On Sat, Dec 15, 2012 at 3:48 PM, Andreas Pokorny <andreas.poko...@gmail.com> wrote: > Hello, > I have been chasing a leak with OpenVG mesa-8.0.5. > > Massif gave me the following stack: > ->09.66% (3,471,568B) 0xB706B80: util_hash_table_set (u_hash_table.c:163) > | ->09.66% (3,471,568B) 0xB4E106F: create_handle (handle.c:83) > | ->09.66% (3,471,568B) 0xB527B7D: vg_init_object (vg_context.c:186) > | ->06.56% (2,358,352B) 0xB4E3599: paint_create (paint.c:201) > | | ->06.56% (2,358,320B) 0xB4DAF4F: vegaCreatePaint (api_paint.c:37) > > Within the first snapshots this stack only occupies a small portion of > the memory below the thresholds. This portion grows to 9% after 100 > seconds of slow mode rendering.The application that creates the > paints, frees them right after use.. I only discovered the issue with > paints, since all other handles (paths, images..) are kept alive and > are reused for the next frame. > > With the patch applied the leak is gone, and the memory usage stays > constant. The patch is against MesaLib-9.0.1 since I could not clone > from annongit.freedesktop.org. > > Andreas Pokorny (1): > vega: fix for object handle leak > > src/gallium/state_trackers/vega/mask.c | 1 + > src/gallium/state_trackers/vega/paint.c | 4 +++- > src/gallium/state_trackers/vega/path.c | 2 ++ > src/gallium/state_trackers/vega/text.c | 2 ++ > 4 files changed, 8 insertions(+), 1 deletion(-)
Looks OK to me. Do you need someone to push this for you? -Brian _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
