This fixes the last false positive from libdrm for valgrind glxgears for me. ---
I'm slightly nervous about removing memsets in favor of VG_CLEAR, in case there are pad fields that might become important later. But the only one I could think of that this ought to be the case for was execbuf, and that explicitly initializes the pad fields. So Chris's patch series is: Reviewed-by: Eric Anholt <[email protected]> though it would be nice if we could teach upstream valgrind about our ioctl parameters instead of just overriding all over this code, so that other direct uses of our ioctls in X and Mesa didn't cause valgrind spew either. intel/intel_bufmgr_gem.c | 1 + 1 files changed, 1 insertions(+), 0 deletions(-) diff --git a/intel/intel_bufmgr_gem.c b/intel/intel_bufmgr_gem.c index 60d54d9..d84da9b 100644 --- a/intel/intel_bufmgr_gem.c +++ b/intel/intel_bufmgr_gem.c @@ -2354,6 +2354,7 @@ drm_intel_bufmgr_gem_init(int fd, int batch_size) (int)bufmgr_gem->gtt_size / 1024); } + VG_CLEAR(gp); gp.param = I915_PARAM_CHIPSET_ID; gp.value = &bufmgr_gem->pci_device; ret = drmIoctl(bufmgr_gem->fd, DRM_IOCTL_I915_GETPARAM, &gp); -- 1.7.9 _______________________________________________ Intel-gfx mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/intel-gfx
