Stefan Winter wrote:
I know 3.0.0-beta1 is the cool new hot topic to work with, but still, I am dealing with a 2.x version which needs to work with SSHA1 and systemd.
Yes...
I can confirm that the SSHA1 double-normalize bug has gone away for me; I keep authenticating some 10.000 users with SSHA1 now, so far without any issues.
That's good.
When I SIGTERM the server - as is required to work with systemd - I still see that while freeing mainconfig, there is a SEGFAULT. It looks like it's still at the same spot as reported earlier: if one has an sql module instance and a custom SQL query for that instance in unlang, then the server tries to "do something" with the custom query as it walks through the vservers; possibly trying to tear down the corresponding sql instance twice or so(?).
I'm not really sure why that's happening. If fixed it in 3.0, but the code is fairly different, and using talloc fixed a lot of problems like this.
See my original message below - and if some more detail is needed, I can work on debugging things in the coming days as it seems to be a calm week :-)
If you have a minimal config which reproduces it, I could take a look. The main thing is to get an interactive gdb session going. Alan DeKok.