[Mesa-dev] Recent Mesa changes broke VTK transparency tests

Submitted by Brad King on April 28, 2014, 2:25 p.m.

Details

Message ID 535E64F3.3030600@kitware.com
State New
Headers show

Not browsing as part of any series.

Commit Message

Brad King April 28, 2014, 2:25 p.m.
Hi Folks,

I run VTK (vtk.org) tests against a nightly build of Mesa
configured without DRI as:

 ./autogen.sh --prefix=$prefix --disable-dri --disable-egl \
    --disable-shared-glapi --enable-xlib-glx --enable-osmesa \
    --without-gallium-drivers

Since commit 234db609544ee521458ce8b648e70cffe2fda6f9, the tests
that use transparency (depth peeling, IIRC) have been failing.
I found an existing open issue reporting another regression from
the same commit:

 https://bugs.freedesktop.org/show_bug.cgi?id=77553

Meanwhile, since commit d1b6f6711095cd94e80372e7488aa0189e328d88
the image looks a bit better but still incorrect.  There is also
an open issue for another regression from that commit:

 https://bugs.freedesktop.org/show_bug.cgi?id=77705

The above two issues report problems with DRI on certain platforms,
but I'd like to report that something is also wrong without DRI.
See attached images for the expected output and the failure after
each of the above two commits.  The hack below avoids the failures,
demonstrating they are due to the new code path.

Thanks,
-Brad

Patch hide | download patch | download mbox

diff --git a/src/mesa/drivers/common/meta.c b/src/mesa/drivers/common/meta.c
index ac27abb..5f734a0 100644
--- a/src/mesa/drivers/common/meta.c
+++ b/src/mesa/drivers/common/meta.c
@@ -2826,7 +2826,7 @@  _mesa_meta_CopyTexSubImage(struct gl_context *ctx, GLuint dims,
    GLint bpp;
    void *buf;

-   if (copytexsubimage_using_blit_framebuffer(ctx, dims,
+   if (0 && copytexsubimage_using_blit_framebuffer(ctx, dims,
                                               texImage,
                                               xoffset, yoffset, zoffset,
                                               rb,