On Mon, 9 Mar 2020, Martijn van Duren wrote:

On 3/9/20 8:15 AM, Andi Vajda wrote:

On Mar 8, 2020, at 23:58, Martijn van Duren <opensm...@list.imperialat.at> 
wrote:

I guess not a lot of opensmtpd developers have a NetBSD machine at hand
(I certainly don't). Could you supply us with a backtrace, which most
likely will be needed from the pony process.

Rebuilding with libressl 3.0.2 (instead of openssl 1.1.1d) worked around the 
problem.

So it's safe for me to assume you're fine with applying the workaround
for every release and letting other people do the same?

Speaking for myself, I am fine with building with libressl instead of openssl, yes.

Here is a stacktrace of the crash in the pony process, without debugging symbols:
(what extra flags do I need to pass to configure to get a build with debug
 symbols ? Adding --with-cflags=-g was not enough)

#0  0x00007cb54f76c8cf in ?? ()
#1  0x0000000000430aa7 in resolver_getaddrinfo (hostname=0x7cb5514b4000 "",
    servname=0x8 <error: Cannot access memory at address 0x8>,
    hints=0x7f7fff5009c0, cb=0x7cb550feac48, arg=0x9)
    at ../../smtpd/resolver.c:91
#2  0x0000000000000000 in ?? ()

Andi..

Reply via email to