On 2017-02-28 6:19 PM, Jeremy Harris wrote: > On 28/02/17 23:08, Daryl Richards via Exim-users wrote:
>> On 2017-02-28 5:58 PM, Jeremy Harris wrote:
>>> On 28/02/17 22:44, Daryl Richards via Exim-users wrote:
>>>> On 2017-02-28 5:16 PM, Daryl Richards via Exim-users wrote:
>>>>
>>>> I got a better one...
>>>> #17 0x000000000049cf60 in cutthrough_headers_send ()
>>>> No symbol table info available.
>>>> #18 0x000000000049cf60 in cutthrough_headers_send ()
>>>> No symbol table info available.
>>>> #19 0x000000000049cf60 in cutthrough_headers_send ()
>>> That's a bit different...
>>> At least we could guess that disabling cutthrough will
>>> be a likely workaround.
>> I saw that as well, and disabling cutthrough doesn't stop the crashes.
> Hmm. I'll trust it more with a stacktrace from a debug build. Okay, yes... Disabling cutthrough delivery seems to have stopped the
constant crashing. Thinking that the extra failures I saw were still
running delivery processes, I looked back through the logs more. Seems
sometimes these messages come in, sit around for a few seconds to
minutes and then start dumping these repeated errors into the log. With
the cutthrough delivery disabled, I'm still seeing the occasional
SSL_write, broken pipe error, but only one line and it looks like the
email is then processed through, although via a normal forward.
Is it possible when it's a cutthrough it gets stuck on some kind of loop?
