Re: A Bug's Life - bug 2920 now fixed? [reposted with corrections]

[Please disregard my earlier post regarding bug 2920 – I got my bug reference numbers mixed up. Too much multitasking. My earlier comments were in reference to a different bug.]

Returning to bug 2920. Things have certainly changed, for the better.

The test cases now are correctly showing up as invalid, and the validator appears to be branching according to CSS version.

1) Fixed? Under selected grammar = CSS2.1, test cases 1-7 all seem to be handled correctly now – all showing as “invalid”.

2) Error message? One small point is that a rather unexpected “no stylesheet” error message is generated if there is nothing else at all after the @charset? Is this as intended?

3) Fault still outstanding? But under selected grammar = CSS3 all the test cases still appear to be _incorrectly_ showing as “valid”, (as if the changes that were made for the CSS2.1 case were not also applied in the CSS3 branch). Needs double-checking, as I only dipped into this briefly.

4) Fixed? Under selected grammar = CSS2 (ie 2.0) the validator’s reported behaviour now I think correctly reflects the CSS2.0 grammar in respect of each of the test cases, that is the validator passes each as valid.

5) Suggestion: However, having said that, reporting these “lax” forms as “valid”  in the case where they are only valid under CSS2.0 is correct but not helpful, and I suggest that ideally a warning or (‘informational’) would be in order. “Formally ‘correct’ but will/may not work” is the idea. Yes, I understand that if the user specifically selected CSS2.0 then that’s what they’re getting, but a general forward-backwards compatibility warning/informational category for output messages would be a pretty useful thing to have.

The (new?) “no special profile” idea, is that like the scheme I suggested in comment following on from the original bug report? UI - Is there a link anywhere to explain to the user precisely what “no special profile means”?

Anyway, could someone take a look another look at the whole set of test cases, as a sanity check, whenever they get a chance?

Cecil Ward.

Received on Thursday, 19 July 2007 19:29:04 UTC