Re: Splaytree fixes in 2.2.STABLE4 [UPDATED PATCH]

From: Arjan de Vet <Arjan.deVet@dont-contact.us>
Date: Tue, 27 Jul 1999 22:50:10 +0200

Michael Rommel wrote:

>Hi Arjan and all you out there,
>
>a while ago I sent a bug report and fix to the squid-bugs mailinglist, but
>they didn't show up ion the web and apparently there are more people
>experiencing problems with the splay trees.

Yes indeed.

>Attached you can find the small patch I wrote to make it work for us again.

My customers today got hit by the problem you fixed in your patch
because this problem was still present in my patch too :-(. I've created
a new patch available from

        http://www.iae.nl/users/devet/squid/

and tested it against all known problem cases I have found so far. Let's
hope this patch is 100% correct now...

>I cannot confirm that aclDomainCompare gives out -1 instead of 0 for
>identical strings, what breaks the splay trees is that the two functions
>for CREATING a tree of type dstdomain and the function for searching
>it do not produce the same result on some occasions, just as you pointed
>out. However I haven't had the time to test all of your code to verify
>the correct processing of bla-foo.bar and blub.foo.bar :-(

That's the kind of problem situation I've seen too (it was the cause of
today's problem).

>Arjan, do you have any idea why these bugfixes don't show up on the
>"Reported Bugs" section on nlanr?

There's a fix now for the "6teen.com is a subdomain of 8teen.com"
problem. But is still not a real solution for overlapping entries and it
still doesn't warn in the ip-address/network case.

Arjan
Received on Tue Jul 27 1999 - 14:33:54 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:47:35 MST