[Piglit] [PATCH 2/5] read-front: modify the test to reproduce an MSAA front buffer bug in st/dri

[Paul Berry]

On 1 June 2013 05:38, Marek Olšák <maraeo at gmail.com> wrote:

> ---
>  tests/all.tests            |    1 +
>  tests/general/read-front.c |   15 ++++++++++++++-
>  2 files changed, 15 insertions(+), 1 deletion(-)
>
> diff --git a/tests/all.tests b/tests/all.tests
> index 417ca08..058253d 100644
> --- a/tests/all.tests
> +++ b/tests/all.tests
> @@ -526,6 +526,7 @@ add_concurrent_test(gl11, 'polygon-mode-offset')
>  add_concurrent_test(gl11, 'push-pop-texture-state')
>  add_concurrent_test(gl11, 'quad-invariance')
>  add_plain_test(gl11, 'read-front')
> +add_plain_test(gl11, 'read-front clear-front-first')
>  add_concurrent_test(gl11, 'readpix-z')
>  add_plain_test(gl11, 'roundmode-getintegerv')
>  add_plain_test(gl11, 'roundmode-pixelstore')
> diff --git a/tests/general/read-front.c b/tests/general/read-front.c
> index e6833dc..bce00c6 100644
> --- a/tests/general/read-front.c
> +++ b/tests/general/read-front.c
> @@ -55,7 +55,6 @@ piglit_display(void)
>
>         piglit_ortho_projection(piglit_width, piglit_height, GL_FALSE);
>
> -       glClearColor(0.0, 0.0, 1.0, 0.0);
>         glClear(GL_COLOR_BUFFER_BIT);
>         glColor4fv(green);
>         piglit_draw_rect(0, piglit_height / 2, piglit_width,
> piglit_height);
> @@ -78,4 +77,18 @@ piglit_display(void)
>  void
>  piglit_init(int argc, char **argv)
>  {
> +       int i;
> +
> +       glClearColor(0.0, 0.0, 1.0, 0.0);
> +
> +       for (i = 1; i < argc; i++) {
> +               if (strcmp(argv[i], "clear-front-first") == 0) {
> +                       /* This should allocate the front buffer in the
> driver
> +                        * if it hasn't been allocated already.
> +                        */
> +                       glDrawBuffer(GL_FRONT);
> +                       glClear(GL_COLOR_BUFFER_BIT);
> +                       glDrawBuffer(GL_BACK);
>

In general accessing the front buffer isn't safe to do until after the
window has been exposed, and the window isn't guaranteed to have been
exposed by the time piglit_init() is called.  I think this clear call
should be moved into piglit_display().  Would that still reproduce your bug?

Note that at the moment, piglit_display() also is called before the window
is exposed, but I have a patch out on the list to fix that: "util/x11: Wait
for expose event in manual mode or with front buffer tests." (
http://lists.freedesktop.org/archives/piglit/2013-May/005880.html)


> +               }
> +       }
>  }
> --
> 1.7.10.4
>
> _______________________________________________
> Piglit mailing list
> Piglit at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/piglit
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/piglit/attachments/20130601/5dd58c1d/attachment-0001.html>