Re: CSS test repo refactored - branch ready for review

Hey Gérard,

To address your concerns, I reorganized css21 and moved all the files that
were at the root into section subdirectories.  I did this in my forked
repo on Github so you could easily review it there before pushing it [1].
(Note: If/when I do land these changes, they’ll be pushed directly to
Mercurial rather then merged from Github).

The commit with all the files moved is here [2]. If you’d like to clone
this repo and browse it locally, you’ll have to clone the ‘organize-css21'
branch:

git clone -b organize-css21 git@github.com:rhauck/csswg-test.git


Or, if you do not have a public key set up:

git clone -b organize-css21 https://github.com/rhauck/csswg-test.git




Let me know if this looks better to you and if you have any other
suggestions. It’s fairly straightforward to make changes.

Also, I added the script I used to check that no paths to reference or
support files were broken [3].  I’ll add this to the main repo when I land
these changes.

Cheers,
-Rebecca


[1] https://github.com/rhauck/csswg-test/tree/organize-css21

[2] 
https://github.com/rhauck/csswg-test/commit/911ca986fe57ce43075cae6b05d11a1

9c55118fb
[3] 
https://github.com/rhauck/csswg-test/blob/organize-css21/tools/path_checker

.py

On 6/2/14, 8:04 PM, "Gérard Talbot" <css21testsuite@gtalbot.org> wrote:

>Le 2014-06-02 21:41, Rebecca Hauck a écrit :
>> Hi Gérard,
>> 
>> [Š]
>> 
>>> 
>>>> All the locations of files within the repository are maintained by
>>>> people (you really don't want an automated process modifying the
>>>> repository).
>>> 
>>> Forgive my question but ... where are *all of my submitted tests* now
>>> ?
>>> in my local repository? and in http://test.csswg.org/source/ ?
>>> 
>>> I ask this because ...
>>> There used to be a /contributors folder (which was in the /src/
>>> folder)
>>> where all contributors had their folder by their username. Now, such
>>> /contributors folder is only visible, only accessible via mercurial
>>> and
>>> has only a few folders.
>>> I can see right now an
>>> /contributors/gtalbot/submitted
>>> but it is empty and this folder is not viewable, not accessible from
>>> http://test.csswg.org/source/

>> 
>> 
>> We are no longer keeping submitted tests in folders by user or company
>> name. We still have a place for work-in-progress that has user/company
>> folders under it, but this is really to support the legacy process when
>> people were pushing directly to Mercurial.
>
>I still use Mercurial.
>
>> Those files needed to be parked
>> somewhere and I chose the name 'work-in-progress' because it was more
>> descriptive than Œincoming¹. We now no longer need special instructions
>> or
>> specific directory names to submit a test. It¹s very simple - if you¹re
>> submitting tests for the Backgrounds & Borders spec (for example), your
>> test goes in the css-backgrounds-3 directory. This makes it very easy
>> to
>> find all of the tests for a given spec in one place rather than being
>> spread across multiple user/company folders.
>
>Understandable.
>
>> It will also make it easier
>> for vendors to import tests for any given spec or set of specs
>> (automated
>> or otherwise) as it eliminates the need to parse the test files to
>> figure
>> out what specs they¹re testing. Implementors really want to access
>> tests
>> by specs and not by who authored them (although we can rely on the
>> metadata for that if needed).
>
>Understandable.
>
>> If you still have contributors folder locally, you likely have some
>> hidden
>> dot files that prevented it from being removed when you updated your
>> local
>> repo.
>
>Yes. After checking the viewing (display) of hidden files, I see hidden
>.directory files in all folders of /contributors/gtalbot/
>
>> It happened to me as well with those pesky .DS_Store files.
>
>.DS_Store files are MacOSX-related, I believe.
>
>> What
>> you see in the web interface at http://test.csswg.org/source/ is
>> accurate
>> and you can safely delete your local contributors folder.
>> 
>> And of course as you know, you can always query all of the tests you
>> authored via Shepherd or grep locally if you wish. Your tests were
>> across
>> multiple specs so they got filed under the appropriate spec
>> directories.
>
>Some of my tests were accross multiple specs. Others were not. In any
>case, ~= 260 of my tests were not moved into appropriate spec
>directories.
>
>Eg
>
>http://test.csswg.org/source/css21/background-position-applies-to-001a.xht

