CSS3 calc: unprefix implementation

• Component: CSS
• Status: RESOLVED FIXED

[Mike Lawther]

IE9 and Mozilla (https://bugzilla.mozilla.org/show_bug.cgi?id=771678) have unprefixed their calc implementations.

Normally we'd wait for the spec (http://www.w3.org/TR/css3-values/) to to to Candidate Rec (CR). It's currently at Last Call, and comments were due 29/03/2012.  

With 2 unprefixed implementations, it's enough to check that ours is compatible with theirs.

IE calc tests: http://samples.msdn.microsoft.com/ietestcenter/#css3valuesandunits
Moz calc tests: http://mxr.mozilla.org/mozilla-central/source/layout/reftests/css-calc/

[Mike Lawther]

I've tested -webkit-calc() against both the IE and Mozilla test suites.

IE : PASS
Moz : PASS everything except:
  - text-indent-intrinsic-1.html : this is a problem in WebKit's text-indent intrinsic
  - width-table-auto-1.html, width-table-fixed-1.html : both table related, and per spec it is acceptable to ignore calc() in tables.

The spec (http://www.w3.org/TR/css3-values/) is now at Candidate Rec.

With the spec at CR, two shipping unprefixed implementations, and broad compat between ours and theirs, I reckon it's OK to go ahead with the unprefixing.

[Binyamin]

Just awaiting for unprefix.

[Mike Lawther]

Created attachment 183701 [details]
Patch

[Mike Lawther]

Testing-wise, I've replaced all -webkit-calc() in the tests with calc(), as the unprefixed version is what we'll support from now on. I've kept a copy of the most basic test using -webkit-calc() to ensure our support for it doesn't completely break.

This seemed a sane approach - keeping an unprefixed copy of the entire test directory seemed like overkill.

[Ryosuke Niwa]

(In reply to comment #1)
> Moz : PASS everything except:
>   - text-indent-intrinsic-1.html : this is a problem in WebKit's text-indent intrinsic
>   - width-table-auto-1.html, width-table-fixed-1.html : both table related, and per spec it is acceptable to ignore calc() in tables.

Please file bugs to track these failures.

[Ryosuke Niwa]

Did you try the Mozilla tests after fixes are landed in https://bugzilla.mozilla.org/show_bug.cgi?id=826582 ?

[Mike Lawther]

Text indent bug filed as https://bugs.webkit.org/show_bug.cgi?id=107416. 

wrt filing bugs about the table differences - both our impls are to spec. All the bug would say is 'this behaviour is different'?

I did not run the tests after Moz upstreamed them to the W3C. I can do this and report back.

[Ojan Vafai]

Comment on attachment 183701 [details]
Patch

<3

[Ryosuke Niwa]

(In reply to comment #7)
> wrt filing bugs about the table differences - both our impls are to spec. All the bug would say is 'this behaviour is different'?

That would be great. It'll helps us document the issue at least. If anything, we should check if there's any compatibility concerns to pick one behavior over another.

[WebKit Review Bot]

Comment on attachment 183701 [details]
Patch

Clearing flags on attachment: 183701

Committed r140300: <http://trac.webkit.org/changeset/140300>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.

[Mike Lawther]

As requested, I tested Mozilla's updated tests from http://mxr.mozilla.org/mozilla-central/source/layout/reftests/w3c-css/submitted/values3/, and the results were the same as I reported in comment #1.

Aside - now that we are both unprefixed, I was able to do this testing without modifying any test source code - yay for compat!

[Ryosuke Niwa]

(In reply to comment #12)
> As requested, I tested Mozilla's updated tests from http://mxr.mozilla.org/mozilla-central/source/layout/reftests/w3c-css/submitted/values3/, and the results were the same as I reported in comment #1.
> 
> Aside - now that we are both unprefixed, I was able to do this testing without modifying any test source code - yay for compat!

Thanks for the follow up!

[Mike Lawther]

Table differences filed as https://bugs.webkit.org/show_bug.cgi?id=107482