Which is why
> we started our own libtool. Not to mention that patches that "may" end up in
> upstream libtool make take a while to be included in the libtool script some
> projects are bundling.

I feel your pain there, having tried to contribute to libtool in the past.  For what it's worth, upstream libtool recently realised that they have had a rather high barrier to contribution and are trying to do something about it.

  http://lists.gnu.org/archive/html/libtool-patches/2013-12/msg00000.html

It might be useful to try again?
Comment 4 Antoine Jacoutot 2014-03-17 13:39:48 UTC
> If projects are hardcoding this in their use of g-ir-scanner then that is their
> fault.  The common use here is to use $(LIBTOOL) (and the real common use it to
> use the .m4 macros -- and they do that).
> 
> I've been dealing with those cases by filing bugs against individual projects
> with patches to move them over to using the macros and everyone has been happy
> to accept those patches (after some light prodding).

OK fair enough. I'll just keep it as a local patch in our ports tree.
Thanks for the comment.

> I feel your pain there, having tried to contribute to libtool in the past.  For
> what it's worth, upstream libtool recently realised that they have had a rather
> high barrier to contribution and are trying to do something about it.
> 
>   http://lists.gnu.org/archive/html/libtool-patches/2013-12/msg00000.html
> 
> It might be useful to try again?

It is too late now, we have moved on with our own libtool(1).
Comment 5 André Klapper 2015-02-07 17:01:38 UTC
[Mass-moving gobject-introspection tickets to its own Bugzilla product - see bug 708029. Mass-filter your bugmail for this message: introspection20150207 ]