Bugzilla Bug 102325

GfsView is a graphical viewer for Gerris hydrodynamic simulation files.

Created an attachment (id=65816) [details]
gfsview-0.2.ebuild

Created an attachment (id=66279) [details]
gfsview-0.2.ebuild 

Please fix the following and reopen:

* LICENSE is invalid.
* No ChangeLog, README etc to dodoc?

Created an attachment (id=66812) [details]
gfsview-0.2.ebuild

Thanks.  I'm learning all the time.  
  
I think I have addressed all issues.  
 

OK, one more remark - no need to dodoc COPYING. ;)

Created an attachment (id=73571) [details]
gfsview-0.4.ebuild

gfsview-0.4 will not compile against the nvidia opengl headers.  Copied code
from gtkglext to change to xorg-11/opengl implementation during compile, and
return again to original opengl implementation post install.  gfsview, after
compiling against the xorg-x11/opengl headers works fine under nvidia/opengl
implementation. 

Depends on the gerris library, currently (11/25/05) in bugs.gentoo.org
http://bugs.gentoo.org/show_bug.cgi?id=102323

Hi Tim,  
  
Thanks for your ebuild! I had a go at the package but, unfortunately, it   
segfaults on my machine (P4, xorg-x11-6.8.2-r6). Currently, it looks  
to me as if I have hit a sse bug in xorg's radeon dri libs, but I will have to   
look more into this. Here's the relevant gdb output  
  
Program received signal SIGFPE, Arithmetic exception.  
[Switching to Thread -1223526736 (LWP 4340)]  
0xb70396ec in _mesa_test_os_sse_exception_support ()  
   from /usr/lib/modules/dri/radeon_dri.so  
 
Maybe the new modular xorg fixes it - I haven't tried yet.  
 

Thanks for looking at it. 
 
I have an AMD64/Nvidia.  I can try it early next week on a K6-2/CyberBlade.  
 
Can we turn off 'sse'? 
 
Thanks 
Tim 

(In reply to comment #10) 
>   
> Can we turn off 'sse'?  
>   
 
I've recompiled the relevant packages with sse turned off and 
that didn't help. I hope I can have a closer look soon. 
 
Thanks, 
Markus 
 
 
 

I found this mention of a problem with _mesa_test_os_sse_exception_support()
and gdb.

https://bugs.freedesktop.org/show_bug.cgi?id=1709

It is a relatively old bug report so I thought that some resolution would have
already been implemented.  It could mean that the actual error is somewhere
else.

Hi Tim,

Thanks for the update. Let's see if this goes away with the 
new modular X. I'll test it once I've everything set up 
properly.

best,
Markus

Hi Tim,

gfsview compiles and works fine for me using the new modular X. 
Hence, as soon as modular X hits ~x86 I'll commit it to the tree
for further testing.

Thanks,
Markus

Created an attachment (id=78778) [details]
gfsview-0.4.ebuild (without switch to opengl headers)

Was able to compile without problem using the nvidia header files, which
simplfied the ebuild a lot.

Included a patch from the developers that prevents a crash that I was having. 
Works great with the patch.

My machine is an amd64, so should at least maintain that keyword.

Thanks!

Created an attachment (id=78779) [details]
patch to fix crash when clicking on menu

Hi Tim,

Thanks for the update and I am glad that we can dispose
of that OpenGL header switch. I'll give it another spin over
the weekend and will move it into portage if everything
works fine.

Thanks,
Markus

Hi Tim,

Thanks for your great work! I just committed gfsview to portage cvs.
Currently, it is still keyworded ~x86, but I will ask somebody in the
sci-team to test and then mark ~amd64 the whole dependency
chain gts/gerris/gfsview.

Thanks,
Markus