W3C

- DRAFT -

CSS WG telcon

01 Dec 2010

See also: IRC log

Attendees

Present
Peter Linss, Tal, David Baron, Koji Ishi, Brad Kemper, Simon Fraser, Fantasai, Bert Bos, Chris Lilley, Arron Eicholz, Tab Atkins, Steve Zilles, HÃ¥kon Wium Lie, John Jansen
Regrets
Chair
Peter
Scribe
Bert

Contents


Agenda

PL: Anything to add?

CSS 2.1 LC

PL: We are past deadline for LC for CSS 2.1. But are there changes still expected?

CL: I think we can publish. No major changes.

<ChrisL> We are better to publish now and dealwith any changes as lc comments

AE: Ditto.

RESOLUTION: publish CSS 2.1 as last call

PL: Are all edits in the text?

BB: Yes.

Tests

TA: How do we note the failures?

DB: I didn't comment on some tests, because I think the test is correct.

PL: Other implementers?

CL: I did some of the Prince tests.

<dbaron> That said, I think it's a lot easier to prove a test incorrect than to prove a test correct.

CL: But some problems with building the suite.
... Worked with some old tests.
... So need to check against new ones.

<dbaron> (and the "some tests" I mentioned are those that are tests for features Gecko implements)

<plinss_> http://test.csswg.org/suites/css2.1/20101027.zip

PL: RC3 is in harness. SVN head version largely the same.

JJ: Yes, same as 1027.

PL: What is in RC3 is largely not changed.
... What shows up on blocking page is not yet marked in database.
... What is on wiki has been marked.

JJ: Do we have updated timeline?

EE: Waiting for some replies from Arron. And a number of comments on MS tests not yet addressed.

AE: Probably just lost in the hundreds of pages... I'm going through hundreds of e-mails, but so far I have zero edits still to make.
... Will reply to you as soon as I checked all e-mail. They are not all tagged correctly.
... Will try this week.

EE: Any that you didn't change?

AE: I've changed all, as far as I know.

PL: Everything that is marked as invalid in harness has been chnaged in RC4?

EE: Do you have a link?

<plinss_> http://test.csswg.org/harness/results?s=CSS21_%HTML_RC3&t=0&f[]=1&f[]=2&f[]=4&f[]=16

JJ: We can move forward without waiting for Arron's mail, I hope?

EE: Let me verify.

JJ: Just wanting to streamline the process.

EE: OK, don't need to wait for Arron.
... I'll go through my e-mails.

JJ: Yes, Arron can do this, but just no need to block RC4.

PL: Only run changed tests, no need to run all tests of RC4. Harness can show that.
... So we're left with the tests we're blocked on...

<TabAtkins> http://test.csswg.org/suites/css2.1/20101027/html4/font-family-rule-011.htm

<plinss_> http://test.csswg.org/suites/css2.1/20101027/html4/font-family-rule-011.htm

TA: Question on one test [see URL above]
... Unquoted "default" keyword.
... Is that a keyword in CSS2?

<oyvind> it's reserved

DB: I think it is reserved.
... In the para "If quoting is omitted..."

<ChrisL> reserved for future expansion

TA: OK, that's is confusing. We were looking at the syntax.
... It will mean either 'initial' or 'inherit', won't it?

<dbaron> http://www.w3.org/TR/CSS21/fonts.html#font-family-prop says:

<dbaron> The keywords 'initial' and 'default' are reserved for future use and must also be quoted when used as font names. UAs must not consider these keywords as matching the '<family-name>' type.

DB: That is the idea, but not sure we'll actually do it.

TA: I think I have all the fonts, but I stil don't get all columns numbered.

CL: That is wrong, then.

TA: But should we test whether the font works, or whether it bolds?

AE: Mail John with the issue.

DB: I get all numbers in Chrome on Linux.

TA: I get some only, this is on Windows.

PL: font-family-rule-011 passes in IE and Chrome, according to your report.

TA: Let me check...
... No, it doesn't. I must have hit the wrong button.

PL: Then I will update the DB accordingly.

HL: I talked to Prince. Maybe they can help with run-in, but I have no results yet.

PL: We have one pass on most of those.

JJ: IE passes about ten of them.

TA: I have one more issue:

<TabAtkins> http://test.csswg.org/suites/css2.1/20101027/html4/run-in-breaking-002.htm

TA: We have the wrong borders. Is that significant?

DB: The test asserts line breaking and borders must be correct.

SF: Webkit fails, it seems.

BK: Looks right in my webkit...

TA: Is there a similar test without run-in?

DB: I'm not sure the border assertion is actually correct.
... It is a ltr run-in and a rtl block. I'll have to check that.

<bradk> I take it back. The missing borders are on the opposite sides of what the text says

JJ: Chrome breaking looks correct.

TA: Not sure. Should it be on the other side?

BK: Maybe the spec is wrong then.

