Upgrade Mesa to latest Mesa-CVS?

[Hans de Goede]

Hi all,

A friend of mine has rescently donated me an ati 9800 to test the new
r300 dri driver. I iimmediatly put it in my machine and ran some tests,
after that a filed a number of bugs in xorg's bugzilla and went back to
my trsutworthy 9200. Yesterday I got a reaction to one of the filed bugs
to try the latest CVS because that should have the bug fixed.

So today I tried the latest Mesa CVS on an 9800, and all the bugs I
previously experienced are gone. This bugs are still present in the
latest Fedora mesa packages.

Would it be an idea to update to mesa CVs, or maybe update the r300
driver to CVS if such a piecemeal upgrade is possible?

Regards,

Hans

[Adam Jackson]

Hans de Goede wrote:
>  So today I tried the latest Mesa CVS on an 9800, and all the bugs I
>  previously experienced are gone. This bugs are still present in the
>  latest Fedora mesa packages.
>  
>  Would it be an idea to update to mesa CVs, 
Requires moderately non-trivial fixes to the X server build system too, 
which is why I haven't done it yet.  Probably should do though.  I just 
hate to be in a situation where we're shipping CVS bits in FC6; I keep 
hoping Mesa will hurry up and release a 6.5.1 already.

>  or maybe update the r300
>  driver to CVS if such a piecemeal upgrade is possible? 
It's not.

- ajax

[Kevin Kofler]

Adam Jackson <ajackson <at> redhat.com> writes:
>  Requires moderately non-trivial fixes to the X server build system
> too, 
>  which is why I haven't done it yet.  Probably should do though.  I just 
>  hate to be in a situation where we're shipping CVS bits in FC6; I keep 
>  hoping Mesa will hurry up and release a 6.5.1 already. 
Maybe that's just me, but I'd say CVS bits which work are better than a release 
that doesn't work. ;-)

        Kevin Kofler

[Hans de Goede]

Hi all,

Adam Jackson <ajackson <at> redhat.com> writes:
>  Requires moderately non-trivial fixes to the X server build system
> too, 
>  which is why I haven't done it yet.  Probably should do though.  I just 
>  hate to be in a situation where we're shipping CVS bits in FC6; I keep 
>  hoping Mesa will hurry up and release a 6.5.1 already.
>   

Anything I can do to help? Are you really planning on updating mesa for
FC-6, or would it be worth my time to see if I can isolate a few
important r300 fixes and backport those?

Regards,

Hans

[Adam Jackson]

Hans de Goede wrote:
>  Adam Jackson <ajackson <at> redhat.com> writes:
> > Requires moderately non-trivial fixes to the X server build system too, 
> > which is why I haven't done it yet.  Probably should do though.  I just 
> > hate to be in a situation where we're shipping CVS bits in FC6; I keep 
> > hoping Mesa will hurry up and release a 6.5.1 already.
>  
>  Anything I can do to help? Are you really planning on updating mesa for
>  FC-6, or would it be worth my time to see if I can isolate a few
>  important r300 fixes and backport those? 
Planning a backport, yes.  There are apparently enough Intel and <=R200 
issues fixed in CVS that it should just be done whole.  Hopefully 6.5.1 
happens Really Freaking Soon so I don't have to ship CVS bits in FC6.

That said, if updated srpms for mesa and xorg-x11-server were to 
magically appear in Bugzilla for me, I certainly wouldn't object ;)

- ajax

[Hans de Goede]

Adam Jackson wrote:
>  Hans de Goede wrote:
> > Adam Jackson <ajackson <at> redhat.com> writes:
> >> Requires moderately non-trivial fixes to the X server build system
> >> too, which is why I haven't done it yet.  Probably should do though. 
> >> I just hate to be in a situation where we're shipping CVS bits in
> >> FC6; I keep hoping Mesa will hurry up and release a 6.5.1 already.
> >
> > Anything I can do to help? Are you really planning on updating mesa for
> > FC-6, or would it be worth my time to see if I can isolate a few
> > important r300 fixes and backport those?
>  
>  Planning a backport, yes.  There are apparently enough Intel and <=R200
>  issues fixed in CVS that it should just be done whole.  Hopefully 6.5.1
>  happens Really Freaking Soon so I don't have to ship CVS bits in FC6.
>  
>  That said, if updated srpms for mesa and xorg-x11-server were to
>  magically appear in Bugzilla for me, I certainly wouldn't object ;)
>   
Magically appearing SRPM for latest mesa-CVS filed here:
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=201365

I've been using this with great success on an x86_64 in both 64 and 32
bit mode (I compiled it in a 32 bit chroot and installed the result on
my 64 bit Fedora for use with googleearth).

