Re: [squid-users] xcalloc Fatal (squid 3.1.9)

On 11/03/11 07:03, Edward Ting wrote:
> Hi Amos,
>
> I read the Bugzilla thread and found the patch may not work for 3.0
> since 3.1 has changed its implementaion using reader callback. (Also
> mentioned by Alex.)
>
> Any advice how to apply the patch to 3.0?

No sorry, I don't. 3.0 has a great many bugs and is no longer maintained
for operational bugs. Mainly because it is so difficult to port patches
like this from 3.1 latest stable down to the old 3.0 code.

3.1.11 is the most production-suitable version of Squid-3 we have. So
same advice goes to you as to Victor. Please try an upgrade to the
supported stable.

>
> In 3.0 it does reading(true) before makeSpaceAvailable(); If I do
> "maybeMakeSpaceAvailable()" after reading(true), I may be trouble if
> "maybeMakeSpaceAvailable()" fails.
>
> Do you think if I do reading(false) following maybeMakeSpaceAvailble()
> failure may be good enough?

I'm unsure the reading(true) mechanism in 3.0 and 2.x relies on
makeSpaceAvailable() always succeeding. You might have some luck making
the maybe*() version unset it o failure.

Amos

-- 
Please be using
   Current Stable Squid 2.7.STABLE9 or 3.1.11
   Beta testers wanted for 3.2.0.5