Subject: Re: Mesa 7.0.1 update/recent commit
To: Dieter Baron <firstname.lastname@example.org>
From: Blair Sadewitz <email@example.com>
Date: 08/14/2007 15:02:44
I don't see a reason we cannot do this.
1) AFAIK, there aren't an inordinate amount of changes to worry about
wrt: the libraries: The GL library (again, as far as I know) has some
minor differences, such as GLX_GLXEXT_LEGACY -> GLX_GLXEXT_PROTOTYPES.
I'm not sure that libglu libglut have changed significantly at all
(other than bug fixes).
2) IMHO, ideally we should take this opportunity to construct a more
advanced OpenGL framework. It would be nice to have
*_VERSION_REQD-type variables for gl/glw/glu/glut libraries.
Moreover, the SGI GLUT implementation is moribund,
and we should allow the user to choose alternative
3) WRT DRI drivers: During a few of my googling sessions on this
topic, I've seen bug reports/patches which give me reason to suspect
that some of the DRI drivers are not fully PIC-compatable (see the
fedora CVS's package, for instance). I do not consider myself
competent to issue a firm judgement call on this, so I'd appreciate it
if someone could help me look into it. Currently, I use
-fvisibility=hidden for the MesaLib target to hide relocs which ld
doesn't like. And while everything seems fine, and hidden visibility
can be an optimization, I'd like to make sure the "right thing" is
Miscellaneous, pseudo-random thoughts:
1) Could libtool versioning help with this in any way?
2) The mesa source has lists of all dependencies for all targets. I'm
aware of the fact that this might take more effort to maintain, but I
can't seem to shake the temptation to write BSD makefiles for Mesa.
Their build system typifies "ad-hoc-ery". For instance, in various OS
distributions/package systems, I see the same patch to keep the build
system from breaking during parallel builds, while attempting to make
dependencies, etc. These patches have been around for years, and for
some reason the Mesa people, AFAIK, steadfastly neglect (that should
be an oxymoron, but unfortunately it's merely paradoxical) to fix
their build system.
3) I'm not sure that this is worth the effort, but I'm intrigued by
projects like graphics/glew.
Also, I'd welcome any comments on my MesaLib-dri package, as I'm a new
pkgsrc developer and have undoubtedly made some errors. The ugly
subst.mk stuff will go away, though; that was just there so I didn't
have to keep track of an insane number of patches while testing