You also speak about xorg-x11-server changes, I haven't got any for you
as things work fine for me with the current xorg-x11-server in RawHide.

Regards,

Hans

[Adam Jackson]

Hans de Goede wrote:
>  
>  Adam Jackson wrote:
> > Hans de Goede wrote:
> >> Adam Jackson <ajackson <at> redhat.com> writes:
> >>> Requires moderately non-trivial fixes to the X server build system
> >>> too, which is why I haven't done it yet.  Probably should do though. 
> >>> I just hate to be in a situation where we're shipping CVS bits in
> >>> FC6; I keep hoping Mesa will hurry up and release a 6.5.1 already.
> >> Anything I can do to help? Are you really planning on updating mesa for
> >> FC-6, or would it be worth my time to see if I can isolate a few
> >> important r300 fixes and backport those?
> > Planning a backport, yes.  There are apparently enough Intel and <=R200
> > issues fixed in CVS that it should just be done whole.  Hopefully 6.5.1
> > happens Really Freaking Soon so I don't have to ship CVS bits in FC6.
> >
> > That said, if updated srpms for mesa and xorg-x11-server were to
> > magically appear in Bugzilla for me, I certainly wouldn't object ;)
> >
>  
>  Magically appearing SRPM for latest mesa-CVS filed here:
>  https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=201365
>  
>  I've been using this with great success on an x86_64 in both 64 and 32
>  bit mode (I compiled it in a 32 bit chroot and installed the result on
>  my 64 bit Fedora for use with googleearth).
>  
>  You also speak about xorg-x11-server changes, I haven't got any for you
>  as things work fine for me with the current xorg-x11-server in RawHide. 
I was expecting that build system fixes would be required, since 
xorg-x11-server builds against the mesa-source package.  Have you 
attempted to rebuild the X server with a mesa-source from CVS?

- ajax

[Hans de Goede]

Adam Jackson wrote:
>  Hans de Goede wrote:
> >
> > Adam Jackson wrote:
> >> Hans de Goede wrote:
> >>> Adam Jackson <ajackson <at> redhat.com> writes:
> >>>> Requires moderately non-trivial fixes to the X server build system
> >>>> too, which is why I haven't done it yet.  Probably should do
> >>>> though. I just hate to be in a situation where we're shipping CVS
> >>>> bits in
> >>>> FC6; I keep hoping Mesa will hurry up and release a 6.5.1 already.
> >>> Anything I can do to help? Are you really planning on updating mesa for
> >>> FC-6, or would it be worth my time to see if I can isolate a few
> >>> important r300 fixes and backport those?
> >> Planning a backport, yes.  There are apparently enough Intel and <=R200
> >> issues fixed in CVS that it should just be done whole.  Hopefully 6.5.1
> >> happens Really Freaking Soon so I don't have to ship CVS bits in FC6.
> >>
> >> That said, if updated srpms for mesa and xorg-x11-server were to
> >> magically appear in Bugzilla for me, I certainly wouldn't object ;)
> >>
> >
> > Magically appearing SRPM for latest mesa-CVS filed here:
> > https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=201365
> >
> > I've been using this with great success on an x86_64 in both 64 and 32
> > bit mode (I compiled it in a 32 bit chroot and installed the result on
> > my 64 bit Fedora for use with googleearth).
> >
> > You also speak about xorg-x11-server changes, I haven't got any for you
> > as things work fine for me with the current xorg-x11-server in RawHide.
>  
>  I was expecting that build system fixes would be required, since
>  xorg-x11-server builds against the mesa-source package.  Have you
>  attempted to rebuild the X server with a mesa-source from CVS?
>   
nope, I didn't know that the xserver is build against mesa-sources,
strange I would have thought the 2 would be pretty independent.

