W3C home > Mailing lists > Public > public-css-testsuite@w3.org > August 2015

Re: [CSS21][CSS22] Seeking review of position:fixed edge case test

From: Linss, Peter <peter.linss@hp.com>
Date: Mon, 17 Aug 2015 23:10:42 +0000
To: Gérard Talbot <css21testsuite@gtalbot.org>
CC: Chris Rebert <csswg@rebertia.com>, Public CSS Test suite mailing list <public-css-testsuite@w3.org>, "chris@rebertia.com" <chris@rebertia.com>
Message-ID: <46F5C328-52C2-4F3C-82AB-A7A832AEFD82@hp.com>

On Aug 17, 2015, at 1:49 PM, Gérard Talbot <css21testsuite@gtalbot.org> wrote:

> Le 2015-08-17 15:08, Chris Rebert a écrit :
>> On Tue, Aug 4, 2015 at 9:06 PM, Gérard Talbot
>> <css21testsuite@gtalbot.org> wrote:
>>> Le 2015-07-27 03:21, Chris Rebert a écrit :
> 
>>> 1- There is already a test with the filename position-fixed-001
>>> http://test.csswg.org/suites/css2.1/nightly-unstable/html4/position-fixed-001.htm
>>> in the test suite. Fortunately for you, your test, in my opinion, is really
>>> about 'left: auto' offset. So, in all fairness, you were destined for a test
>>> filename change anyway.
>> A. This "testcase filename itself must be globally unique" requirement
>> doesn't seem to be documented anywhere on Test the Web Forward.
> 
> When the first version of the file format documentation was created, zero-filled number suffix was explained to keep filename unique. As more and more test suites for 40+ specifications were created, there is a need to keep filename unique... otherwise test searching (with CSS Test Suite Manager Shepherd system: http://test.csswg.org/shepherd/ ) gets more difficult, confusing and returning several occurences.
> 
> But, you are right; I agree with you that this "testcase filename itself must be globally unique" guideline or requirement should be stated somewhere in the Test the Web Forward documentation.

Agreed, this used to be stated but apparently has gotten lost as the docs have merged and transitioned.

We have this requirement because test sources get built into one or more test suites (depending on their spec links). The built test suites have a flat directory structure and do not attempt to preserve the source directory structure, so filename collisions can happen.

> 
>> B. The "Automatic validation checks" performed by W3C Sync Bot
>> (https://github.com/w3c/csswg-test/pull/813#issuecomment-124963013 )
>> didn't complain about this. If it's really a requirement, this seems
>> like something that the bot could and should enforce.
>> C. Okay, I renamed it to "left-offset-position-fixed-001.xht"
>> following your example.
>>> 2-
>>> <!DOCTYPE html SYSTEM "about:legacy-compat">
>>> Please use
>>> <!DOCTYPE html>
>> Done.
>>> 3-
>>>    <link rel="help"
>>> href="https://drafts.csswg.org/css2/visuren.html#choose-position" />
>>>    <link rel="help"
>>> href="https://drafts.csswg.org/css2/visudet.html#abs-non-replaced-width" />
>>>    <link rel="help"
>>> href="https://drafts.csswg.org/css2/visudet.html#abs-non-replaced-height" />
>>> Do not link to draft spec as they can change or will change in the future.
>> Fixed.
>> Sorry, I had gotten confused since web-platform-tests uses the exact
>> opposite convention.
>> For the record, http://testthewebforward.org/docs/css-metadata.html
>> does not explicitly mention this rule, although its example does use
>> non-draft links.
> 
> Then, in my opinion, it should state such guideline and then be updated accordingly.
> 
> Such guideline is not mentioned in the documentation. As a test author, you want to provide a link to the most stable version of a spec so that your test remains as much reliable and complete as possible. Once a spec becomes Proposed Recommendation or Technical Recommendation, then *I think (speculation)* working drafts no longer exist and links are not redirected.
> 
> "
> This is a draft document and may be updated, replaced or obsoleted by other documents at any time. It is inappropriate to cite this document as other than work in progress.
> "
> https://drafts.csswg.org/css2/#status
> 

This is actually not a requirement. While it's best for test to link to the most stable version of a spec, linking to a draft is fine, and sometimes necessary if the tested features are not published on /TR yet. Our tools deal with draft links just fine.

It is possible however for anchors to change in drafts, and if that happens the tests will beed to be updated, but as drafts mature we do try to avoid it.

> 
>> Also, again, this seems like something that W3C Sync Bot ought to validate.

It does check the links, if it's not complaining about the drafts links, then they're fine.

> 
> I do not use or know about that W3C Sync Bot; I think (just a hunch/speculation) such software only synchronized between github and Mercurial ... but I really don't know...

SyncBot is an account on GutHub that Shepherd uses to sync the GitHub and Mercurial repositories. Shepherd also looks at submitted pull requests and runs the same validation checks that it does when files are pushed to the repo, these "pre-warnings" are what SyncBot generates on GitHub.



Received on Monday, 17 August 2015 23:12:41 UTC

This archive was generated by hypermail 2.4.0 : Friday, 20 January 2023 19:58:21 UTC