On Fri, Mar 14, 2008, Alex Rousskov wrote:
> > I think it actually is a bug in the Vary handling in Squid-3.
> > The condition:
> >
> > if (!has_vary || !entry->mem_obj->vary_headers) {
> > if (vary) {
> > /* Oops... something odd is going on here.. */
> >
> > .. needs to be looked at.
>
> But it is not the condition getting hit according to Aurimas' log, is
> it?
There's two confusing log messages with the same message, which is confusing
things.
Aurimas, in varyEvaluateMatch() in client_side.cc, change the first debugs()
statement here:
if (!has_vary || !entry->mem_obj->vary_headers) {
if (vary) {
/* Oops... something odd is going on here.. */
debugs(33, 1, "varyEvaluateMatch: Oops. Not a Vary object on
second attempt, '" <<
entry->mem_obj->url << "' '" << vary << "'");
safe_free(request->vary_headers);
return VARY_CANCEL;
}
.. replace that "second attempt" with "first attempt", recompile, and see whats
logged.
Adrian
--
- Xenion - http://www.xenion.com.au/ - VPS Hosting - Commercial Squid Support -
- $25/pm entry-level VPSes w/ capped bandwidth charges available in WA -
