> Weird. That was totally contrary to the way I remember it working. In
> trying to figure out why I remembered incorrectly, I realized the
> reason why you (in general) want to resolve the path to a canonical
> path is that you can seriously break include_once and require_once if
> you dont. Otherwise require_once wont correctly work if you do
>
> <?
> require_once("/home/george/foo.php");
> require_once("/home/george/../george/foo.php");
> ?>
Right, but guarding against this edge case is costing us a lot on the
normal case.
100,000 realpath() calls on Linux on file /home/rasmus/foo/u2.inc:
real 0m1.286s
user 0m0.190s
sys 0m1.030s
100,000 stat() calls on Linux on file /home/rasmus/foo/u2.inc:
real 0m0.421s
user 0m0.070s
sys 0m0.290s
The deeper the directory hierachy, the larger the difference.
On FreeBSD we get slammed! Looks like they don't have the same
dcache-like thing that Linux has.
100,000 realpath() calls on /home/rasmus/foo/u2.inc on FreeBSD:
real 0m51.274s
user 0m6.752s
sys 0m26.478s
100,000 stats() calls on the same file on FreeBSD:
real 0m8.873s
user 0m0.165s
sys 0m8.042s
So yes, it does look like we don't need to worry too much on Linux, but
clearly on FreeBSD avoiding stat() calls and calls that create cascading
stat()'s are to be avoided at all costs.
-Rasmus
--
PHP Development Mailing List <http://www.php.net/>
To unsubscribe, visit: http://www.php.net/unsub.php
