Hi,
OK I will have another look at my code. My concern was why the backtrace
shows a crash so far into libcrypto. I would normally expect something to
indicate "things were going wrong" before that? - i.e. some previous call
returning error etc.
The code is simple - we wrote our own as it has to interface to a
non-standard database and we don't use validation. And yes, I have read and
understand the API documentation :-)
Looks like I will have to try to compile the whole library chain with
debugging in order to try to see what is going on.
The code has no memory leaks - and the messages it fails on are random.
Coredumps are all similar / identical - they all seem to indicate crashes
in SHA1_Init() in libcrypto. I was thinking that was so far away from my
code - it was unlikely to be something I had done? As surely dkim_eom() and
the code around it would have complained if something had been wrong - at
this point I am just along for the ride it would seem (and the coredump).
Jon
On 16 May 2017 at 10:44, A. Schulze <sca_at_andreasschulze.de> wrote:
>
> John Fretby:
>
> Just trying to cut down wasted time chasing down mine vs. libdkim vs.
>> openssl possible issues been a while since I have had to debug anything :(
>>
>
> In general the milter and the library are very stable. My opendkim signer
> processes
> are usually running as long as a machines uptime, and that may be months...
> (same for valdiators). So I guess it's your your code.
>
> you have your own program to sign plain messages, right?
> Why? Are you really unable to use the milter?
> Does it happen on random messages? Are the messages which result in a
> coredump different from other?
> What about the memory usage over time?
>
> I assume, you know (and read + understood) http://opendkim.org/libopendki
> m/index.html
>
> Andreas
>
>
>
Received on Tue May 16 2017 - 10:46:49 PST