PlatformAppBrowserTest.ShellWindowRestorePosition: Test create after window close

PlatformAppBrowserTest.ShellWindowRestorePosition: Test create after window close

BUG=238310

[stevenjb, 2013-05-03 23:20:27]

I discovered that the save geometry tests weren't actually waiting for the
window to close before opening, so restore wasn't actually being tested, which
appears to break OSX.

I believe this should fix the issue correctly.

[benwells, 2013-05-06 07:21:56]

On 2013/05/03 23:20:27, stevenjb (chromium) wrote:
> I discovered that the save geometry tests weren't actually waiting for the
> window to close before opening, so restore wasn't actually being tested, which
> appears to break OSX.
> 
> I believe this should fix the issue correctly.

Does this break Mac? I don't see any mac specific code to fiddle with the
bounds. I know mac had problems with this.

[stevenjb, 2013-05-06 16:32:59]

On 2013/05/06 07:21:56, benwells wrote:
> On 2013/05/03 23:20:27, stevenjb (chromium) wrote:
> > I discovered that the save geometry tests weren't actually waiting for the
> > window to close before opening, so restore wasn't actually being tested,
which
> > appears to break OSX.
> > 
> > I believe this should fix the issue correctly.
> 
> Does this break Mac? I don't see any mac specific code to fiddle with the
> bounds. I know mac had problems with this.

Correcting the test broke mac.

Moving the Insert subtraction from ShellWindow to NativeAppWindow (gtk and views
only) should fix it, although trybots appear to be flakey. Will re-run mac and
win.

[stevenjb, 2013-05-06 18:00:22]

On 2013/05/06 16:32:59, stevenjb (chromium) wrote:
> On 2013/05/06 07:21:56, benwells wrote:
> > On 2013/05/03 23:20:27, stevenjb (chromium) wrote:
> > > I discovered that the save geometry tests weren't actually waiting for the
> > > window to close before opening, so restore wasn't actually being tested,
> which
> > > appears to break OSX.
> > > 
> > > I believe this should fix the issue correctly.
> > 
> > Does this break Mac? I don't see any mac specific code to fiddle with the
> > bounds. I know mac had problems with this.
> 
> Correcting the test broke mac.
> 
> Moving the Insert subtraction from ShellWindow to NativeAppWindow (gtk and
views
> only) should fix it, although trybots appear to be flakey. Will re-run mac and
> win.

Hmm. Mac appears to still break. I may need to set up a Mac dev environment or
solicit some help here and disable the test on OSX for now (I think having the
test disabled would be better than it not correctly testing the behavior).

[benwells, 2013-05-07 02:22:29]

On 2013/05/06 18:00:22, stevenjb (chromium) wrote:
> On 2013/05/06 16:32:59, stevenjb (chromium) wrote:
> > On 2013/05/06 07:21:56, benwells wrote:
> > > On 2013/05/03 23:20:27, stevenjb (chromium) wrote:
> > > > I discovered that the save geometry tests weren't actually waiting for
the
> > > > window to close before opening, so restore wasn't actually being tested,
> > which
> > > > appears to break OSX.
> > > > 
> > > > I believe this should fix the issue correctly.
> > > 
> > > Does this break Mac? I don't see any mac specific code to fiddle with the
> > > bounds. I know mac had problems with this.
> > 
> > Correcting the test broke mac.
> > 
> > Moving the Insert subtraction from ShellWindow to NativeAppWindow (gtk and
> views
> > only) should fix it, although trybots appear to be flakey. Will re-run mac
and
> > win.
> 
> Hmm. Mac appears to still break. I may need to set up a Mac dev environment or
> solicit some help here and disable the test on OSX for now (I think having the
> test disabled would be better than it not correctly testing the behavior).

As you've taken code out of the platform independent bit, I assume you need to
add code into the corresponding mac bit.

Why did you remove the code? The CL description doesn't explain (it just
mentions the test).

[stevenjb, 2013-05-07 15:44:53]

On 2013/05/07 02:22:29, benwells wrote:
> On 2013/05/06 18:00:22, stevenjb (chromium) wrote:
> > On 2013/05/06 16:32:59, stevenjb (chromium) wrote:
> > > On 2013/05/06 07:21:56, benwells wrote:
> > > > On 2013/05/03 23:20:27, stevenjb (chromium) wrote:
> > > > > I discovered that the save geometry tests weren't actually waiting for
> the
> > > > > window to close before opening, so restore wasn't actually being
tested,
> > > which
> > > > > appears to break OSX.
> > > > > 
> > > > > I believe this should fix the issue correctly.
> > > > 
> > > > Does this break Mac? I don't see any mac specific code to fiddle with
the
> > > > bounds. I know mac had problems with this.
> > > 
> > > Correcting the test broke mac.
> > > 
> > > Moving the Insert subtraction from ShellWindow to NativeAppWindow (gtk and
> > views
> > > only) should fix it, although trybots appear to be flakey. Will re-run mac
> and
> > > win.
> > 
> > Hmm. Mac appears to still break. I may need to set up a Mac dev environment
or
> > solicit some help here and disable the test on OSX for now (I think having
the
> > test disabled would be better than it not correctly testing the behavior).
> 
> As you've taken code out of the platform independent bit, I assume you need to
> add code into the corresponding mac bit.
> 
> Why did you remove the code? The CL description doesn't explain (it just
> mentions the test).

The test was failing on OSX when I added the wait-for-close to the test (and no
other change). I assumed that the subtraction was the failure so I moved it to
every platform except OSX, but apparently that was not the problem.

[scheib, 2013-06-05 20:49:04]

stevenjb: Is this change still active, or should it be closed?

[stevenjb, 2013-06-05 21:05:50]

On 2013/06/05 20:49:04, scheib wrote:
> stevenjb: Is this change still active, or should it be closed?

It needs to be fixed on OSX. I haven't had a chance to follow up on that.