update-w3c-deps: Fix a bug of wrong revisions in a CL description

update-w3c-deps: Fix a bug of wrong revisions in a CL description.

"git show-ref origin/master" ran on a wrong directory.

BUG=

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=196372

[tkent, 2015-06-03 02:18:06]

tkent@chromium.org changed reviewers:
+ dpranke@chromium.org, jsbell@chromium.org

[tkent, 2015-06-03 02:18:07]

dpranke@, jsbell@, please review this.

Motivation is to import new directory without updating existing tests.

[Dirk Pranke, 2015-06-03 02:37:34]

On 2015/06/03 02:18:07, tkent wrote:
> dpranke@, jsbell@, please review this.
> 
> Motivation is to import new directory without updating existing tests.

By design we explicitly do *not* have the ability to do this. 

It was always the intent that we would roll to tip-of-tree css and wpt every
time we rolled anything, 
to limit any possible confusion of what had or hadn't been imported, and at what
reviisions.

That way the entire contents of LayoutTests/imported could be trivially
recreated by running a single command
at any point.

I am greatly afraid of how unmaintainable things would become if we allowed
people to import parts of
the repo piecemeal at different times.

Does that make sense?

[Dirk Pranke, 2015-06-03 02:40:23]

+joelo for important context since you'll be taking over this :)

[Dirk Pranke, 2015-06-03 03:26:48]

Two other important comments on this topic ...

1) in a perfect world, we'd be running most if not all of the upstream tests, 
and we'd pull the repos in via DEPS rather than manually copying files into 
Blink (you may recall we did pull them in via DEPS for a while, which made
imports very painful; it will be much less painful and viable again once the 
Blink merge is done). 

In that world, since they are git repos, it's not really possible to pull in 
directory A at one rev and directory B at another; the whole repo 
needs to see the same version.

It would be possible to have wpt and css at different revisions, but I'm 
not sure that it's worth handling that case given that css is merging into
wpt at some point.

2) As a result of one, updating the w3c repos is (or should be) really 
more like gardening than it is like something individuals do on their
own.

We should give thought to setting up auto-rollers and some process
for triaging new failures, so the person who wants new versions of X 
doesn't get stuck having to deal with new failures in Y.

[kojii, 2015-06-03 03:41:19]

It may be irrelevant here, but please feel free to ping me if you're stuck with
CSS Writing Modes tests. It's quite active these days, oftentimes 50-100 tests
are updated/added. 200 tests were added in these 3 months, a small team is
trying to complete the test suites by the end of September.

I'm planning to import once a week or two until it stabilize, to minimize the
troubles these updates might cause, but I'm more than happy to help out when it
troubles anyone anytime.

[tkent, 2015-06-03 03:48:27]

dpranke,

ok, I withdraw the skip feature and the revision feature.  Now this CL handles
only the CL description bug. Please review this.

kojii,
Yes, the massive update of CSS Writing Modes tests was the trigger.  If we run
udpate-w3c-deps now, we'll have thousand lines diff though we ran it yesterday
:(

[Dirk Pranke, 2015-06-03 04:28:47]

lgtm. I'm surprised that bug has been around for so long.

Also, I hope I was not too harsh earlier; that was not my intent. 

It's just that it is easy for people to not have the full context of how I had
hoped for this stuff to work. Hopefully if we get more in the habit of importing
things regularly (like we're doing at the moment), this will become common
knowledge.

[tkent, 2015-06-03 04:42:40]

The CQ bit was checked by tkent@chromium.org

[commit-bot: I haz the power, 2015-06-03 04:43:05]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1164833003/20001

[commit-bot: I haz the power, 2015-06-03 05:29:42]

Committed patchset #2 (id:20001) as
https://src.chromium.org/viewvc/blink?view=rev&revision=196372