Hi Brian,
On 06/21/2017 04:20 PM, Brian Paul wrote:
On 06/21/2017 06:24 AM, Thomas Hellstrom wrote:
Window systems (read dri3) that allocate a fake frontbuffer on demand
will
effectively read out from the real front before the first swapbuffers
after
fake front creation, and the real front buffer content is subject to
errors
caused by, among other things, window reparenting. So increase the
likelyhood
of not reading out from the real front by advertizing early that we will
access the front buffer. After all, real front consistency is not the
objective
of this test.
Fixes sporadic failures due to window reparenting on dri3/vmwgfx.
Cc: Brian Paul <bri...@vmware.com>
Signed-off-by: Thomas Hellstrom <thellst...@vmware.com>
---
tests/spec/gl-1.0/swapbuffers-behavior.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/tests/spec/gl-1.0/swapbuffers-behavior.c
b/tests/spec/gl-1.0/swapbuffers-behavior.c
index d914020..15a07ac 100644
--- a/tests/spec/gl-1.0/swapbuffers-behavior.c
+++ b/tests/spec/gl-1.0/swapbuffers-behavior.c
@@ -62,6 +62,7 @@ piglit_display(void)
/* Clear back buffer to green */
glDrawBuffer(GL_BACK);
+ glReadBuffer(GL_FRONT);
The comment no longer agrees with the code.
I'll fix that.
glClearColor(green[0], green[1], green[2], green[3]);
glClear(GL_COLOR_BUFFER_BIT);
@@ -69,7 +70,6 @@ piglit_display(void)
piglit_swap_buffers();
/* Front buffer sanity-check */
- glReadBuffer(GL_FRONT);
Same here.
Actually that comment still makes sense since it's followed by a
frontbuffer piglit_probe
if (!piglit_probe_rect_rgb_silent(0, 0, piglit_width,
piglit_height,
green)) {
printf("SwapBuffers apparently failed!\n");
These changes seem to defeat the purpose of the test.
Would a different approach work? How about after the initial
glViewport call we simply clear the front buffer? That should trigger
creation of the fake front buffer, right?
Yes, as long as we access the front buffer before SwapBuffers we should
be safe. But I don't quite understand how creating the fake front buffer
using glReadBuffer() would differ form creating it by a write access
with respect to the test functionality? All this patch is doing is
calling glReadBuffer() a little earlier than before?
/Thomas
-Brian
_______________________________________________
Piglit mailing list
Piglit@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/piglit