Eitherway I'll be gone for a couple of days now, so I cannot try
building the X-server. Why don't you give it a shot and let me know the
result? If it turns out it needs some work, but is fixable and you would
be willing to update to mesa CVS if the work gets done for you I'm
probably willing to fix this once I'm back.

Regards,

Hans

[David Nielsen]

lør, 05 08 2006 kl. 00:12 +0200, skrev Hans de Goede:
>  
>  Adam Jackson wrote:
>  > Hans de Goede wrote:
>  >>
>  >> Adam Jackson wrote:
>  >>> Hans de Goede wrote:
>  >>>> Adam Jackson <ajackson <at> redhat.com> writes:
>  >>>>> Requires moderately non-trivial fixes to the X server build system
>  >>>>> too, which is why I haven't done it yet.  Probably should do
>  >>>>> though. I just hate to be in a situation where we're shipping
> CVS
>  >>>>> bits in
>  >>>>> FC6; I keep hoping Mesa will hurry up and release a 6.5.1 already.
>  >>>> Anything I can do to help? Are you really planning on updating mesa for
>  >>>> FC-6, or would it be worth my time to see if I can isolate a few
>  >>>> important r300 fixes and backport those?
>  >>> Planning a backport, yes.  There are apparently enough Intel and <=R200
>  >>> issues fixed in CVS that it should just be done whole.  Hopefully 6.5.1
>  >>> happens Really Freaking Soon so I don't have to ship CVS bits in FC6.
>  >>>
>  >>> That said, if updated srpms for mesa and xorg-x11-server were to
>  >>> magically appear in Bugzilla for me, I certainly wouldn't object ;)
>  >>>
>  >>
>  >> Magically appearing SRPM for latest mesa-CVS filed here:
>  >> https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=201365
>  >>
>  >> I've been using this with great success on an x86_64 in both 64 and 32
>  >> bit mode (I compiled it in a 32 bit chroot and installed the result on
>  >> my 64 bit Fedora for use with googleearth).
>  >>
>  >> You also speak about xorg-x11-server changes, I haven't got any for you
>  >> as things work fine for me with the current xorg-x11-server in RawHide.
>  > 
>  > I was expecting that build system fixes would be required, since
>  > xorg-x11-server builds against the mesa-source package.  Have you
>  > attempted to rebuild the X server with a mesa-source from CVS?
>  > 
>  
>  nope, I didn't know that the xserver is build against mesa-sources,
>  strange I would have thought the 2 would be pretty independent.
>  
>  Eitherway I'll be gone for a couple of days now, so I cannot try
>  building the X-server. Why don't you give it a shot and let me know the
>  result? If it turns out it needs some work, but is fixable and you would
>  be willing to update to mesa CVS if the work gets done for you I'm
>  probably willing to fix this once I'm back.
>  
>  Regards,
>  
>  Hans 
I decided to give this a spin on my x86_64 machine and it hit the
following:

mklib: Making Linux shared library:  libglut.so.3.7.1
/usr/bin/ld: cannot find -lXi
collect2: ld returnerede afslutningskoden 1
mklib: Installing libglut.so.3.7.1 libglut.so.3 libglut.so
in ../../../lib64
mv: kan ikke udføre stat() 'libglut.so.3.7.1': Ingen sådan fil eller
filkatalog
make[3]: *** [../../../lib64/libglut.so] Fejl 1
make[3]: Forlader katalog '/usr/src/redhat/BUILD/Mesa-6.5/src/glut/glx'
make[2]: *** [subdirs] Fejl 1
make[2]: Forlader katalog '/usr/src/redhat/BUILD/Mesa-6.5/src'
make[1]: *** [default] Fejl 1
make[1]: Forlader katalog '/usr/src/redhat/BUILD/Mesa-6.5'
make: *** [linux-dri-x86-64] Fejl 2
fejl: Fejl-afslutningsstatus fra /var/tmp/rpm-tmp.44934 (%build)


- David Nielsen

[Adam Jackson]

