Summary: | [regression, bisected] arb_shader_image_load_store.compiler.declaration-format-qualifier-duplicate | ||
---|---|---|---|
Product: | piglit | Reporter: | Mark Janes <mark.a.janes> |
Component: | tests | Assignee: | Timothy Arceri <t_arceri> |
Status: | CLOSED FIXED | QA Contact: | Piglit Mailing List <piglit> |
Severity: | normal | ||
Priority: | medium | CC: | chrisf, mark.a.janes |
Version: | unspecified | ||
Hardware: | x86-64 (AMD64) | ||
OS: | Linux (All) | ||
Whiteboard: | |||
i915 platform: | i915 features: | ||
Attachments: | output from arb_shader_image_size |
Description
Mark Janes
2016-01-19 22:42:57 UTC
It may be that the test is wrong, and Timothy's patch to Mesa is correct. It may be that the test is incorrect and Timothy's patch is correct. Fixed by: commit 0a6a05c8eaeda570891fdece2d86e8e0b0e4d56f glsl: add missing explicit_image_format flag to has_layout() Fixes piglit regression after fixes to duplicate layout rules. Previously catching multiple layouts was relying on the code meant to catch duplicates within a single layout(...), this change triggers the rules for multiple layouts. On hsw/ivb, the following tests now fail: piglit.spec.arb_shader_image_load_store.max-images piglit.spec.arb_shader_image_load_store.indexing piglit.spec.arb_shader_image_load_store.semantics piglit.spec.arb_shader_image_load_store.shader-mem-barrier piglit.spec.arb_shader_image_size.builtin skl/bdw fails: piglit.spec.arb_shader_image_size.builtin This is with piglit=e792463, which fixed many other failures associated with 0a6a05. stderr for shader_image_load_store tests show: Failed to compile compute shader: 0:38(1): error: storage qualifiers must come after precise, invariant, interpolation, layout and auxiliary storage qualifiers for shader_image_size stderr is: Failed to compile fragment shader: 0:22(26): error: syntax error, unexpected NEW_IDENTIFIER, expecting '{' I'm changing the product of this bug to piglit, as per Tim's comments. I've tested on my IVB and there is indeed more issues with the tests. They seem to take different paths in the tests on different hardware, so I didn't notice these when fixing on my BDW. Should have a fix shortly. HSW/IVB issues resolve by piglit: commit 9bd8454a24304eabadcc4a5f6766158bbb8eb508 arb_shader_image_load_store: fix location of storage qualifier In GLSL 1.50 layout qualifiers must come before the storage qualifier. A recent fix in Mesa exposed this issue. However I can't reproduce the last problem on either skl or bdw, any chance you could copy the shader that's failing into the bug report? Created attachment 121192 [details]
output from arb_shader_image_size
Looks like there's some left-over IMAGE_T usage... pushed a fix. commit 1e7ca7fc900fbf59a37e84b46c2a2fb69686dc1f Author: Ilia Mirkin <imirkin@alum.mit.edu> Date: Thu Jan 21 13:51:21 2016 -0500 arb_shader_image_size: fix last use of IMAGE_T Trivial |
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.