Additional fix for Sandboxed WebKit on OS X.

Fix for Sandboxed WebKit on OS X:
* Cache Gestalt's value since it opens files.
* Do a little cleanup in sys_info while I'm there and add a unit test.
* Label some methods as not being threadsafe on OSX/POSIX.

[pink (ping after 24hrs), 2009-02-24 14:21:00]

LGTM with comments.

http://codereview.chromium.org/27088/diff/1/4
File base/sys_info_mac.cc (right):

http://codereview.chromium.org/27088/diff/1/4#newcode21
Line 21: // Gestalt can't be called in the sandbox, so we cache it's return
value.
its

http://codereview.chromium.org/27088/diff/1/4#newcode43
Line 43: int32 dummy_major_version;
why bother with the nested scope? They're just ints, there are no side effects
of the additional scoping.

[Mark Mentovai, 2009-02-24 14:31:23]

http://codereview.chromium.org/27088/diff/1/3
File base/sys_info.h (right):

http://codereview.chromium.org/27088/diff/1/3#newcode17
Line 17: // WARNING: On POSIX, this method uses static variables and is not
threadsafe.
Not threadsafe until initialized by having been called once without a race.  You
should make that point clear.  Same below.

http://codereview.chromium.org/27088/diff/1/4
File base/sys_info_mac.cc (right):

http://codereview.chromium.org/27088/diff/1/4#newcode46
Line 46: OperatingSystemVersionNumbers(&dummy_major_version,
You could make OperatingSystemVersionNumbers accept and check for NULL pointers
which would simplify this call, or you could just have a single "int32 dummy"
that you pass for all three parameters.

http://codereview.chromium.org/27088/diff/1/7
File base/sys_info_win.cc (right):

http://codereview.chromium.org/27088/diff/1/7#newcode120
Line 120: *bugfix_version = 0;
info.dwBuildNumber, I'd think.

[jeremy, 2009-02-24 15:11:50]

http://codereview.chromium.org/27088/diff/1/3
File base/sys_info.h (right):

http://codereview.chromium.org/27088/diff/1/3#newcode17
Line 17: // WARNING: On POSIX, this method uses static variables and is not
threadsafe.
On 2009/02/24 14:31:23, Mark Mentovai wrote:
> Not threadsafe until initialized by having been called once without a race. 
You
> should make that point clear.  Same below.

Done.

http://codereview.chromium.org/27088/diff/1/4
File base/sys_info_mac.cc (right):

http://codereview.chromium.org/27088/diff/1/4#newcode21
Line 21: // Gestalt can't be called in the sandbox, so we cache it's return
value.
On 2009/02/24 14:21:00, pink wrote:
> its

Done.

http://codereview.chromium.org/27088/diff/1/4#newcode43
Line 43: int32 dummy_major_version;
On 2009/02/24 14:21:00, pink wrote:
> why bother with the nested scope? They're just ints, there are no side effects
> of the additional scoping.

Done.

http://codereview.chromium.org/27088/diff/1/4#newcode46
Line 46: OperatingSystemVersionNumbers(&dummy_major_version,
On 2009/02/24 14:31:23, Mark Mentovai wrote:
> You could make OperatingSystemVersionNumbers accept and check for NULL
pointers
> which would simplify this call, or you could just have a single "int32 dummy"
> that you pass for all three parameters.

Done.

http://codereview.chromium.org/27088/diff/1/7
File base/sys_info_win.cc (right):

http://codereview.chromium.org/27088/diff/1/7#newcode120
Line 120: *bugfix_version = 0;
This mirrors the current behavior in webkit_glue.cc, I'm not comfortable
changing that.

[pink (ping after 24hrs), 2009-02-24 15:17:12]

lgtm

[Mark Mentovai, 2009-02-24 15:19:45]

Oh, I see.  I thought we were going to throw away the 6 in 10.5.6 because that's
what Firefox does, but now I see that we use it on the Mac because Safari does. 
Whatever.

LGTM, carry on.

http://codereview.chromium.org/27088/diff/1003/1005
File base/sys_info.h (right):

http://codereview.chromium.org/27088/diff/1003/1005#newcode18
Line 18: // until its been initialized by being called once without a race.
it's this time :)

http://codereview.chromium.org/27088/diff/1003/1005#newcode50
Line 50: // until its been initialized by being called once without a race.
ditto