<div dir="auto">Yes, as I mentioned this makes some tests assert.<div dir="auto"><br></div><div dir="auto">They were passing before, but it was through luck since the actual images were never accessed.</div></div><div class="gmail_extra"><br><div class="gmail_quote">On Mar 1, 2018 04:04, "Timothy Arceri" <<a href="mailto:tarceri@itsqueeze.com">tarceri@itsqueeze.com</a>> wrote:<br type="attribution"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">This causes the CTS tests to assert on radeonsi where they previously passed. If that expected?<br>
<br>
On 27/02/18 16:19, Ilia Mirkin wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Ideally the st_finalize_texture call would take care of that, but it<br>
doesn't seem to with KHR-GL45.shader_image_size.adv<wbr>anced-nonMS-*. This<br>
assertions makes sure that no such values are passed to the driver.<br>
<br>
Signed-off-by: Ilia Mirkin <<a href="mailto:imirkin@alum.mit.edu" target="_blank">imirkin@alum.mit.edu</a>><br>
---<br>
<br>
This will trigger asserts in CTS, but I think that's better than feeding<br>
broken values to driver backends.<br>
<br>
src/mesa/state_tracker/st_atom<wbr>_image.c | 1 +<br>
1 file changed, 1 insertion(+)<br>
<br>
diff --git a/src/mesa/state_tracker/st_at<wbr>om_image.c b/src/mesa/state_tracker/st_at<wbr>om_image.c<br>
index 1c4980173f4..421c926cf04 100644<br>
--- a/src/mesa/state_tracker/st_at<wbr>om_image.c<br>
+++ b/src/mesa/state_tracker/st_at<wbr>om_image.c<br>
@@ -97,6 +97,7 @@ st_convert_image(const struct st_context *st, const struct gl_image_unit *u,<br>
img->resource = stObj->pt;<br>
img->u.tex.level = u->Level + stObj->base.MinLevel;<br>
+ assert(img->u.tex.level <= img->resource->last_level);<br>
if (stObj->pt->target == PIPE_TEXTURE_3D) {<br>
if (u->Layered) {<br>
img->u.tex.first_layer = 0;<br>
<br>
</blockquote>
</blockquote></div></div>