Chris Rankin wrote: > Hi, > > I think there's a problem with Mesa git. This was with a clean rebuild: > > *** glibc detected *** glxgears: double free or corruption (!prev): > 0x08444bc0 *** > ======= Backtrace: ========= > /lib/libc.so.6[0x42718ac1] > /lib/libc.so.6(cfree+0x90)[0x4271c0f0] > /usr/X11R6/lib/modules/dri/r300_dri.so(_mesa_free+0x1d)[0xb7d86f4b] > /usr/X11R6/lib/modules/dri/r300_dri.so(_mesa_delete_program+0x106)[0xb7df676f] > /usr/X11R6/lib/modules/dri/r300_dri.so[0xb7d46fa6] > /usr/X11R6/lib/modules/dri/r300_dri.so(_mesa_reference_program+0x74)[0xb7df619f] > /usr/X11R6/lib/modules/dri/r300_dri.so(_mesa_free_context_data+0xbd)[0xb7d5787e] > /usr/X11R6/lib/modules/dri/r300_dri.so(_mesa_destroy_context+0x26)[0xb7d57a13] > /usr/X11R6/lib/modules/dri/r300_dri.so(radeonCleanupContext+0x24)[0xb7d2921e]
I added reference counting of vertex/fragment programs last night. There was some feeble attempt at it before but now it's done properly. Though, I guess there's probably a bug or two to solve. My guess here is we're double-freeing a program. Could you re-run with valgrind to get an idea of where that might be happening? Also, if you could use gdb to print the fields of the program object being deleted, that would be helpful. -Brian ------------------------------------------------------------------------- This SF.net email is sponsored by the 2008 JavaOne(SM) Conference Don't miss this year's exciting event. There's still time to save $100. Use priority code J8TL2D2. http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone _______________________________________________ Mesa3d-dev mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/mesa3d-dev
