Fix fragment input deductions in max_variables_count,input

[ErichDonGubler]

Issue: #4538

See also #4539.

---

Requirements for PR author:

• All missing test coverage is tracked with "TODO" or .unimplemented().
• New helpers are /** documented */ and new helper files are found in helper_index.txt.
• Test behaves as expected in a WebGPU implementation. (If not passing, explain above.)
• Test have be tested with compatibility mode validation enabled and behave as expected. (If not passing, explain above.)

Requirements for reviewer sign-off:

• Tests are properly located in the test tree.
• Test descriptions allow a reader to "read only the test plans and evaluate coverage completeness", and accurately reflect the test code.
• Tests provide complete coverage (including validation control cases). Missing coverage MUST be covered by TODOs.
• Helpers and types promote readability and maintainability.

When landing this PR, be sure to make any necessary issue status updates.

[kainino0x]

nit: Test -2 and -3 because those are the cases on either side of the boundary. (Helps catch off-by-one errors or if an implementation tries to merge two of the variables but not the third.)

Suggested change

	{ numVariablesDelta: 0, useExtraBuiltinInputs: true },
	{ numVariablesDelta: -1, useExtraBuiltinInputs: true },
	{ numVariablesDelta: -3, useExtraBuiltinInputs: true },
	{ numVariablesDelta: -2, useExtraBuiltinInputs: true },
	{ numVariablesDelta: -3, useExtraBuiltinInputs: true },

[ErichDonGubler]

Added the -2 case with the latest force-push to 7c3fdf7. I want to clarify, though: your suggestion block would strip the -1 and 0 cases, and I'm guessing that's not intentional. Is that right?

[kainino0x]

IIUC (please let me know if I'm misunderstanding the test):

It used to test -1 and 0 because those were on either side of the boundary (-1 would give space for the builtins, 0 would not). Now the boundary is between -3 (max valid) and -2 (min invalid). I think it should be pretty safe to assume that if the implementation gets -2 correct, then it will get -1 and 0 correct as well.

To be fair this is not clear because of the structure of the test since it computes success in the test body instead of passing it along with each case. It would probably be a good idea to do a little refactor here and make success a parameter (_success: true / _success: false) so we can actually tell that the test is testing the boundaries it intends to.

[ErichDonGubler]

Makes sense. Changed cases (but not the _success refactor) as suggested with latest force-push to f9fd2cc.

[ErichDonGubler]

Ah, but wait: If this is compat, then -1 is still an interesting boundary case to test, isn't it? It seems like that's still interesting to keep. That also means that we can't naively add _success to the case parameters, because whether it should succeed or not would actually depend on how many built-ins are being added, which is variable.

[ErichDonGubler]

Force-pushed to e0157af to include -1.

[kainino0x]

very good point... I should have actually looked at the code again before suggesting things.

I pulled it locally and experimented with another fix idea. Couldn't push to your branch so I made a new PR here that includes your commits and mine. PTAL: #4555

[github-actions]

Results for build job (at e0157af):

-webgpu:api,validation,render_pipeline,inter_stage:max_variables_count,input:* - 8 cases, 8 subcases (~1/case)
+webgpu:api,validation,render_pipeline,inter_stage:max_variables_count,input:* - 10 cases, 10 subcases (~1/case)
-TOTAL: 284495 cases, 2341917 subcases
+TOTAL: 284497 cases, 2341919 subcases

i'm editing the PR

[greggman]

AFAICT numVariablesDelta is how many to add to maxInterStageShaderVariables. I don't think you can really decide where how many to subtract and statically determine success without changing the rest of the test. If the test is in compatibility mode then only 1 extra builtin will be added. If it's in core, 3 will be added.

[kainino0x]

There is a thread about this above #4546 (comment)
and that's how we ended up with #4555. Actually I should close this PR (at least for now) in favor of that one.