Re: [PATCH] Truncate HTTP response bodies to match clen

Alex Rousskov wrote:
> On 06/28/2009 01:50 PM, Henrik Nordstrom wrote:
>> lör 2009-06-27 klockan 08:39 -0600 skrev Alex Rousskov:
>>
>>> Do you think this change should go in?
>> A strong + from me.
>
> Committed to trunk. Amos, please commit to 3.1 (using trunk) and 3.0
> (using the patch I originally posted as it was based on 3p0-plus).
>
>> For what it's worth Squid-2 was fixed long ago. But it took quite some
>> iterations before all cases were covered.. When this condition is
>> detected it not only discards excess data and closes the connection, it
>> also throws out the current response from the cache as it's malformed.
>
> I agree that we should not cache the truncated response. Unfortunately,
> we cannot use HttpStateData::cacheableReply for that because it is
> called before we know about the body truncation.
>
> Will everything work as expected if we call entry->makePrivate() later
> in the process, when the extra body starts coming in?
>
> Thank you,
>
> Alex.

Just in passing on one of the squid-users question research I noted a
call that seemed to cause the RELEASE to happen for an object. If thats
available it would seem to be the right thing.
We are not making the URL no-cachable only the currently received object
or variant.

Amos

-- 
Please be using
   Current Stable Squid 2.7.STABLE6 or 3.0.STABLE16
   Current Beta Squid 3.1.0.9