W3C home > Mailing lists > Public > public-css-testsuite@w3.org > November 2010

Re: [RC3] Multiple rendering references for testcases: eg block-in-inline-*

From: Gérard Talbot <css21testsuite@gtalbot.org>
Date: Sat, 20 Nov 2010 14:53:20 -0800
Message-ID: <1bee73cdb1d7ffe7df4380e32a4d20e6.squirrel@cp3.shieldhost.com>
To: "L. David Baron" <dbaron@dbaron.org>
Cc: "Boris Zbarsky" <bzbarsky@mit.edu>, "public-css-testsuite@w3.org" <public-css-testsuite@w3.org>

Le Sam 20 novembre 2010 14:15, L. David Baron a écrit :
> On Saturday 2010-11-20 13:44 -0800, "Gérard Talbot" wrote:
>> I am confused about how "Rendering References" are supposed to work.
>> E.g.:
>> http://test.csswg.org/suites/css2.1/20101027/html4/block-in-inline-insert-001a.htm
>> has 2 rendering references:
>> http://test.csswg.org/suites/css2.1/20101027/html4/block-in-inline-insert-001-nosplit-ref.htm
>> http://test.csswg.org/suites/css2.1/20101027/html4/block-in-inline-insert-001-ref.htm
>> and those 2 differ.
>> If you examine
>> http://test.csswg.org/suites/css2.1/20101027/html4/chapter-9.html#s9.2.1.1
>> you will see that several testcases have 2 rendering references as
>> well.
>> Same thing with RC1 (20100917):
>> http://test.csswg.org/suites/css2.1/20100917/html4/chapter-9.html#s9.2.1.1
>> How is it that all implementation reports can then report that they
>> fail
>> or pass a test with 2 different rendering references?
> In the normal use of reftest, each pair of files constitutes a test
> (typically referred to as the test and the reference, though these
> names really only refer to which of the two files uses the feature
> the author was trying to test and which of the two files uses some
> other feature, which is often also worth testing).  This permits
> construction of a second test using one of the files also in another
> pair, along with a third file.

Reusing a rendering reference for other tests is fine with me. I do not
mind this; in fact, this should be encouraged.

> However, the reftests that have been contributed were incorporated
> into the test suite on the assumption that the test and reference
> names were exact, and every test should have exactly one reference.
> This isn't an unreasonable rule, but it wasn't one that I was aware
> of at the time Mozilla contributed the tests.
> Probably the pairings of the tests should be adjusted to match this
> assumption.

I still am unsure about all this.

In the example I gave, what's the rendering reference for


How is an ordinary tester under normal circumstances going to figure
this out on his/her own?

regards, Gérard
Contributions to the CSS 2.1 test suite:

CSS 2.1 test suite (RC3; October 27th 2010):

CSS 2.1 test suite contributors:
Received on Saturday, 20 November 2010 22:53:58 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 19:26:52 UTC