TA: We still not pass background intrinsic, but I'll need to study that more carefully first.

BK: Contradiction? Test says there should be no border there.

DB: I don't think the test is ambiguous. But it may be incorrect.

TA: But it says there should be a border between "header" and "Start".

DB: Ah yes.

PL: Are we agreeing on the spec?

DB: I think Chrome is correct. The test assertion needs to be corrected.

JJ: IE has a catastrophic break, but if we ignore that, it matches Chrome.

PL: Does somebody ping Boris about the test?

DB: Yes, I will.

PL: Any other tests we can review?

DB: Boris says the test dates from before our decision on inheritance.

TA: But inheritance is always the document tree. Anything else would be insane.

EE: There are sometimes anonymous boxes, which don't exist in the doc tree.

DB: Boris agrees the text about vertical borders in the test is backwards.

<dbaron> bz says he can fix the test in svn

PL: Can we look at the margin collapse tests?

<dbaron> http://test.csswg.org/suites/css2.1/20101027/html4/bidi-breaking-002.htm

EE: We had resolved that line separator behavior is undefined.
... One test fails, but it is a "may."

DB: A bunch of tests without passes. Not sure we can discuss them.

<dbaron> http://test.csswg.org/suites/css2.1/20101027/html4/abspos-non-replaced-width-margin-000.htm

<dbaron> http://test.csswg.org/suites/css2.1/20101027/html4/abspos-replaced-width-margin-000.htm

DB: [see URLs]. These are almost exhaustive tests for all combinations. They are ref tests.
... We can probably split them and get more passes. But that probably still not leads to 2 passes for all of them.

PL: I see reports that IE7 passes, but also that it fails...

<smfr> TabAtkins: not <link rel="alternate" href="http://mxr.mozilla.org/mozilla-central/source/layout/reftests/abspos-non-replaced-width-offset-margin.html"> ?

TA: Where is the reference? I don't see it linked.

<TabAtkins> smfr: Well, that file doesn't exist at all.

<plinss_> http://test.csswg.org/suites/css2.1/20101027/html4/reftest.list

<dbaron> http://test.csswg.org/suites/css2.1/20101027/html4/abspos-non-replaced-width-margin-000-ref.htm

DB: Add "-ref" to the URl to find the reference.

<dbaron> http://test.csswg.org/suites/css2.1/20101027/html4/abspos-replaced-width-margin-000-ref.htm

PL: Working on improvement to harness. But see the manifest [URL above].
... Some tests have multiple refs.

TA: That doesn't make sense.

PL: There are multiple ways to make the same result.

DB: Reftests test equivalence between features. There may be more than two equivalent features.

PL: Manifest has several lines for same test.

DB: But indeed not marked in metadata in the test. Could be reworked, but haven't done it.

PL: I'm planning to add links in the wrapper in the harness.
... We'll also have negative refs in RC4, what should not match.

SF: Useful to have sometihing in manifest that says there are mutliple refs.

EE: Like what?

SF: Some keyword I can look for.

EE: The manifest has a fixed format. Can put it in the metadata.

SF: That would probably be OK.

DB: The test we were talking about, I'd like to ask implementers to look at their bugs.

SF: It's a complex test. Splitting it out would indeed help.

DB: Relatively straightforward to edit test and reference. They are aligned line by line.

SF: I'll try to get Hyatt to look at this.

JJ: A propos of the IE7 result: IE7 fails.

DB: We might soon have two additional tests on the zero-passes list, because one fix I made caused two other tests to now fail.

EE: Update to spec in 2007 had not been incorporated in the tests.

PL: That will then give us one pass. Still need a second one...

TA: I'm willing to try...

EE: Hyatt wasn't sure about making the change. Somebody should talk to him.

<fantasai> He didn't want to change margin collapsing code without adequate test coverage

<fantasai> which we didn't have when I talked about it with him last

DB: Table background tests: it looks like Opera is very close.

<dbaron> http://test.csswg.org/suites/css2.1/20101027/html4/table-backgrounds-bs-row-001.htm

DB: Some spacing in horizontal direction only.

<dbaron> http://test.csswg.org/suites/css2.1/20101027/html4/table-backgrounds-bs-rowgroup-001.htm

DB: While webkit doesn't pass at all.
... Can Opera fix that?

<dbaron> http://test.csswg.org/suites/css2.1/20101027/html4/table-backgrounds-bs-row-001-ref.htm

DB: I haven't looked at IE.

<dbaron> http://test.csswg.org/suites/css2.1/20101027/html4/table-backgrounds-bs-rowgroup-001-ref.htm

<smfr> sorry, have to run to another mtg

HL: You think it is the same issue in both?

DB: That is what it looks like, yes.

HL: I will take a look at that.

PL: I will work on harness and reftest. It looks like the web spider attack on it is under control.

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.135 (CVS log)
$Date: 2010/12/02 10:43:08 $