Dean Brooks wrote:
> On Wed, Feb 14, 2007 at 01:11:55PM +0100, Magnus Holmgren wrote:
>> On Wednesday 14 February 2007 12:51, Renaud Allard wrote:
>>> This is a _new_ mail sent from the internal server (192.168.100.12) from
>>> fpouchain@??? to chardon@??? using exim 4.66 as a mail
>>> relay between the internal server and the internet.
>>>
>>> As we can see, a first attempt is made on the primary MX: SMTP1, it gets
>>> a 450 error for greylisting. Quite obviously exim retries to SMTP2 and
>>> also gets a 450 error. But exim never retries and interprets this as a
>>> permanent error.
>> This may be another instance of the address_retry_include_sender bug,
>> introduded in 4.64 and fixed in CVS:
>>
>> PH/19 Change 4.64/PH/36 introduced a bug: when address_retry_include_sender
>> was true (the default) a successful delivery failed to delete the retry
>> item, thus causing premature timeout of the address. The bug is now
>> fixed.
>
> It this fix something the rest of us should be concerned about?
>
> That is, is this considered an emergency hotfix we should all install,
> or does it only affect unusual configurations?
>
For my part, it only happens in a very peculiar situation where a server
never appears to be up because there are two MX using greylisting in a
very stupid way (never whitelisting even after one mail passes). I still
think this is a serious bug (just because interpreting 4xx errors as 5xx
is something weird) but you will only see drawbacks on very special
cases where the remote servers are really strangely configured.
