If I pass setWindowRect as an alwaysMatch capability, what should behaviour be?
Currently, under "For each name and value corresponding to capability’s own properties:", this falls through out of the "Otherwise" clause because:
"If name is the name of an additional WebDriver capability" doesn't apply because an additional WebDriver capability is one defined in another spec, and setWindowRect is defined in this spec
"Otherwise, if name is the key of an extension capability" also obviously doesn't apply
This effectively means that we end up not doing any matching and we also don't fail as a result, and the returned value may well be different to the alwaysMatch value.
From a protocol point-of-view, it is exceptionally surprising that we end up with the returned value being different to the alwaysMatch value.
(Note I'm not super interested in setWindowRect specifically here—and indeed this is arguably a bug—but we do want to allow for other output-only capabilities, and we should define in general what should happen when they are passed in as a capability.)
whimboo
commented
4 months ago
Split from discussion on https://github.com/w3c/webdriver-bidi/issues/446 (c.f. https://github.com/w3c/webdriver/pull/1790#issuecomment-1935920012).
Not everything in https://w3c.github.io/webdriver/#capabilities has defined matching behaviour under https://w3c.github.io/webdriver/#dfn-matching-capabilities.
If I pass
setWindowRectas analwaysMatchcapability, what should behaviour be?Currently, under "For each name and value corresponding to capability’s own properties:", this falls through out of the "Otherwise" clause because:
"If name is the name of an additional WebDriver capability" doesn't apply because an additional WebDriver capability is one defined in another spec, and setWindowRect is defined in this spec
"Otherwise, if name is the key of an extension capability" also obviously doesn't apply
This effectively means that we end up not doing any matching and we also don't fail as a result, and the returned value may well be different to the
alwaysMatchvalue.From a protocol point-of-view, it is exceptionally surprising that we end up with the returned value being different to the
alwaysMatchvalue.(Note I'm not super interested in
setWindowRectspecifically here—and indeed this is arguably a bug—but we do want to allow for other output-only capabilities, and we should define in general what should happen when they are passed in as a capability.)