David Nielsen wrote:
>  lør, 05 08 2006 kl. 00:12 +0200, skrev Hans de Goede:
> > nope, I didn't know that the xserver is build against mesa-sources,
> > strange I would have thought the 2 would be pretty independent.
> >
> > Eitherway I'll be gone for a couple of days now, so I cannot try
> > building the X-server. Why don't you give it a shot and let me know the
> > result? If it turns out it needs some work, but is fixable and you would
> > be willing to update to mesa CVS if the work gets done for you I'm
> > probably willing to fix this once I'm back.
> >
> > Regards,
> >
> > Hans
>  
>  I decided to give this a spin on my x86_64 machine and it hit the
>  following:
>  
>  mklib: Making Linux shared library:  libglut.so.3.7.1
>  /usr/bin/ld: cannot find -lXi 
I'll take a swing at this this weekend.  Thanks!

- ajax

[Hans de Goede]

Install libXi-devel and try again.

Regards,

Hans


David Nielsen wrote:
>  lør, 05 08 2006 kl. 00:12 +0200, skrev Hans de Goede:
> > Adam Jackson wrote:
> >> Hans de Goede wrote:
> >>> Adam Jackson wrote:
> >>>> Hans de Goede wrote:
> >>>>> Adam Jackson <ajackson <at> redhat.com> writes:
> >>>>>> Requires moderately non-trivial fixes to the X server build
> system
> >>>>>> too, which is why I haven't done it yet.  Probably should do
> >>>>>> though. I just hate to be in a situation where we're shipping
> CVS
> >>>>>> bits in
> >>>>>> FC6; I keep hoping Mesa will hurry up and release a 6.5.1
> already.
> >>>>> Anything I can do to help? Are you really planning on updating mesa
> for
> >>>>> FC-6, or would it be worth my time to see if I can isolate a few
> >>>>> important r300 fixes and backport those?
> >>>> Planning a backport, yes.  There are apparently enough Intel and
> <=R200
> >>>> issues fixed in CVS that it should just be done whole.  Hopefully 6.5.1
> >>>> happens Really Freaking Soon so I don't have to ship CVS bits in
> FC6.
> >>>>
> >>>> That said, if updated srpms for mesa and xorg-x11-server were to
> >>>> magically appear in Bugzilla for me, I certainly wouldn't object ;)
> >>>>
> >>> Magically appearing SRPM for latest mesa-CVS filed here:
> >>> https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=201365
> >>>
> >>> I've been using this with great success on an x86_64 in both 64 and 32
> >>> bit mode (I compiled it in a 32 bit chroot and installed the result on
> >>> my 64 bit Fedora for use with googleearth).
> >>>
> >>> You also speak about xorg-x11-server changes, I haven't got any for you
> >>> as things work fine for me with the current xorg-x11-server in RawHide.
> >> I was expecting that build system fixes would be required, since
> >> xorg-x11-server builds against the mesa-source package.  Have you
> >> attempted to rebuild the X server with a mesa-source from CVS?
> >>
> > nope, I didn't know that the xserver is build against mesa-sources,
> > strange I would have thought the 2 would be pretty independent.
> >
> > Eitherway I'll be gone for a couple of days now, so I cannot try
> > building the X-server. Why don't you give it a shot and let me know the
> > result? If it turns out it needs some work, but is fixable and you would
> > be willing to update to mesa CVS if the work gets done for you I'm
> > probably willing to fix this once I'm back.
> >
> > Regards,
> >
> > Hans
>  
>  I decided to give this a spin on my x86_64 machine and it hit the
>  following:
>  
>  mklib: Making Linux shared library:  libglut.so.3.7.1
>  /usr/bin/ld: cannot find -lXi
>  collect2: ld returnerede afslutningskoden 1
>  mklib: Installing libglut.so.3.7.1 libglut.so.3 libglut.so
>  in ../../../lib64
>  mv: kan ikke udføre stat() 'libglut.so.3.7.1': Ingen sådan fil eller
>  filkatalog
>  make[3]: *** [../../../lib64/libglut.so] Fejl 1
>  make[3]: Forlader katalog '/usr/src/redhat/BUILD/Mesa-6.5/src/glut/glx'
>  make[2]: *** [subdirs] Fejl 1
>  make[2]: Forlader katalog '/usr/src/redhat/BUILD/Mesa-6.5/src'
>  make[1]: *** [default] Fejl 1
>  make[1]: Forlader katalog '/usr/src/redhat/BUILD/Mesa-6.5'
>  make: *** [linux-dri-x86-64] Fejl 2
>  fejl: Fejl-afslutningsstatus fra /var/tmp/rpm-tmp.44934 (%build)
>  
>  
>  - David Nielsen
>

