W3C home > Mailing lists > Public > www-style@w3.org > November 2010

[CSSWG] Minutes and Resolutions 2010-11-24

From: fantasai <fantasai.lists@inkedblade.net>
Date: Wed, 24 Nov 2010 21:51:51 -0800
Message-ID: <4CEDF977.7090202@inkedblade.net>
To: "www-style@w3.org" <www-style@w3.org>
Summary:

   - RESOLVED: fantasai is the liaison with EPUB
   - RESOLVED: Publish FPWD of CSS3 Writing Modes
   - Reviewed lists of CSS2.1 tests without enough passes to exit CR
       http://wiki.csswg.org/test/css2.1/results
       http://test.csswg.org/harness/results?s=CSS21_%25HTML_RC3&t=0&f[]=1&f[]=4&f[]=8&f[]=16

====== Full minutes ======

Present:

   Tab Atkins
   David Baron
   Bert Bos
   Arron Eicholz
   Elika J. Etemad
   Koji Ishii
   Håkon Wium Lie
   Peter Linss
   Geoffrey Sneddon (Opera)

<RRSAgent> logging to http://www.w3.org/2010/11/24-CSS-irc

Scribe: Tab Atkins

EPUB Liaison
------------

   glazou: Before testsuite discussion, anything else?
   ChrisL: Brief item.
   ChrisL: Liaison with EPUB, and whether the group is willing to do it.
   fantasai: We already resolved to do it at the FtF.
   <ChrisL> but please can I have a minuted resolution that the liaison
            is Elika, not just a generic 'liaison' with no named person
   RESOLVED: Elika is the liaison with epub.
   * ChrisL thanks

CSS3 Writing Modes
------------------

   fantasai: I'd like to publish FPWD of Writing Modes.
   fantasai: jdaggett said he was okay with it.
   RESOLVED: Publish FPWD of Writing MOdes.

CSS2.1 Test Suite Results
-------------------------

   <gsnedders> Just to note, I've been busy with university stuff over
               the past few days, and am not entirely up-to-date on
               CSS 2.1 testsuite stuff.
   <glazou> http://wiki.csswg.org/test/css2.1/results
   <glazou> http://test.csswg.org/harness/results?s=CSS21_%25HTML_RC3&t=0&f[]=1&f[]=4&f[]=8&f[]=16
   <fantasai> http://test.csswg.org/source/filename-list
   <fantasai> links to the latest by filename

   glazou: We have 107 tests that aren't passed by two impls.
   glazou: And over 9k tests that *are* passed by at least two.
   glazou: At the last call, vendors were asked to review these tests.
   dbaron: There's a lot of activity on the wiki and on the list.
   glazou: Tab, you sent an email about not having time to review yet?
   TabAtkins: Yeah, I'll review after I run the next RC in Chrome.
   glazou: What about tests that have no comments at all?
   dbaron: In some cases I didn't look at them because I don't really
           know about them.
   glazou: When is the next RC going to be published?
   fantasai: I'm still working through the backlog of reported problems -
             I'm in mid October for feedback (going backwards), and I think
             I have about two days of work left to fix all the errors.
   fantasai: So probably next week, but possibly I can make it happen this
             weekend.
   gsnedders: From our point of view, Oyvind was looking through the tests
              and was asking if we're delaying the next RC of the testsuite,
              what we'll doa bout the results deadline for the RC.
   * dbaron notes Thanksgiving is tomorrow
   glazou: I'm somewhat disappointed that we're not yet ready with all the
           reviews by now.
   <gsnedders> Nah, I was asking what we're doing about the expected date
               for IRs for the next RC
   <gsnedders> Given that the RC is later than the date mentioned in the
               resolutions from the last F2F
   <ChrisL> puzzled by links to tests
   <ChrisL> http://test.csswg.org/harness/testcase?s=CSS21_%HTML_RC3&c=at-charset-013
+howcome
   <ChrisL> <b>Fatal error</b>:  Call to a member function get_test_suite()
            on a non-object in <b>/sites/csswg.org/test/htdocs/harness/testcase.php</b>
            on line <b>237</b><br />
   plinss: You need to use a report that doesn't combine the HTML and
           XML versions
   <plinss> http://test.csswg.org/harness/results?s=CSS21_HTML_RC3&t=0&f[]=1&f[]=4&f[]=8&f[]=16

   arron: Tab, can you run the tests that are on the wiki for now?
   TabAtkins: Yeah, I can do that.
   <Bert> (A number of tests is also marked invalid on the wiki, that also
           reduces the issues list considerably, doesn't it?)
   dbaron: Note that I wrote in a few places that Beta8 passes but not
           Beta6, because there are a few places where we fixed bugs.
   ChrisL: If you follow the link in the agenda to the test, you get an
           error message.
   plinss: If you're just running a few tests for a new browser, please
           do the test in the harness.
   <ChrisL> bug in at-charset-013 - pass criteria are ambiguous
   <fantasai> ChrisL: please report any errors to public-css-testsuite /
              wiki error page
   * fantasai cannot track errors via IRC
   * fantasai has a hard enough time tracking them through email
   <ChrisL> ok, thought we might fix some on the call
   arron: Peter, did you hook up the logic to display reftests correctly
          in the harness?
   plinss: Not yet

