Fix custom cursors on Mac. Null and 0 turn out not to be opposites....

Fix custom cursors on Mac.  Null and 0 turn out not to be opposites.

I also added an assertion to webcursor_mac.mm that is probably a bit
redundant with the one in AppKit, but this way the assertion is more
visible to someone reading this file.

BUG=16346
TEST=load a page that uses a custom css cursor.
R=pinkerton


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=20327

[pink (ping after 24hrs), 2009-07-09 21:30:30]

LGTM

http://codereview.chromium.org/155320/diff/1/2
File webkit/glue/webcursor_mac.mm (right):

http://codereview.chromium.org/155320/diff/1/2#newcode52
Line 52: DCHECK(!custom_data.empty());
is it worth being over-protective (since this takes out the browser process
hard) and returning nil if this check fails? Can callers handle that? Is it
better than throwing a cocoa exception here in release builds?

[darin (slow to review), 2009-07-09 21:35:45]

http://codereview.chromium.org/155320/diff/1/2
File webkit/glue/webcursor_mac.mm (right):

http://codereview.chromium.org/155320/diff/1/2#newcode52
Line 52: DCHECK(!custom_data.empty());
we should only reach here if webkit tries to send us a custom cursor that has no
data.  that situation shouldn't be possible.

will our crash handler catch the cocoa exception?

return early just gimps the feature silently.  i'd much rather get a crash
report.

[pink (ping after 24hrs), 2009-07-09 21:42:25]

Yeah, trouble is we don't get a crash report. Cocoa exceptions aren't
like C++ ones. They hork up your app and leave you in some random
state where the app barely works. In this particular case, the browser
pegs the CPU in the message pump. Yay.

I'm fine with not being too paranoid, just understand we'll never
catch this out in the field because it'll never generate a crash
report.

On Thu, Jul 9, 2009 at 5:35 PM, <darin@chromium.org> wrote:
>
> http://codereview.chromium.org/155320/diff/1/2
> File webkit/glue/webcursor_mac.mm (right):
>
> http://codereview.chromium.org/155320/diff/1/2#newcode52
> Line 52: DCHECK(!custom_data.empty());
> we should only reach here if webkit tries to send us a custom cursor
> that has no data. =A0that situation shouldn't be possible.
>
> will our crash handler catch the cocoa exception?
>
> return early just gimps the feature silently. =A0i'd much rather get a
> crash report.
>
> http://codereview.chromium.org/155320
>



--=20
Mike Pinkerton
Mac Weenie
pinkerton@google.com

[darin (slow to review), 2009-07-09 22:21:39]

OK, I'm going to change that DCHECK to a CHECK so that we get a crash report. 
Thanks!