In <list-14390617@2rosenthals.com>, on 10/08/25
at 12:07 PM, "Doug Bissett" <ecs-isp@2rosenthals.com> said:
Hi Doug,
>>Offhand, it looks like weasel ran out of heap memory and weasel did not
>>catch the allocation failure and the resulting null pointer. One
>>interesting feature of the modula-2 runtime is that the heap size is fixed
>>when the binaries are built. This means that weasel can run out of heap
>>space while there are still large amounts of OS/2 address space available.
>Since I am the only user, I would find that theory to be unlikely.
I understand that you are the only user. However, threads use address
space, even if they are doing nothing. A thread that is processing a
message is going to use more, but that's expected.
With the addition of DKIM and TLS support the per thread memory usage has
increased a bit. A thread count that was OK in older Weasel versions
might be too much under some conditions in 3.08.
IAC, I took another look at the errinfo.$$$ and this changed the results
of my analysis.
Your trap occurred while processing an SMTP AUTH command and it looks like
a defect in the semaphore handling logic. I'll explain what I think I am
seeing to Peter, when he returns.
>I
>should mention that the client, and server, are connected by a gigabit
>LAN, which may have something to do with it.
Unlikely. This does not look like a data rate issue.
>The bad part is, that it did not terminate, so it was not restarted
>automatically.
This is another feature of the Modula-2 runtime. This kind of exception
restarts the failing thread, but this typically is not sufficient get keep
weasel working properly.
back to list