savage: glxinfo makes xgl crash

It would probably be more useful to repeat this with a debug build of Xgl. Can this be provided ? Building it myself is a little out of reach ATM.

Just want to add that everything (glxinfo, DRI, GL applications) works really nice with the standard xserver-xorg/x11-common.

The server log (I hope this is the correct one). If I understand correctly, this is from the Xorg server which Xgl starts up automatically. Does the Xgl spit out anything interesting somewhere else?

I have a Savage8 chip (Via KM266 chipset) and OpenGL-apps crash here. glxinfo says Direct Rendering is enabled (this would be way cool, earlier I could only use OpenGL software emulation), but then it crashes. glxgears also segfaults. I'm using the 'savage' driver in my xorg.conf

root@dapperdrake:/var/log# glxinfo
name of display: :0.0
display: :0 screen: 0
direct rendering: Yes
server glx vendor string: SGI
server glx version string: 1.2
server glx extensions:
    GLX_ARB_multisample, GLX_EXT_visual_info, GLX_EXT_visual_rating,
    GLX_EXT_import_context, GLX_OML_swap_method, GLX_SGI_make_current_read,
    GLX_SGIS_multisample, GLX_SGIX_hyperpipe, GLX_SGIX_swap_barrier,
    GLX_SGIX_fbconfig
client glx vendor string: SGI
client glx version string: 1.4
client glx extensions:
    GLX_ARB_get_proc_address, GLX_ARB_multisample, GLX_EXT_import_context,
    GLX_EXT_visual_info, GLX_EXT_visual_rating, GLX_MESA_allocate_memory,
    GLX_MESA_swap_control, GLX_MESA_swap_frame_usage, GLX_OML_swap_method,
    GLX_OML_sync_control, GLX_SGI_make_current_read, GLX_SGI_swap_control,
    GLX_SGI_video_sync, GLX_SGIS_multisample, GLX_SGIX_fbconfig,
    GLX_SGIX_pbuffer, GLX_SGIX_visual_select_group,
    GLX_EXT_texture_from_pixmap
GLX version: 1.2
GLX extensions:
    GLX_ARB_get_proc_address, GLX_ARB_multisample, GLX_EXT_import_context,
    GLX_EXT_visual_info, GLX_EXT_visual_rating, GLX_OML_swap_method,
    GLX_SGIS_multisample, GLX_SGIX_fbconfig, GLX_SGIX_visual_select_group
OpenGL vendor string: S3 Graphics Inc.
OpenGL renderer string: Mesa DRI ProSavageDDR 20050829 AGP 1x
OpenGL version string: 1.2 Mesa 6.4.1
OpenGL extensions:
    GL_ARB_imaging, GL_ARB_multisample, GL_ARB_multitexture,
    GL_ARB_texture_compression, GL_ARB_texture_env_add,
    GL_ARB_texture_mirrored_repeat, GL_ARB_transpose_matrix,
    GL_ARB_window_pos, GL_EXT_abgr, GL_EXT_bgra, GL_EXT_blend_color,
    GL_EXT_blend_minmax, GL_EXT_blend_subtract, GL_EXT_clip_volume_hint,
    GL_EXT_compiled_vertex_array, GL_EXT_convolution, GL_EXT_copy_texture,
    GL_EXT_draw_range_elements, GL_EXT_histogram, GL_EXT_packed_pixels,
    GL_EXT_polygon_offset, GL_EXT_rescale_normal,
    GL_EXT_separate_specular_color, GL_EXT_stencil_wrap, GL_EXT_subtexture,
    GL_EXT_texture, GL_EXT_texture3D, GL_EXT_texture_edge_clamp,
    GL_EXT_texture_env_add, GL_EXT_texture_lod_bias, GL_EXT_texture_object,
    GL_EXT_vertex_array, GL_APPLE_packed_pixels, GL_IBM_rasterpos_clip,
    GL_IBM_texture_mirrored_repeat, GL_MESA_window_pos,
    GL_NV_light_max_exponent, GL_NV_texgen_reflection, GL_OES_read_format,
    GL_SGI_color_matrix, GL_SGI_color_table, GL_SGIS_texture_edge_clamp,
    GL_SGIS_texture_lod
glu version: 1.3
glu extensions:
    GLU_EXT_nurbs_tessellator, GLU_EXT_object_space_tess
Segmentation fault

OK, I have the same bug, but in my case the graphics card is an Nvidia 6200, glxgears and all opengl applications make xgl crash.

with a prosavage (compaq presario 701ea laptop) I can't start xgl, but the same happens with aiglx

It's the same here...
I'm using aiglx and my xServer crashes everytime I start any opengl-test or app or glxinfo... I've got a GeForce 6800 and I'm using the nvidia proprietary drivers...

There seem to be a number of different bugs here. There are 2 reports of Xgl crashing when running glxinfo, on Savage & nVidia hardware. Can either of you test with Gutsy's version of Xgl? I cannot reproduce this here.

Two other people with savage cards see glxinfo crashing, but not under Xgl. This isn't an Xgl bug, and seems to be a duplicate of bug #45307 which appears to be fixed.

Finally, there's Tom with the "nvidia + Composite + OpenGL crashes X" bug #130325

Yes, also in current Gutsy: Xgl crashes as soon as I run for instance glxinfo inside it. It leaves a useless core dump.

Hm. This may be due to the savage drivers not supporting pbuffer objects which Xgl uses to accelerate 3d. The cards the savage supports have fairly minimal 3D hardware, right?

You could test this by editing the /usr/share/xserver-xgl/Xgl-session file, and changing the 'XGL_OTHER_ACCEL_OPTS="-accel xv:pbuffer -accel glx:pbuffer"' to read 'XGL_OTHER_ACCEL_OPTS=""'. If this fixes it, we may need to add savage to the Xgl blacklist.

No, this does not fix it. I also tried running an Xgl "manually" (non fullscreen) inside a failsafe xterm session.

I also tried :fbo and :none as well as -noglx. It doesn't need glxinfo to crash, just filling up an xterm so it should scroll will crash the server.

I think savage can safely be blacklisted.

xserver-xgl (1:1.1.99.1~git20070727-0ubuntu3) gutsy; urgency=low

  * debian/Xgl-session
    + Add "savage" to driver blacklist (LP: #32588)
    + Find an unused DISPLAY to spawn Xgl on (LP: #140870)
  * debian/Xgl-lockfile-wrapper
    + Add (another) wrapper to clean up the lockfiles Xgl leaves behind on
      logout (LP: #139852)

 -- Christopher James Halse Rogers <email address hidden> Sun, 23 Sep 2007 22:23:15 +1000