On 02/27/2012 04:40 PM, Peter Hutterer wrote:

The test case was:
   valgrind --leak-check=full --show-reachable=yes Xvfb
forcing two server regenerations by running xterm against it 3 times.

Before:
==31654== LEAK SUMMARY:
==31654==    definitely lost: 3,720 bytes in 11 blocks
==31654==    indirectly lost: 287,964 bytes in 2,273 blocks
==31654==      possibly lost: 0 bytes in 0 blocks
==31654==    still reachable: 136,104 bytes in 1,733 blocks
==31654==         suppressed: 0 bytes in 0 blocks

After:
==13190== LEAK SUMMARY:
==13190==    definitely lost: 960 bytes in 8 blocks
==13190==    indirectly lost: 123,552 bytes in 572 blocks
==13190==      possibly lost: 0 bytes in 0 blocks
==13190==    still reachable: 27,724 bytes in 77 blocks
==13190==         suppressed: 0 bytes in 0 blocks

The majority of the indirectly lost are in glx, but afaict they are false
positives:
==13190== 61,344 bytes in 284 blocks are indirectly lost in loss record 52
of 54
==13190==    at 0x4A074CD: malloc (vg_replace_malloc.c:236)
==13190==    by 0x4E1478: createModeFromConfig (glxdricommon.c:131)
==13190==    by 0x4E16B1: glxConvertConfigs (glxdricommon.c:187)
==13190==    by 0x4E12D7: __glXDRIscreenProbe (glxdriswrast.c:472)
==13190==    by 0x4E0220: GlxExtensionInit (glxext.c:328)
==13190==    by 0x41CA14: InitExtensions (miinitext.c:471)
==13190==    by 0x5ACBC8: main (main.c:208)

I understand enough to review the following without diving in further than I'm able right now: 1, 2, 4, 7. For those:

Reviewed-by: Chase Douglas <[email protected]>
_______________________________________________
[email protected]: X.Org development
Archives: http://lists.x.org/archives/xorg-devel
Info: http://lists.x.org/mailman/listinfo/xorg-devel

Reply via email to