[David Nielsen]

lør, 05 08 2006 kl. 06:32 +0200, skrev Hans de Goede:
>  Install libXi-devel and try again. 
Yep, that did indeed solve the problem, might I suggest BR'ing it?

- David

[Hans de Goede]

David Nielsen wrote:
>  lør, 05 08 2006 kl. 06:32 +0200, skrev Hans de Goede:
> > Install libXi-devel and try again.
>  
>  Yep, that did indeed solve the problem, might I suggest BR'ing it?
>  
Actually the build so be fixed not to build the included glut as that is 
not oss, but as a quick hack I just rm the builded glut in %install.

Regards,

Hans

[Mike A. Harris]

Hans de Goede wrote:
>  
>  
>  David Nielsen wrote:
> > lør, 05 08 2006 kl. 06:32 +0200, skrev Hans de Goede:
> >> Install libXi-devel and try again.
> >
> > Yep, that did indeed solve the problem, might I suggest BR'ing it?
> >
>  
>  Actually the build so be fixed not to build the included glut as that is 
>  not oss, but as a quick hack I just rm the builded glut in %install. 
Mesa does not include GLUT in the MesaLib tarball anymore, and hasn't
for about a year and a half.  It is included in MesaGLUT tarball, which
is not shipped in Fedora Core.

Fedora Core has shipped 'freeglut' all along, and in RHL prior to that.
freeglut itself is moving to Extras for FC6 (if it hasn't already been
moved).


-- 
Mike A. Harris  *  Open Source Advocate  *  http://mharris.ca
                       Proud Canadian.

[Hans de Goede]

Mike A. Harris wrote:
>  Hans de Goede wrote:
> >
> >
> > David Nielsen wrote:
> >> lør, 05 08 2006 kl. 06:32 +0200, skrev Hans de Goede:
> >>> Install libXi-devel and try again.
> >>
> >> Yep, that did indeed solve the problem, might I suggest BR'ing it?
> >>
> >
> > Actually the build so be fixed not to build the included glut as that
> > is not oss, but as a quick hack I just rm the builded glut in %install.
>  
>  Mesa does not include GLUT in the MesaLib tarball anymore, and hasn't
>  for about a year and a half.  It is included in MesaGLUT tarball, which
>  is not shipped in Fedora Core.
>   
Well appearantly it does include GLUT in a CVS checkout.

Regards,

Hans

[Adam Jackson]

Hans de Goede wrote:
>  Well appearantly it does include GLUT in a CVS checkout. 
Which is why we hack the buildsystem to not build it.  It's 
redistributable but not free.  And of course, since we don't build it, 
we never needed to add the BuildReq on libXi-devel.

Kristian started merging our patchset into Mesa 6.5.1 yesterday, should 
be out soonish I hope after we do some smoketesting.

- ajax

[Hans de Goede]

Adam Jackson wrote:
>  Hans de Goede wrote:
> > Well appearantly it does include GLUT in a CVS checkout.
>  
>  Which is why we hack the buildsystem to not build it.  It's
>  redistributable but not free.  And of course, since we don't build it,
>  we never needed to add the BuildReq on libXi-devel.
>  
>  Kristian started merging our patchset into Mesa 6.5.1 yesterday, should
>  be out soonish I hope after we do some smoketesting.
>   
So 6.5.1 is out, cool! May I ask for your attention to the
mesa-cvs-r300-smooth-line.patch patch which is in the SRPM I "attached" too:
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=201365

Get the SRPM directly here:
http://people.atrpms.net/~hdegoede/mesa-6.5-20060802.src.rpm

This might be a good starting point for the patchset as it contains the
patchset in Fedora's SRPM ported to a pretty rescent mesa-CVS.

