Fix CheckFalseTest.CheckFails on Linux after my change to ui_test.

Fix CheckFalseTest.CheckFails on Linux after my change to ui_test.

This makes PathService clear its cache after overriding a path.
We have many paths depending on each other, so this is necessary
to avoid inconsistencies.

TEST=ui_tests in Release mode
BUG=49838
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=59021

[Lei Zhang, 2010-07-22 20:38:57]

Look ok to me with nits below. Deferring to Brett.

http://codereview.chromium.org/2805100/diff/1/2
File base/path_service.cc (right):

http://codereview.chromium.org/2805100/diff/1/2#newcode238
base/path_service.cc:238: // Path providers may rely on the overriden path
value. To be sure we update
spelling police says: s/overriden/overridden/

http://codereview.chromium.org/2805100/diff/1/3
File base/path_service.h (right):

http://codereview.chromium.org/2805100/diff/1/3#newcode47
base/path_service.h:47: // Do not call in production code.
But Override is used in production code in a couple places, i.e.
chrome_dll_main.cc. Rephrase?

[brettw, 2010-07-22 20:39:36]

How about a quick unit test that getting the value, overriding, and then getting
it again gives you the right value? (Or whatever you used to have to do to get
the un-overridden value.)

LGTM with that.

[Paweł Hajdan Jr., 2010-08-06 22:13:12]

Please take another look. This version of the patch doesn't break browser_tests.

[brettw, 2010-08-08 05:21:41]

LGTM

On Fri, Aug 6, 2010 at 3:13 PM,  <phajdan.jr@chromium.org> wrote:
> Please take another look. This version of the patch doesn't break
> browser_tests.
>
> http://codereview.chromium.org/2805100/show
>

[Paweł Hajdan Jr., 2010-08-18 22:55:13]

Please take a look. This revision attempts to fix some weird failures.

[Lei Zhang, 2010-08-18 23:08:12]

http://codereview.chromium.org/2805100/diff/14001/15005
File chrome/test/ui/ui_test.cc (right):

http://codereview.chromium.org/2805100/diff/14001/15005#newcode396
chrome/test/ui/ui_test.cc:396: env->SetVar("XDG_CACHE_HOME",
user_data_dir().value());
Can you override DIR_USER_CACHE instead of relying on the current implementation
of DIR_USER_CACHE?

[Lei Zhang, 2010-08-19 17:43:17]

http://codereview.chromium.org/2805100/diff/14001/15005
File chrome/test/ui/ui_test.cc (right):

http://codereview.chromium.org/2805100/diff/14001/15005#newcode396
chrome/test/ui/ui_test.cc:396: env->SetVar("XDG_CACHE_HOME",
user_data_dir().value());
On 2010/08/18 23:08:13, Lei Zhang wrote:
> Can you override DIR_USER_CACHE instead of relying on the current
implementation
> of DIR_USER_CACHE?

OTOH, if messing with DIR_USER_CACHE is really hard, then just go ahead and go
with what you have here, but put a note here saying this depends on the Linux
implementation of DIR_USER_CACHE.

[Lei Zhang, 2010-08-19 17:47:03]

Oh, and LGTM with the additional comments.

[Paweł Hajdan Jr., 2010-08-19 18:00:11]

Patch updated. I'm assuming LGTM from both of you, but if you have any comments
I will apply them (if you have comments after I commit this I will do a
follow-up).

Or maybe it will get reverted again, ...

[evanm, 2010-11-12 20:54:58]

http://codereview.chromium.org/2805100/diff/28001/chrome/browser/profile_impl.cc
File chrome/browser/profile_impl.cc (left):

http://codereview.chromium.org/2805100/diff/28001/chrome/browser/profile_impl...
chrome/browser/profile_impl.cc:313: if
(!PathService::IsOverridden(chrome::DIR_USER_DATA)) {
FYI, this regressed having separate cache directories for different profiles on
Linux.

That is, even if you pass --user-data-dir=foo, this uses ~/.cache/google-chrome
for all running profiles.  They stomp on each other's caches and cause various
page corruption.

[Paweł Hajdan Jr., 2010-11-13 15:00:17]

On Fri, Nov 12, 2010 at 21:54, <evanm@google.com> wrote:
>
> FYI, this regressed having separate cache directories for different
> profiles on Linux.
>
> That is, even if you pass --user-data-dir=foo, this uses
> ~/.cache/google-chrome for all running profiles.  They stomp on each
> other's caches and cause various page corruption.


Oops indeed. However, I'm not sure how to fix it. If the cache lives in
.cache and the profile in .config, it's not obvious, and using IsOverridden
is hacky and fragile. How about just moving the cache to .config? Or maybe
there's a better solution.

[evanm, 2010-11-13 19:04:14]

On Sat, Nov 13, 2010 at 6:59 AM, Paweł Hajdan, Jr.
<phajdan.jr@chromium.org> wrote:
> Oops indeed. However, I'm not sure how to fix it. If the cache lives in
> .cache and the profile in .config, it's not obvious, and using IsOverridden
> is hacky and fragile. How about just moving the cache to .config? Or maybe
> there's a better solution.

My plan is to move the .config/.cache decision into path service,
which will then have access to the private IsOverridden.
I will CC you on it.  (I didn't want to do it in this change because
it requires refactoring on Windows and Mac too...)

[Paweł Hajdan Jr., 2010-11-13 21:21:16]

On Sat, Nov 13, 2010 at 20:04, Evan Martin <evanm@google.com> wrote:

> On Sat, Nov 13, 2010 at 6:59 AM, Paweł Hajdan, Jr.
> <phajdan.jr@chromium.org> wrote:
> > Oops indeed. However, I'm not sure how to fix it. If the cache lives in
> > .cache and the profile in .config, it's not obvious, and using
> IsOverridden
> > is hacky and fragile. How about just moving the cache to .config? Or
> maybe
> > there's a better solution.
>
> My plan is to move the .config/.cache decision into path service,
> which will then have access to the private IsOverridden.
> I will CC you on it.  (I didn't want to do it in this change because
> it requires refactoring on Windows and Mac too...)
>

SGTM, thanks for working on this!