Revert 289526 "Fix Mac sandbox meta data access"

Revert 289526 "Fix Mac sandbox meta data access"

Speculative revert as a lot of Mac bots broke when this landed:

http://build.chromium.org/p/chromium.mac/builders/Mac%2010.6%20Tests%20%28dbg%29%282%29/builds/54365
http://build.chromium.org/p/chromium.mac/builders/Mac%2010.6%20Tests%20%28dbg%29%284%29/builds/43947
http://build.chromium.org/p/chromium.mac/builders/Mac%2010.6%20Tests%20%28dbg%29%283%29/builds/52314

> Fix Mac sandbox meta data access
> 
> Sandbox::AllowMetadataForPath() currently allow all metadata access due to
> https://codereview.chromium.org/10539009/ made the for loop comparison
> in Sandbox::AllowMetadataForPath() always false, when we actually only
> want to allow access to the path and all its parent path until root.
> 
> Turn the for loop to a do/while loop instead as it's a better fit, also
> add a test case for Sandbox::AllowMetadataForPath().
> 
> It should only affect component builds on OS X 10.6 and utility process
> as no other process is using this mechanism.
> 
> Review URL: https://codereview.chromium.org/472513002

TBR=jiangj@opera.com

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=289541

[bartfab (slow), 2014-08-14 13:05:34]

Committed patchset #1 manually as r289541 (tree was closed).

[Jiang Jiang, 2014-08-14 13:18:04]

On 2014/08/14 13:05:34, bartfab wrote:
> Committed patchset #1 manually as r289541 (tree was closed).

Hi, are all the test failures happen on OS X 10.6?

[bartfab (slow), 2014-08-14 13:24:56]

On 2014/08/14 13:18:04, Jiang wrote:
> On 2014/08/14 13:05:34, bartfab wrote:
> > Committed patchset #1 manually as r289541 (tree was closed).
> 
> Hi, are all the test failures happen on OS X 10.6?

As just discussed on IRC: It looks like this is OS X 10.6 only, yes. I
speculatively reverted your CL and am monitoring the bots. If the CL is
vindicated, I will re-land it.

[Jiang Jiang, 2014-08-14 15:41:47]

On 2014/08/14 13:24:56, bartfab wrote:
> On 2014/08/14 13:18:04, Jiang wrote:
> > On 2014/08/14 13:05:34, bartfab wrote:
> > > Committed patchset #1 manually as r289541 (tree was closed).
> > 
> > Hi, are all the test failures happen on OS X 10.6?
> 
> As just discussed on IRC: It looks like this is OS X 10.6 only, yes. I
> speculatively reverted your CL and am monitoring the bots. If the CL is
> vindicated, I will re-land it.

I see. So it's basically http://crbug.com/127465 happening again as the original
fix at r141020 didn't really work but only hide the issue.