On 03/21/2012 08:53 PM, Marc-André Lureau wrote:
Note: I can't reproduce the leak, even after dozen of resolution switch.
Always delete shared memory segment of primary surfaces when
destroying its canvas.
Ack. Tested.
Just notice that it is not only for the primary surface, but for every
surface.
---
gtk/channel-display.c | 6 +-----
1 files changed, 1 insertions(+), 5 deletions(-)
diff --git a/gtk/channel-display.c b/gtk/channel-display.c
index 152d6a7..8269705 100644
--- a/gtk/channel-display.c
+++ b/gtk/channel-display.c
@@ -705,6 +705,7 @@ static void destroy_canvas(display_surface *surface)
#ifdef HAVE_SYS_SHM_H
else {
shmdt(surface->data);
+ shmctl(surface->shmid, IPC_RMID, 0);
}
#endif
surface->shmid = -1;
@@ -815,11 +816,6 @@ static void display_handle_mode(SpiceChannel *channel,
SpiceMsgIn *in)
surface->size = surface->height * surface->stride;
surface->primary = true;
create_canvas(channel, surface);
-#ifdef HAVE_SYS_SHM_H
- if (surface->shmid != -1) {
- shmctl(surface->shmid, IPC_RMID, 0);
- }
-#endif
So for very old spice-server and for the old qxl driver, there wouldn't
have been a leak (since the primary surface was created with
SPICE_MSG_DISPLAY_MODE msg, which is no longer used).
}
/* coroutine context */
Regards,
Yonit.
_______________________________________________
Spice-devel mailing list
Spice-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/spice-devel
