Re: squid 1.NOVM.22 bug

squid-users-list@salford.ac.uk writes:

>Squid 1.NOVM.22 dies with:
>
>FATAL: storeUnlockObject: unlocked with pending clients
>
>if I open a large number of sockets (200-400 [varies]) in quick succession,
>with a little
>test program I knocked together, and request a file down each
>one. If I merely open the sockets it's happy. It is the requesting of
>an object down each which causes it to die everytime.
> I tried to pin-point if it was at a particular number of sockets that
>causing it die, but it isn't. Sometimes dies with ~280 opened sometimes
>survives with upto 400 be opened.
> This is really annoying as I can't really justify using squid at
>this site if it can be brought down so easily. Hope someone has an
>idea why this is happending.
> This under a newly cvsupped FreeBSD 2.2.6-RELEASE.

Are you requesting the same URL on every connection? If so...
try this patch:

cvs -q diff -wu store.h
Index: store.h
===================================================================
RCS file: /surf1/CVS/squid/src/Attic/store.h,v
retrieving revision 1.76.2.16
diff -w -u -r1.76.2.16 store.h
--- store.h 1997/12/31 04:52:21 1.76.2.16
+++ store.h 1998/07/20 15:37:30
@@ -229,10 +229,10 @@
     swap_status_t swap_status:3;
     method_t method:4;
 
- /* WARNING: Explicit assummption that fewer than 256
+ /* WARNING: Explicit assummption that fewer than 65536
      * WARNING: clients all hop onto the same object. The code
      * WARNING: doesn't deal with this case. */
- unsigned char lock_count;
+ unsigned short lock_count;
 
 };

Duane W.
Received on Mon Jul 20 1998 - 08:38:56 MDT