|
From: Timothy B. T. <tte...@em...> - 2006-06-17 01:29:14
|
I suspect this will be more of an "interesting" report than one I actually expect anyone to do anything about. I was recently trying to debug an OpenGL program via valgrind 2.4.0. I wanted to attach a debugger at the place where the errors occurred, but the prompts at the various unsuprressed errors in my X11 and OpenGL libraries would cause the display to lock up (I could still ssh into the machine and kill valgrind to recover, however). So I upgraded to valgrind 3.1.1 to get the --gen-supressions=all switch to eliminate these errors, and now every time the program exited, valgrind would spin using 100% of the CPU (at least judging by the fan speed), the display would stop updating, and the network would stop responding (I could not even ssh in to kill the process). This would occur just as it began searching for not-freed blocks, e.g.: ==28722== For counts of detected errors, rerun with: -v ==28722== searching for pointers to 20127 not-freed blocks. At least those were the last lines output to the display. Even fewer got written to a file when I redirected stderr. After a number of hard power-off's I managed to get the supression contexts so I could revert to 2.4.0 and solve my original problem. For now I'm happy to continue using 2.4.0, but this makes newer versions of valgrind essentially useless for me. The kernel I am running is: Linux mado 2.6.14-gentoo-r5 #1 PREEMPT Wed Jan 18 19:06:52 EST 2006 i686 Intel(R) Pentium(R) M processor 1600MHz GenuineIntel GNU/Linux The X server is X.org's: X Window System Version 6.8.2 The video drivers are ATI's binary drivers: compiled for 6.8.0, module version = 8.23.7 (this is the real reason I'm not expecting much actual help) The thing that makes it odd, and the only reason I'm bothering to report it at all was that the problem occurred with 100% reproducibility with valgrind 3.1.1, and has never occurred with 2.4.0. |