This moves log output for ChromeOS to safer locations.

This moves log output for ChromeOS to safer locations.

For BWSI (incognito) mode, we want to make sure that we don't leak
private information between sessions, so we need to have chrome write
log output to the tmpfs created for that mode.

Also, the chrome log in "logged in" mode should reside in the "log"
subdir instead of in the user data dir.

BUG=http://crosbug.com/6908
TEST=Ran on device in incognito and regular modes, and watched where
log output ended up.

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=64638

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=64826

[DaveMoore, 2010-11-01 18:08:29]

http://codereview.chromium.org/4194005/diff/9001/10005
File chrome/common/logging_chrome.cc (right):

http://codereview.chromium.org/4194005/diff/9001/10005#newcode157
chrome/common/logging_chrome.cc:157: if (env->GetVar(env_vars::kSessionLogDir,
&log_dir_str) &&
It surprises me that environment variables would superceed command line
arguments. Can't we just pass different command line args?

[Greg Spencer (Chromium), 2010-11-01 18:16:41]

http://codereview.chromium.org/4194005/diff/9001/10005
File chrome/common/logging_chrome.cc (right):

http://codereview.chromium.org/4194005/diff/9001/10005#newcode157
chrome/common/logging_chrome.cc:157: if (env->GetVar(env_vars::kSessionLogDir,
&log_dir_str) &&
On 2010/11/01 18:08:29, davemoore wrote:
> It surprises me that environment variables would superceed command line
> arguments. Can't we just pass different command line args?

We could... Currently however, the Chrome log file is passed only as an
environment variable (CHROME_LOG_FILE), and I was following that example for
consistency.  If we add a command line flag for the session log file, we should
also add one for the chrome system log file.

In the case here, a command line arg isn't being superseded by the environment
variable, I'm just building a credible fallback using other information (which
happens to come from the command line, but the option is not related to
logging).

[DaveMoore, 2010-11-01 18:22:00]

Ok. LGTM

[Greg Spencer (Chromium), 2010-11-01 22:59:59]

Dave, can you take a quick look at this again?  I removed the DCHECK to make
sure that the log dir exists, and added some code to keep the number of warnings
down in the test code.

Alternatively, I could just remove the warnings, but they seemed like something
we'd want to know on a running system.

http://codereview.chromium.org/4194005/diff/21001/22007
File chrome/test/in_process_browser_test.cc (right):

http://codereview.chromium.org/4194005/diff/21001/22007#newcode195
chrome/test/in_process_browser_test.cc:195: FilePath log_dir =
logging::GetSessionLogFile(*command_line).DirName();
This is only to keep the number of warnings down during tests.

[DaveMoore, 2010-11-02 17:50:53]

LGTM.

Note: For the next release we are going to go back and clean up the whole
chromeos / logging story. It's gotten totally out of hand and is way too
complicated.

[Greg Spencer (Chromium), 2010-11-02 18:02:36]

On Tue, Nov 2, 2010 at 10:50 AM, <davemoore@chromium.org> wrote:

> LGTM.
>
> Note: For the next release we are going to go back and clean up the whole
> chromeos / logging story. It's gotten totally out of hand and is way too
> complicated.


Completely agreed.  I filed a bug to that effect already (
http://code.google.com/p/chromium/issues/detail?id=61143)