>
>is not in
>
>http://test.csswg.org/source/css21/backgrounds/

>
>... where I think it should be instead.
>
>
>> If you want the complete list of where everything went, the HG
>> changeset
>> is here [1], but the Github interface actually gives you a little nicer
>> view of the the renaming [2]. Everything you had in
>> /contributors/gtalbot/incoming moved to /work-in-progress/gtalbot [3].
>> The
>> full description of the changes that were made are at the top of this
>> thread [4].
>> 
>> On somewhat of a side note, since we¹re now in the Github world, any
>> new
>> tests that land in the repo should start with a pull request where they
>> will be reviewed, approved, and merged from there. We are favoring this
>> over using the mailing list for test reviews for reasons I outlined
>> here
>> [5].
>
>
>I'm sorry. I am still outdated then.
>
>
>> All W3C test submissions are now done this way and it¹s a much
>> cleaner approach than relying on an external system or directory naming
>> convention to reflect test status.  There are still a few who push
>> directly to Mercurial, but we are not broadcasting this workflow any
>> longer and and even the veterans are discouraged from doing this if
>> what
>> they¹re submitting needs review. I personally only do so for
>> housekeeping
>> tasks that don¹t require a review. With this new workflow, all tests
>> that
>> are merged into the repo can be assumed reviewed and approved.
>> 
>> Now, I realize that we still have many tests unreviewed from before we
>> adopted Github. We can still use Shepherd for tracking these-- either
>> its
>> API or the web interface. However, at some point, we¹ll have to decide
>> how
>> to reconcile these tests as it¹s probably not realistic to expect that
>> thousands of tests will ever be reviewed by humans. Peter and I have
>> had
>> some offline discussions about how to address this, but this is a issue
>> to
>> solve later.
>
>I have ideas on this.
>
>> We have Shepherd in the meantime (luckily). We wanted to make
>> these changes first to move closer to the way the rest of the W3C
>> manages
>> tests. We¹re now in a better position to merge/move into the main W3C
>> web-platform-tests repo [6]. That¹s also a separate discussion that
>> only
>> just began at the last CSSWG F2F and it will certainly pick up again
>> soon.
>>  We just had to do this part first.
>> 
>> Let me know if you have any other questions & thanks again for your
>> incredible attention to detail. :)
>
>:)
>
>Gérard
>
>> Cheers,
>> -Rebecca
>> 
>> 
>> [1] http://hg.csswg.org/test/rev/8ed45b2c892f

>> [2]
>> 
>>https://github.com/w3c/csswg-test/commit/70cfca08acf7fdb1119eb2e7ecbccd91

>>15
>> cd81c7
>> [3]
>> 
>>https://github.com/w3c/csswg-test/commit/55ecf8c9c7bfcb67d059ea68dc8041d8

>>a8
>> 0cdf7f
>> [4]
>> 
>>http://lists.w3.org/Archives/Public/public-css-testsuite/2014May/0000.htm

>>l
>> [5]
>> 
>>http://lists.w3.org/Archives/Public/public-css-testsuite/2013Nov/0014.htm

>>l
>> [6] https://github.com/w3c/web-platform-tests

>
>-- 
>Web authors' contributions to CSS 2.1 test suite
>http://www.gtalbot.org/BrowserBugsSection/css21testsuite/web-authors-contr

>ibutions-css21-testsuite.html
>CSS 2.1 Test suite RC6, March 23rd 2011
>http://test.csswg.org/suites/css2.1/20110323/html4/toc.html

Received on Wednesday, 4 June 2014 00:12:13 UTC