Anyways this patch does the following (as described in the referenced BZ
ticket):

---

There is one peculiar patch in my SRPM which needs some clarification:
mesa-cvs-r300-smooth-line.patch

This comments this 1 line in r300_renderer.c:
FALLBACK_IF(ctx->Line.SmoothFlag); // GL_LINE_SMOOTH

the result of this is that if an application has requested anti-aliased
lines,
which currently are not implemented in the r300 driver, the application gets
normal (not antialiased) lines instead of falling back to the software
renderer.

This patch is needed to make googleearth and smc
(http://www.secretmaryo.org/)
usable IMHO having none antialiased lines is way better then 2 fps.

The same could be argued for anti-aliased points (the line below this
one), but
I haven't encountered any apps using this one yet.

Notice that this change was suggested by upstream and that upstream has even
concidered doing this themselves, they haven;t however because they are
planning
to fix this by actually implementing antialiased lines soon.

---

Thanks & Regards,

Hans

[Adam Jackson]

Hans de Goede wrote:
>  Adam Jackson wrote:
> > Hans de Goede wrote:
> >> Well appearantly it does include GLUT in a CVS checkout.
> > Which is why we hack the buildsystem to not build it.  It's
> > redistributable but not free.  And of course, since we don't build it,
> > we never needed to add the BuildReq on libXi-devel.
> >
> > Kristian started merging our patchset into Mesa 6.5.1 yesterday, should
> > be out soonish I hope after we do some smoketesting.
>  So 6.5.1 is out, cool! May I ask for your attention to the
>  mesa-cvs-r300-smooth-line.patch patch which is in the SRPM I "attached" too:
>  https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=201365 
Uh, I wrote this wrong, should have been "merging our patchset into the 
CVS SRPM you did".  I only which 6.5.1 were out...

>  Get the SRPM directly here:
>  http://people.atrpms.net/~hdegoede/mesa-6.5-20060802.src.rpm
>  
>  This might be a good starting point for the patchset as it contains the
>  patchset in Fedora's SRPM ported to a pretty rescent mesa-CVS. 
Yeah, we'll pop that in.  Thanks!

- ajax

[Adam Jackson]

Adam Jackson wrote:
>  Hans de Goede wrote:
> > Adam Jackson wrote:
> >> Hans de Goede wrote:
> >>> Well appearantly it does include GLUT in a CVS checkout.
> >> Which is why we hack the buildsystem to not build it.  It's
> >> redistributable but not free.  And of course, since we don't build it,
> >> we never needed to add the BuildReq on libXi-devel.
> >>
> >> Kristian started merging our patchset into Mesa 6.5.1 yesterday, should
> >> be out soonish I hope after we do some smoketesting.
> > So 6.5.1 is out, cool! May I ask for your attention to the
> > mesa-cvs-r300-smooth-line.patch patch which is in the SRPM I 
> > "attached" too:
> > https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=201365
>  
>  Uh, I wrote this wrong, should have been "merging our patchset into the 
>  CVS SRPM you did".  I only which 6.5.1 were out... 
Wish.  Brain no work good.

- ajax

[Hans de Goede]

Adam Jackson wrote:
>  Adam Jackson wrote:
> > Hans de Goede wrote:
> >> Adam Jackson wrote:
> >>> Hans de Goede wrote:
> >>>> Well appearantly it does include GLUT in a CVS checkout.
> >>> Which is why we hack the buildsystem to not build it.  It's
> >>> redistributable but not free.  And of course, since we don't build it,
> >>> we never needed to add the BuildReq on libXi-devel.
> >>>
> >>> Kristian started merging our patchset into Mesa 6.5.1 yesterday, should
> >>> be out soonish I hope after we do some smoketesting.
> >> So 6.5.1 is out, cool! May I ask for your attention to the
> >> mesa-cvs-r300-smooth-line.patch patch which is in the SRPM I
> >> "attached" too:
> >> https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=201365
> >
> > Uh, I wrote this wrong, should have been "merging our patchset into
> > the CVS SRPM you did".  I only which 6.5.1 were out...
>  
>  Wish.  Brain no work good.
>   
No problem, any verdict on the anti-aliased lines hack I included? I
really want to see it in, otherwise I will still have to roll my own
version with it, and don't underestimate the popularity of googleearth
(closed sourced unfortunatly).

