CSS Validator can't parse .dfn-panel.activated.

[mikewest]

Bikeshed's definition panel CSS includes the rule .dfn-panel.activated. That throws the following error in the live pubrules checker:

[generator.unrecognize] Parse Error (https://w3c.github.io/webappsec-credential-management/WD.html, line 29 in '.dfn-panel.activated')

The CSS is legal, validator or not.

[mikewest]

Oh, I see. It's not choking on the rule itself. It's choking on calc(...).

I'd prefer not to add an exception for every WebAppSec spec. Think this is something y'all could either fix or globally exempt?

[deniak]

I'm not sure when calc() will be properly supported in the CSS validator so in the meantime, I suggest to use the exception system.

[mikewest]

This means that every bikeshed spec will need to talk to you about getting an exception before publication. That seems like a lot of overhead, both for you and for spec authors who will be confused by error messages that don't actually represent errors.

[deniak]

We can easily ignore specific types of error reported by the CSS validator (in this case generator.unrecognize) for all the specs. One drawback is it will ignore real errors as well.

[mikewest]

So far, the validator has prevented me from publishing both specs I've tried. Neither time has been due to a valid error (see #319 for the other one). In my personal set of ancedotes, this hasn't been a helpful check.

[deniak]

This issue will be fix if we resolve to skip the CSS validator until it's updated (see #322). @plehegar what do you think?

[mikewest]

Ping! I'm interested in seeing some resolution here; I have a number of Bikeshed-driven specs whose /TR version I'd like to refresh. :)

[deniak]

I'm being told the CSS validator will very soon be able to handle calc() properly.

[plehegar]

I suggest ignoring the CSS validator for the next 10 days. If it's still giving errors away after that, we'll need to start ignoring the CSS validator altogether.

[mikewest]

@plehegar: I'm fine with that. I'd simply like to be able to submit some updates to /TR without landing exemptions for each spec. :) @deniak, WDYT?

[deniak]

@mikewest : I talked to @plehegar and what I meant is let's wait 10 days to see if the validator gets patched. After that, if it still fails on calc(), we'll disable the CSS validator for all the specs.
In the meantime, I'm happy to publish your specs on Tuesdays or Thursdays.

[plehegar]

To give more background: I've been told that the calc problem would be fixed within 10 days in the css validator, so let's not try to make changes we would have to undo after in that timeframe. The truth is that we're catching more false positive errors in the css validation than real errors nowadays so the value of having the css validation step is clearly negative at the moment. If we can't change that within the 10 days, my proposal is to deactivate the css validation step entirely.

[ylafon]

Should be fixed

[tripu]

👍, @ylafon!

@mikewest, https://jigsaw.w3.org/css-validator/validator?uri=https%3A%2F%2Fw3c.github.io%2Fwebappsec-credential-management%2FWD.html&profile=css3&usermedium=all&warning=1&vextwarning=&lang=en is green now. Does that solve the issue?

[mikewest]

It does, thank you. :)

[mikewest]

Reopening: the CSS validator now fails on left: calc(-1 * (3.5rem - 26px)); (and on a number of SVG attributes that I think we agreed to simply ignore for the moment?): https://jigsaw.w3.org/css-validator/validator?uri=https%3A%2F%2Fw3c.github.io%2Fwebappsec-secure-contexts%2FWD.html&profile=css3&usermedium=all&warning=1&vextwarning=&lang=en

This is basically teaching me never to upgrade Bikeshed, which certainly shouldn't be the lesson. :)

@ylafon: Where should I be filing bugs against the validator?

[ylafon]

@mikewest https://github.com/w3c/css-validator/issues but I just fixed the calc() issue with the latest commit. Thx for the report!

[ylafon]

Note that SVG properties are not part of http://www.w3.org/TR/2015/NOTE-css-2015-20151013/ maybe it's a bug.

[deniak]

Closing this ticket as calc() is now correctly implemented in the CSS validator. The SVG properties are not defined by the CSS WG. We might need to add special exceptions to ignore the errors for these properties.

[tabatkins]

The SVG properties are defined by the SVGWG; there's no requirement that all CSS must be defined by the CSSWG. They're valid and well-defined properties, and need to be supported by the validator. (This is currently blocking publication of Media Queries 4, will block CSS Syntax as soon as I publish that again, and will continue to block anything that uses Bikeshed's railroad-diagram support.)