css/selectors/attribute-selectors/attribute-case/syntax.html test has inconsistent names of subtests

[foolip]

https://wpt.fyi/results/css/selectors/attribute-selectors/attribute-case/syntax.html?sha=71fe0a6c3d

In Edge, in addition to the failures, the subtests names are different, some are missing and there's an extra "[foo='bar' i" test at the end which I cannot find in the source.

This seems likely to be a test bug.

@jugglinmike is this a sort of anomaly that shows up in any report we have?

[gsnedders]

I don't really understand how this could be a test bug? It's a pretty simple sequence of sync tests…

Almost all the missing tests are ones with null bytes in them, so I presume something somewhere is failing with that?

[foolip]

Hmm, possibly an infra problem then. If null bytes are involved, then somewhere along the line a string is stored as a C string. Maybe that's an Edge bug? Or Sauce? Would need to debug.

[jugglinmike]

I'm on the case!

[jugglinmike]

This is the result of a deficiency in Edge/Edgedriver: string literals transmitted via the "Execute Script" and "Execute Async Scripts" commands are truncated at the first null byte.

I've submitted a WebDriver conformance test to alert the Edge maintainers, and I've suggested a more immediate solution for WPT via gh-14245.

[jugglinmike]

@foolip We never did come to a conclusion for unprintable characters in test names, but now that Edge has a new WebDriver implementation which tolerates that case (as evidenced by recent test
results), the issue is less relevant. I'm going to un-assign myself and recommend my fix, unpopular though it may be--gh-14245.

[foolip]

Alright, I'll just go ahead and close this because without something to surface the problem it won't ever be a priority for anyone.