FWIW I've created a bug report. http://bugs.squid-cache.org/show_bug.cgi?id=3731 So far the patch I mentioned has worked beautifully on my production machines. On Fri, Dec 28, 2012 at 6:59 PM, Amos Jeffries <squid3@xxxxxxxxxxxxx> wrote: > On 29/12/2012 3:49 p.m., Andrew Beverley wrote: >> >> On Fri, 2012-12-28 at 18:13 -0800, Nick Rogers wrote: >>> >>> I was able to come up with a patch that works in my environment by >>> looking at some of the changes between 3.1 and 3.2. It seems that >>> sizeof(tos_t) does not result in a valid setsockopt() argument, >>> whereas sizeof(int) that was used in 3.1.x does. >> >> Thanks for that. I vaguely remember making that change when creating the >> v3.2 patch set, probably for no reason other than tidying up the code. >> I'll have a closer look in the next couple of weeks, unless anyone else >> does before then. > > > This is a bit of a strange error. Since the tos_t type is an 8-bit char. > Passing in a larger size than the object being passed means some garbage off > the stack will be passed to the setsockopt() internals. > > Amos >
