GTTF: Make ui_tests more reliable by using a temporary profile directory

GTTF: Make ui_tests more reliable by using a temporary profile directory
for each test.

TEST=ui_tests
BUG=none

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=53202

[Evan Martin, 2010-07-20 20:33:28]

+some process singleton people

[Evan Martin, 2010-07-20 20:53:12]

Can you make the change description say more?  It made sense after I read the
change, but not before.  Include something like: "previously, we attempted to
always use the same profile directory, cleaning it out between each run.  now we
just use a new temporary profile directory on each run"

http://codereview.chromium.org/2856047/diff/1/2
File chrome/browser/process_singleton_linux_uitest.cc (right):

http://codereview.chromium.org/2856047/diff/1/2#newcode111
chrome/browser/process_singleton_linux_uitest.cc:111: NotifyOtherProcess(url));
It seems like most of these can fit on one line now.

http://codereview.chromium.org/2856047/diff/1/3
File chrome/test/ui/ui_test.cc (right):

http://codereview.chromium.org/2856047/diff/1/3#newcode1121
chrome/test/ui/ui_test.cc:1121: temp_profile_dir_->path().ToWStringHack());
Sorry about this, I will fix it soon.

http://codereview.chromium.org/2856047/diff/1/3#newcode1225
chrome/test/ui/ui_test.cc:1225:
temp_profile_dir_->path().AppendASCII("Default").AppendASCII("History");
why not user_data_dir()  ?

[Evan Martin, 2010-07-20 20:53:21]

Oops, and LGTM with those nits

[mattm, 2010-07-20 21:03:41]

LGTM too.

[huanr, 2010-07-20 23:49:28]

LG

[Paweł Hajdan Jr., 2010-07-21 00:06:47]

Applied the fixes, and made some changes after trybot failure on Windows (not
tested yet, but should work). Could you take another look?

[Evan Martin, 2010-07-21 00:14:18]

http://codereview.chromium.org/2856047/diff/1/3
File chrome/test/ui/ui_test.cc (right):

http://codereview.chromium.org/2856047/diff/1/3#newcode730
chrome/test/ui/ui_test.cc:730: PathService::Get(chrome::FILE_LOCAL_STATE,
&local_state_path);
Hm, this makes me think of a better approach -- I believe you can force the
USER_DATA_DIR value that PathService returns to a value of your choosing (the
temp dir), and then it will return the proper values for data derived from that
USER_DATA_DIR.  I forget the API but it should be in the PathService header.

[Paweł Hajdan Jr., 2010-07-21 00:39:17]

Good idea, patch updated.

[Evan Martin, 2010-07-21 00:51:06]

LGTM