DOMMatrix2DInit validate-fixup test need some improvement

[fserb]

Using canvas set/getTransform as a proxy for the test has some weird consequences. Namely, it seems consesus on Canvas side that we should not throw when getting an invalid input on those cases. Which means the first 8 tests on css/geometry-1/DOMMatrix2DInit-validate-fixup are wrong. Even thought they make sense looking at Geometry spec only.

A similar thing happens to the "infinity" or the "-0" input tests. Canvas has no guarantee to maintain those. So they can't be tested like this.

I could remove them, but it could be worth to discuss what to do long term. On principle, I'm ok with just testing what Canvas can see.

[fserb]

@zcorpan ?

[zcorpan]

See
whatwg/html#2975
#7355

[zcorpan]

-0 maybe ought to be clarified if it is roundtripped by getTransform, can you file an issue for html please? (edit: I filed whatwg/html#3060.)

[zcorpan]

Closing this since #7473 is merged. Please reopen or comment if there's still something to do here.