   glazou: I suggested in the agenda email to review some tests in the conf call.
   glazou: In the list of tests with 0 passes, we have a few with "invalid?"
           markers.  Looking at these could be useful.
   <glazou> http://test.csswg.org/suites/css2.1/20101027/html4/background-intrinsic-004.htm
   dbaron: Looks like Oyvind marked it as invalid.
   <gsnedders> oyvind: ping?
   fantasai: Given that the other tests pass, I'm guessing this will pass
             as soon as it's fixed.
   <oyvind> first of all, the version in the RC has been updated by fantasai
   <oyvind> the "invalid" word links to the mail I sent about the remaining issue -
            http://lists.w3.org/Archives/Public/public-css-testsuite/2010Nov/0119.html
   <oyvind> I don't have all the browsers listed (XP, not win7, for one),
            but IIRC Chrome still failed
   ChrisL: I dropped in a link to a particular test I thought might be quick to do.
   <ChrisL> http://test.csswg.org/harness/testcase?s=CSS21_HTML_RC3&c=at-charset-013
   ChrisL: The text is two lines with identical text, and the pass condition
           says it passed if "Filler Text" is green.  One line of "Filler Text"
           is. I think the pass condition should just be that both lines
           should be red.

   fantasai: Some of the printing-related tests might pass on Prince, for example.
   howcome: I think you may be right, especially for cases where we know
           the test is valid and just need a pass.
   glazou: Print-related tests should also pass in Print Preview.
   ChrisL: There's some page-break-* tests that I suspect Prince will be
           more likely to pass.
   glazou: When could we get a Prince report?
   howcome: Dunno.
   glazou: You don't need the whole suite - just the 100 or less that don't
           have two passes yet.
   <ChrisL> here is the list of 100 tests
   <ChrisL> http://test.csswg.org/harness/results?s=CSS21_%25HTML_RC3&t=0&f[]=1&f[]=4&f[]=8&f[]=16
   <glazou> howcome: http://wiki.csswg.org/test/css2.1/results

   glazou: What other tests can we fix now?
   <glazou> http://test.csswg.org/suites/css2.1/20101027/html4/first-page-selectors-003.htm
   glazou: This is mentioned by Oyvind as probably invalid.
   <gsnedders> http://lists.w3.org/Archives/Public/public-css-testsuite/2010Nov/0121.html
               is oyvind's email about it
   fantasai: The test is correct - that's the intention of the spec.
   arron: But is the spec clear enough to say that?
   howcome: I don't see :first-page as applying...
   fantasai: If you force a break before the first page, it just puts
             the first page on a particular side.  So the left page
             would be the first page, it's just on a different page than
             it would be otherwise.
   <oyvind> (this text was introduced as part of issue 160 btw)
   <Bert> (I agree with Opera's interpretation: not the first page.)
   ACTION howcome to check first-page-selectors-003 to make sure it's valid.
   <trackbot> Created ACTION-277
   <fantasai> Wrt page-breaking, CSS3 Page is rather explicit about
              forcing a :left :first page
   <fantasai> "However, to force a ':left' or ':right' first page, authors
              MAY insert a page break before the first generated box (e.g.,
              in HTML, specify this for the BODY element)."
   <fantasai> http://www.w3.org/TR/css3-page/
   <oyvind> fantasai: "page break" still suggests to me that there will
            actually be a page before it (though empty)

   plinss: Some tests are somewhat ambiguous if someone isn't familiar with
           CSS's use of the term "box", because there's no visible box drawn
           on the page.  Suggest updating to say "paragraph" or similar.
   <fantasai> plinss: if you notice more such pages, please send a list to
              public-css-testsuite
   glazou: What about allowed-page-breaks-002?
   glazou: Should we defer to next week after the review?
   <oyvind> allowed-page-breaks-002 assumes 4em is between three and four
            lines, but other than that I couldn't see anything obviously
            wrong (there's a negative margin whith trips us up a bit)
   glazou: I am *insisting* on these reviews.
   glazou: Hopefully this should further drastically reduce the number of
           <2 pass tests, so we can focus our attention on the handful of
           truly invalid ones.
   <Bert> (Yes, oyvind, that's what I thought, too. It needs a 'line-height'
          property to be sure.)

   arron: I think there's another easy one to check.
   arron: overflow-applies-to-010
   <arron> http://test.csswg.org/suites/css2.1/20101027/html4/overflow-applies-to-010.htm
   <dbaron> Well, I was against that decision in the first place.
   arron: The spec currently matches this test, but no impl does it.
          I think the spec was changed here about a year ago.
   arron: So do we want to just reverse the spec change?
   dbaron: I wanted the spec to go the other way when we discussed it the
           last time.
   arron: And all impls would pass if we went back to what we had before.
   fantasai: Could we just make it undefined in CSS2.1?
   glazou: If we make it undefined we just remove the test.
   fantasai: I think we need a coherent model for outside/parent/whatever
             bullets before we can determine what this *should* do.
   RESOLVED: Change text surrounding overflow and bullets to make it
             explicitly undefined in CSS2.1.

