Add regression test for bug 7900 and necessary infrastructure....

Add regression test for bug 7900 and necessary infrastructure.

The test simulates following scenario: user opens a local html page with Chrome when a browser instance is already running. The new Chrome process should notify the existing process about new page and exit. Bug 7900 broke this behavior causing the new process to crash.

I verified that the ui_tests infrastructure would catch this crash and cause the test to fail.

BUG=7900, 8056

[Pam (message me for reviews), 2009-03-06 19:51:05]

Please expand the checkin comment (description) so it's self-explanatory.
Someone looking at 'svn log' shouldn't have to refer to a bug to see what a
patch was for.

Looks reasonable apart from that, but it's not my area of expertise. Someone
else -- Nicolas? -- should take a look too.

[Nicolas Sylvain, 2009-03-06 20:50:45]

http://codereview.chromium.org/39257/diff/1/4
File chrome/app/chrome_main_uitest.cc (right):

http://codereview.chromium.org/39257/diff/1/4#newcode66
Line 66: net::FilePathToFileURL(test_file_path), action_timeout_ms()));
did we understand why it's creating a window instead of a tab? My testing were
not really useful. I'd like to understand the problem.

And if you feel like it, the test just above (SecondLaunch) had the same problem
IIRC, so you might want to do the samething there and reenable it. (That might
need refactoring a little bit your code though). If you don't want it's ok too.

But I would still like to understand the problem..

Other than that the code LGTM.

thanks

Nicolas

[Paweł Hajdan Jr., 2009-03-06 22:34:21]

Pam: I updated the commit message. It's still not perfect, but contains more
information. If you have ideas how to improve it I'll gladly do so.

On 2009/03/06 20:50:45, Nicolas Sylvain wrote:
> http://codereview.chromium.org/39257/diff/1/4
> File chrome/app/chrome_main_uitest.cc (right):
> 
> http://codereview.chromium.org/39257/diff/1/4#newcode66
> Line 66: net::FilePathToFileURL(test_file_path), action_timeout_ms()));
> did we understand why it's creating a window instead of a tab? My testing were
> not really useful. I'd like to understand the problem.

What I know now is that there is some problem when the screen is locked and you
try to find existing Chrome window. That's why the results were different on my
machine and on the trybot. I managed to reproduce the failure on my machine by
launching the tests and quickly locking the screen.

So I would say that "why it's creating a window and not a tab" is still
unanswered. I also think it's worth investigating at some point, but it's not
very easy.

About SecondLaunch: I will look, but eventually in a separate changeset.

[Pam (message me for reviews), 2009-03-06 23:22:52]

On 2009/03/06 22:34:21, Paweł Hajdan Jr. wrote:
> Pam: I updated the commit message. It's still not perfect, but contains more
> information. If you have ideas how to improve it I'll gladly do so.

Even something like "Add a test to make sure that opening a local HTML page
while the browser is already running doesn't crash the new process" would be
sufficient for me. This is plenty good, thanks.