Bug 26180 – If "otherOrigin == this" in SecurityOrigin::equal, it's obviously equal.

• Status: RESOLVED
• Resolution: FIXED
• Component: New Bugs

[Jeremy Orlow]

If "otherOrigin == this" in SecurityOrigin::equal, it's obviously equal.

From http://lists.macosforge.org/pipermail/webkit-dev/2009-June/008038.html

> SecurityOrigin was created to cleanup the way we do same-origin checks in
> the JavaScript bindings, and as such, carry some of that baggage (eg. domain
> relaxing).  The main design constraint then was that it was a cheap compare
> for two origins representing the same document (pointer compare) as that is
> common case.  The code evolved from there.

Makes sense.  Note that SecurityOrigin::equal does not currently do a fast
path check of "this == other" (to just do a pointer compare as you
mentioned) so this is probably low hanging fruit to make things a bit
faster.  (Though, as you said, it's not showing up on current profiles...so
it's probably not a huge deal.)

[Jeremy Orlow]

Created attachment 30937 [details]
A fix.

[Sam Weinig]

Wouldn't it makes more sense to do this in isSameSchemeHostPort?

[Jeremy Orlow]

I don't think so.  I only see isSameSchemeHostPort used a couple places, but they don't seem like they'd benefit from such an optimization [1].  ::equal, on the other hand, is called pretty often (it's used a lot in hash tables).  It seems a bit painful to require another function call plus a branch or two for a function that's not called very often.

We could put the == check in both places.  Doing one additional branch isn't going to hurt the slow path.  It doesn't seem necessary though.

[1] http://www.google.com/codesearch?hl=en&q=isSameSchemeHostPort+package%3A%22git%3A%2F%2Fandroid.git.kernel.org%2Fplatform%2Fexternal%2Fwebkit.git%22

[Darin Adler]

Comment on attachment 30937 [details]
A fix.

Seems fine. r=me

[Brent Fulgham]

Assign for landing.

[Brent Fulgham]

Landed in @r44420.