Memleak in the 4K buffers in 1.2.22?

--MimeMultipartBoundary
Content-Type: text/plain; charset=us-ascii

At

 http://www.cuci.nl/~srb/memleak.html

you'll find a memory usage statistic of a Squid 1.2.22 that has been
running for 29 hours:

Connection information for squid:
        Number of HTTP requests received: 541848
        Number of ICP messages received: 176561
        Number of ICP messages sent: 197833
        Number of queued ICP replies: 45
        Request failure ratio: 0.00%
        HTTP requests per minute: 308.0
        ICP messages per minute: 212.8
        Select loop called: 63558616 times, 1.661 ms avg
File descriptor usage for squid:
        Maximum number of file descriptors: 1024
        Largest file desc currently in use: 245
        Number of file desc currently in use: 145
        Available number of file descriptors: 879
        Reserved number of file descriptors: 100
Internal Data Structures:
        631173 StoreEntries
          3077 StoreEntries with MemObjects
          3075 StoreEntries with MemObject Data
           133 storeSwapLogData temporaries
          2994 Hot Object Cache Items
        631061 Filemap bits set
        631044 on-disk objects

Now, on the memory statistics page, you see that the 4K-byte buffer blocks
already occupy 40MB (and rising). It very much looks like a memory leak.
The leak seems to be leaking at a much higher rate now than in the preceding
period of five days (which was the same squid). The only difference between
now and then is that the disk-empty/full distributions were different.

Any ideas?

-- 
Sincerely,                                                          srb@cuci.nl
           Stephen R. van den Berg (AKA BuGless).
"Good moaning!"
--MimeMultipartBoundary--