Bring up Firefox Password Import Unittest on OS X.

Bring up Firefox Password Import Unittest on OS X.

The NSSDecrptor class needs to be run in a child process on OS X, so we need to do some fancy footwork in the form of a bridge class.

Also a few small fixes for surrounding code.
BUG=18775
TEST=none

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=23971

[jeremy, 2009-08-21 00:40:00]

Tons of files in this CL since I needed to checkin two sets of NSS binaries
(checked with wtc and he's is fine with this) but otherwise there's not that
much to review.

[John Grabowski, 2009-08-21 01:13:18]

LGTM
Only minor comments

http://codereview.chromium.org/174206/diff/62/65
File chrome/browser/importer/firefox_importer_unittest_messages_internal.h
(right):

http://codereview.chromium.org/174206/diff/62/65#newcode4
Line 4: 
Add #ifndef to mirror convention you maintain in
chrome/browser/importer/firefox_importer_unittest_utils.h?

http://codereview.chromium.org/174206/diff/62/66
File chrome/browser/importer/firefox_importer_unittest_utils.h (right):

http://codereview.chromium.org/174206/diff/62/66#newcode5
Line 5: #ifndef CHROME_BROWSER_IMPORTER_UNITTEST_UTILS_H_
CHROME_BROWSER_IMPORTER_UNITTEST_UTILS_H_ -->
CHROME_BROWSER_IMPORTER_FIREFOX_IMPORTER_UNITTEST_UTILS_H_

http://codereview.chromium.org/174206/diff/62/66#newcode19
Line 19: // On OS X NSSDecryptor needs to run in a separate process, to allow us
to use
More comments would be nice.  E.g. brief example of what messages go back and
forth.

http://codereview.chromium.org/174206/diff/62/66#newcode20
Line 20: // the same uni test on all platforms we use a proxy class which spawns
a
uni test --> unit test (and run on sentence)

http://codereview.chromium.org/174206/diff/62/66#newcode27
Line 27: FFUnitTestDecryptorProxy(std::wstring &nss_path, bool* success);
We don't use C++ exceptions so constructors in general aren't expected to fail. 
That's why, for example, base/thread.h doesn't start the thread in Thread(), but
in Start().  Perhaps move the constructor logic which might fail into a
different method which can return a bool.  Init()?

http://codereview.chromium.org/174206/diff/62/67
File chrome/browser/importer/firefox_importer_unittest_utils_mac.cc (right):

http://codereview.chromium.org/174206/diff/62/67#newcode43
Line 43: dyld_override.first = "DYLD_FALLBACK_LIBRARY_PATH";
Add comment explaining use of DYLD_FALLBACK_LIBRARY_PATH

http://codereview.chromium.org/174206/diff/62/67#newcode68
Line 68: // This class relies needs to be called on a single thread.
remove "relies"

http://codereview.chromium.org/174206/diff/62/67#newcode117
Line 117: IPC::Message::Sender* sender_;
// weak

http://codereview.chromium.org/174206/diff/62/67#newcode164
Line 164: const int64 kLoopTimeoutMS = 3 * 1000;  // 3 seconds.
Jack up to help prevent flaky tests

http://codereview.chromium.org/174206/diff/62/68
File chrome/browser/importer/firefox_importer_utils.h (right):

http://codereview.chromium.org/174206/diff/62/68#newcode35
Line 35: bool GetFirefoxDylibPath(FilePath *ff_dylib_dir);
Why return bool and not return FilePath?
FilePath::empty() to see if it is valid.

http://codereview.chromium.org/174206/diff/62/69
File chrome/browser/importer/firefox_importer_utils_mac.mm (right):

http://codereview.chromium.org/174206/diff/62/69#newcode48
Line 48: ff_library_path = [ff_library_path
[[NSBundle executablePath] stringByDeletingLastPathComponent]

[kuchhal, 2009-08-21 16:24:57]

lgtm

http://codereview.chromium.org/174206/diff/62/67
File chrome/browser/importer/firefox_importer_unittest_utils_mac.cc (right):

http://codereview.chromium.org/174206/diff/62/67#newcode35
Line 35: IPC::Channel& channel, base::ProcessHandle* handle) {
make nss_path const and may be channel too.

http://codereview.chromium.org/174206/diff/62/67#newcode167
Line 167: // wire and the client to reply.  If the client does relpy by a given
timeout
does *not reply*

http://codereview.chromium.org/174206/diff/62/71
File chrome/browser/importer/nss_decryptor_mac.mm (right):

http://codereview.chromium.org/174206/diff/62/71#newcode8
Line 8: #include </usr/include/mach-o/dyld.h>
is that the right convention? not just mach-o/dyld.h?

http://codereview.chromium.org/174206/diff/62/71#newcode24
Line 24: // To do s, the process this function runs in must have the
To do so