-
Notifications
You must be signed in to change notification settings - Fork 20
Make validation errors a test error. #258
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
@spall I will eventually want to get this change in. I noticed that some DXC issues showed up, and not all of them are problems in the compiler. I'll get the Vulkan API code clean before I ask to merge this. Failing tests will get marks an XFAIL with a bug opened against them. Is that reasonable? |
@s-perron, I was discussing related issues with @spall last night. We definitely want to get this in. Sarah spent time yesterday chasing a failure that had validation errors which didn't cause the test to fail outright, instead it was failing in verification. Filing a tracking issue on the relevant compiler(s) and XFAILing the tests seems like the right approach to me. |
c3bac03
to
cf9299d
Compare
The current use of the validation layer will emit a message, but it will not case a test to fail. By adding a call back that returns true for error messages and warnings, test that have a validation error or warning will fail. I've made other changes to keep tests that already pass passing. xfail test/Basic/TestPipeline.test. Modify the VkBufferUsageFlagBits for cbuffers. They are uniform buffers not uniform texel buffers.
FYI: I'm trying to track down why the Intel tests had validation error for f2aadab. I do not have an Intel GPU available right now, so I'll be updating to PR to see what works. |
I saw it go purple right as I was about to comment about the assertion failures. So, the assertion is due to support for handlefromimplicitbinding not being added yet to spirv. I can fix these tests so they don't fail by providing a register for the RWBuffer's. I believe the issue you filed in llvm is already being tracked in @hekota 's work. |
Sorry, I say a green checkmark from you above, and did not realize that was still the first review from you, and not the second. I was too quick to merge. |
The current use of the validation layer will emit a message, but it will
not cause tests to fail. By adding a call back that returns true for
error messages and warnings, test that have a validation error or
warning will fail.