   TabAtkins: text-transform-bicameral-008 is invalid and easy to fix -
              the test is explicitly stating that the same capital letter
              should downcase to two different lowercase letters.
   <dbaron> -004 has the same problem
   <dbaron> (but it's in the 0-pass list)
   TabAtkins: We should just remove the couple of final-form letters that
              it's trying to test.
   <gsnedders> text-transform-bicameral-004 apparently passes in IE9, FWIW
   <gsnedders> oh, wait, 008
   <gsnedders> I can't read :)
   <dbaron> gsnedders, we were discussing both

   fantasai: Question on the run-in-contains tests.
   fantasai: Nobody's said they're invalid, so what are we gonna do about them?
   dbaron: I've always been in favor of removing run-in from the spec, but...
   <fantasai> Basically, if we're reviewing tests today, we should review
              tests that are not known invalid :)
   <fantasai> Because that doesn't depend on RC4
   glazou: Do you have a suggestion for what do about the tests, elika?
   fantasai: Nope.
   fantasai: Let's assume the run-in-contains tests are valid.
             There's still only one pass.
   fantasai: Does Prince support run-ins?
   howcome: ^^^
   howcome: I'm not sure, but I can check.
   fantasai: Prince does support run-ins.
   * gsnedders just opened his mouth to say that when fantasai did
   <ChrisL> yes it does http://www.princexml.com/doc/7.0/properties/
   fantasai: So the way forward here would be to run these tests on Prince.
   ChrisL: I've got the testsuite SVN pulled down locally, so I'll run them
           in Prince 7.1.
   <gsnedders> Make sure they're run with the file encoding as utf-8, if
               running locally

   fantasai: I think I updated word-spacing-characters-001, and I don't
             recall getting many passes on it.
   <fantasai> http://test.csswg.org/source/contributors/fantasai/submitted/css2.1/word-spacing-characters-001.xht
   <fantasai> FF passes
   arron: I think IE9 passes all but the final one.
   TabAtkins: That one is pre, and also includes three kinds of whitespace -
              spaces, nbsp, and tabs.
   glazou: white-space-control-characters-001 is passed by FFb and Opera 11b.
   <dbaron> That's one of the ones I never managed to check.
   <gsnedders> white-space-control-characters-001 seems to have been broken
               in RC2, so maybe coming up in the list due to lack of reported
               results?
   * glazou tests http://test.csswg.org/suites/css2.1/20101027/html4/before-after-table-parts-001.htm and collapses :-)
   <dbaron> glazou, it's a reftest
   <dbaron> glazou, just load 001-ref in another tab and switch between them

   fantasai: The margin-collapse-clear tests were due to a spec change,
             since the earlier behavior made no sense (clear could
             sometimes make an element move up).
   fantasai: The tests are valid, they're just not passed by anyone yet.
   fantasai: At this point we either need to change the spec (which I don't
             think is a good idea) or we need impls to make a fix, so we
             can evaluate the web compat impact.
   glazou: How long do people think it will take to fix implementations?
   arron: I doubt IE9 will take this change, at least right now.
   gsnedders: We're unlikely to make it for Opera 11.
   glazou: What happens if we don't get two passes?  This could potentially
           take months.
   TabAtkins: We attempt to justify why the current spec is good, despite
              the lack of passes?
   glazou: The feature isn't at-risk, so we need to pass.
   fantasai: I think if we get at least one impl so we can evaluate web
             compat, we can make a case as to why we don't have two passes
             for this case.
   glazou: When you say "make a case", do you mean when we discuss IRs with
          the director?
   fantasai: Yes.
   <bradk> Move to errata after REC?
   glazou: We have exit criteria decided *long* ago, and if we start to
           special-case things it's no good.
   fantasai: I don't think we'll have to do this for *any* other feature,
             just this one.

   glazou: Next week, *have reviews ready please*.

Meeting closed.

<bradk> I mean, change the spec to match implementations, and then have
         an errata later on to change it back?
<TabAtkins> bradk: That's such disgusting politics-maneuvering.  >_<
<bradk> Yeah, I know, but it is only this one test...
<TabAtkins> If it's only one test, I agree with elika that we should just
             get at least one impl and then try to justify it on its merits.
<fantasai> I think it's more honest to just say we don't have the passes
<bradk> It would get us a pass, if we are not willing to overlook
         implementations not being ready in time.

<dbaron> http://test.csswg.org/source/contributors/fantasai/submitted/css2.1/word-spacing-characters-001.xht
          passes in Firefox trunk but not in 4.0b6; I don't know why
<dbaron> er, never mind, I think it's related to the default font size
          requirement
<dbaron> It passes with Firefox's default default font sizes, but not
          the default font sizes required by the test suite.
<fantasai> dbaron: ???
<fantasai> dbaron: The test should be font-size independent
<dbaron> fantasai, well, it's somehow profile-dependent
<fantasai> @_@
<dbaron> fantasai, it passes in a freshly created profile and fails in the
          profile I used for the test suite that has the preference tweaks
Received on Thursday, 25 November 2010 05:52:31 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 17:20:34 GMT