Regards,

Hans

[Hans de Goede]

David Nielsen wrote:
>  lør, 05 08 2006 kl. 06:32 +0200, skrev Hans de Goede:
> > Install libXi-devel and try again.
>  
>  Yep, that did indeed solve the problem, might I suggest BR'ing it?
>   
And that it has been build, does it work for you? Any improvements /
regressions?

Regards,

Hans

[David Nielsen]

lør, 05 08 2006 kl. 09:47 +0200, skrev Hans de Goede:
>  
>  David Nielsen wrote:
>  > lør, 05 08 2006 kl. 06:32 +0200, skrev Hans de Goede:
>  >> Install libXi-devel and try again.
>  > 
>  > Yep, that did indeed solve the problem, might I suggest BR'ing it?
>  > 
>  
>  And that it has been build, does it work for you? Any improvements /
>  regressions? 
It seems about the same on my r300, that is a good thing though, it's
stable enough to run Compiz on a day to day basis and everything is
generally peachy.

I still a few glitzes when displaying menus in quick succesion but
that's about the only thing that doesn't work as expected I can think of
(aside the known issues like lack of xv support).

- David

[Kristian Kristensen]

David Nielsen wrote:
>  lør, 05 08 2006 kl. 09:47 +0200, skrev Hans de Goede:
> > David Nielsen wrote:
> >> lør, 05 08 2006 kl. 06:32 +0200, skrev Hans de Goede:
> >>> Install libXi-devel and try again.
> >> Yep, that did indeed solve the problem, might I suggest BR'ing it?
> >>
> > And that it has been build, does it work for you? Any improvements /
> > regressions?
>  
>  It seems about the same on my r300, that is a good thing though, it's
>  stable enough to run Compiz on a day to day basis and everything is
>  generally peachy.
>  
>  I still a few glitzes when displaying menus in quick succesion but
>  that's about the only thing that doesn't work as expected I can think of
>  (aside the known issues like lack of xv support). 
This is not a mesa glitch, though, it's a known compiz problem.

Kristian

[David Nielsen]

man, 07 08 2006 kl. 10:56 -0400, skrev Kristian Høgsberg:
>  David Nielsen wrote:
>  > lør, 05 08 2006 kl. 09:47 +0200, skrev Hans de Goede:
>  >> David Nielsen wrote:
>  >>> lør, 05 08 2006 kl. 06:32 +0200, skrev Hans de Goede:
>  >>>> Install libXi-devel and try again.
>  >>> Yep, that did indeed solve the problem, might I suggest BR'ing it?
>  >>>
>  >> And that it has been build, does it work for you? Any improvements /
>  >> regressions?
>  > 
>  > It seems about the same on my r300, that is a good thing though, it's
>  > stable enough to run Compiz on a day to day basis and everything is
>  > generally peachy.
>  > 
>  > I still a few glitzes when displaying menus in quick succesion but
>  > that's about the only thing that doesn't work as expected I can think of
>  > (aside the known issues like lack of xv support).
>  
>  This is not a mesa glitch, though, it's a known compiz problem. 
Things took a bad turn somewhere, now using compiz the entire system
hangs if I look at it wrong. Even after I downgraded Mesa.. very scary
indeed.

I have no idea what causes this and it's not like I can grab a backtrace
easily as it locks up the machine in the really bad cases or locks up X
in the mild cases.. I have no idea what caused it but I have a feeling I
was involved.

