gpuweb/cts

`webgpu:shader,validation,parse,diagnostic:duplicate_attribute_same_location:*` expects failure when `s1` and `s2` are the same, but should succeed

Closed this issue · 2 comments

ErichDonGubler commented

According to the WGSL spec.:

Two diagnostic filters conflict when:

  • (AR1 = AR2), and

  • (TR1 = TR2), and

  • (NS1 ≠ NS2).

...where NS1 and NS2 correspond to the s1 and s2 matrix values in webgpu:shader,validation,parse,diagnostic:duplicate_attribute_same_location:*. This means that when s1 === s2 in the duplicate_attribute_same_location test, we should expect success, in addition to the cases already expected where same_rule === false.

alan-baker commented

The attribute has an additional rule:

More than one diagnostic attribute may be specified on a syntactic form, but they must specify different triggering rules.

So same severity is allowed for the global filter, but disallowed for attributes.

ErichDonGubler commented

Oof! My mistake, it seems. 😩 Thanks for the patient response.