- David

[Kristian Kristensen]

David Nielsen wrote:
>  man, 07 08 2006 kl. 10:56 -0400, skrev Kristian Høgsberg:
> > David Nielsen wrote:
> >> lør, 05 08 2006 kl. 09:47 +0200, skrev Hans de Goede:
> >>> David Nielsen wrote:
> >>>> lør, 05 08 2006 kl. 06:32 +0200, skrev Hans de Goede:
> >>>>> Install libXi-devel and try again.
> >>>> Yep, that did indeed solve the problem, might I suggest BR'ing it?
> >>>>
> >>> And that it has been build, does it work for you? Any improvements /
> >>> regressions?
> >> It seems about the same on my r300, that is a good thing though, it's
> >> stable enough to run Compiz on a day to day basis and everything is
> >> generally peachy.
> >>
> >> I still a few glitzes when displaying menus in quick succesion but
> >> that's about the only thing that doesn't work as expected I can think of
> >> (aside the known issues like lack of xv support).
> > This is not a mesa glitch, though, it's a known compiz problem.
>  
>  Things took a bad turn somewhere, now using compiz the entire system
>  hangs if I look at it wrong. Even after I downgraded Mesa.. very scary
>  indeed.
>  
>  I have no idea what causes this and it's not like I can grab a backtrace
>  easily as it locks up the machine in the really bad cases or locks up X
>  in the mild cases.. I have no idea what caused it but I have a feeling I
>  was involved. 
The r300 driver in 6.5 has a bad memory leak, which will screw up your system 
if you run compiz (or any GL app) a few times in indirect mode.  This is fixed 
in mesa cvs, though, so you shouldn't see that there.

Kristian

[Kristian Kristensen]

Hans de Goede wrote:
>  
>  Adam Jackson wrote:
> > Hans de Goede wrote:
> >> Adam Jackson wrote:
> >>> Hans de Goede wrote:
> >>>> Adam Jackson <ajackson <at> redhat.com> writes:
> >>>>> Requires moderately non-trivial fixes to the X server build system
> >>>>> too, which is why I haven't done it yet.  Probably should do
> >>>>> though. I just hate to be in a situation where we're shipping
> CVS
> >>>>> bits in
> >>>>> FC6; I keep hoping Mesa will hurry up and release a 6.5.1 already.
> >>>> Anything I can do to help? Are you really planning on updating mesa for
> >>>> FC-6, or would it be worth my time to see if I can isolate a few
> >>>> important r300 fixes and backport those?
> >>> Planning a backport, yes.  There are apparently enough Intel and <=R200
> >>> issues fixed in CVS that it should just be done whole.  Hopefully 6.5.1
> >>> happens Really Freaking Soon so I don't have to ship CVS bits in FC6.
> >>>
> >>> That said, if updated srpms for mesa and xorg-x11-server were to
> >>> magically appear in Bugzilla for me, I certainly wouldn't object ;)
> >>>
> >> Magically appearing SRPM for latest mesa-CVS filed here:
> >> https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=201365
> >>
> >> I've been using this with great success on an x86_64 in both 64 and 32
> >> bit mode (I compiled it in a 32 bit chroot and installed the result on
> >> my 64 bit Fedora for use with googleearth).
> >>
> >> You also speak about xorg-x11-server changes, I haven't got any for you
> >> as things work fine for me with the current xorg-x11-server in RawHide.
> > I was expecting that build system fixes would be required, since
> > xorg-x11-server builds against the mesa-source package.  Have you
> > attempted to rebuild the X server with a mesa-source from CVS?
> >
>  
>  nope, I didn't know that the xserver is build against mesa-sources,
>  strange I would have thought the 2 would be pretty independent.
>  
>  Eitherway I'll be gone for a couple of days now, so I cannot try
>  building the X-server. Why don't you give it a shot and let me know the
>  result? If it turns out it needs some work, but is fixable and you would
>  be willing to update to mesa CVS if the work gets done for you I'm
>  probably willing to fix this once I'm back. 
Something like the attached (untested) patch might be enough to make the 
xserver build with the cvs mesa rpm.  It's basically just a backport of the 
GLcore Makefile.am changes from the xserver git master branch since 1.1 branched.

Kristian

[David Nielsen]

man, 07 08 2006 kl. 12:25 -0400, skrev Kristian Høgsberg:

>  
>  Something like the attached (untested) patch might be enough to make the 
>  xserver build with the cvs mesa rpm.  It's basically just a backport of the 
>  GLcore Makefile.am changes from the xserver git master branch since 1.1 branched. 
It compiles against the new Mesa so by all accounts you did good work.

- David