Re: iACS ODBC Driver on Linux: Segfault in libcwbcore.so

Hello Richard,

Am 22.09.2025 um 19:20 schrieb Richard Schoen <richard@xxxxxxxxxxxxxxxxx>:

Not knowing your app I wonder if It would be worthwhile creating a REST service on IBMi to communicate with if the Linux driver is not fully stabilized.

There's also the new MapePire WebSockets services that Jesse/Liam created.

The dev has had no former exposure to IBM i. Learning from almost zero would have been too long of a process. Also, he's very fond of C#/.NET. I'm not in the position to argue with him. :-)

We already sank hours and hours of debugging time into the issue for what we perceived being a simple application. Honestly: We're fed up. It works, and that's it. Since the latest change, there was no crash anymore.

Perhaps the background thread processing could be separated into a separate Linux service that actually spawns a new process for each run. (Depending on volume of course).

That would have been the next step.

Another thing they could do is potentially move the app to a Windows server. Perhaps the Windows driver is less prone to issues since Linux relies on unixODBC.

This is a "no no" from the big boss. He came to like Linux as being sleek and rather unproblematic in terms of automation, updates, "call home", etc., compared to the mess Microsoft produces lately.

It's always fun to try and troubleshoot these types of issues.

Did you miss putting quotes around the word fun? ;-)

However it's making me more hesitant to roll out IBMi OBDC based services on Linux.

Well, I've used ODBC privately and extensively for many years with only one (also mysterious) problem:

https://archive.midrange.com/midrange-l/202108/msg00502.html

Since accessing my 150 with Perl works fine since I started doing so ca. 15 years ago, — and I'm doing things with Perl anyway —, I'm good. Note: The described error with the uneven memory access is not apparent with 7.2 and newer!

Over the years, I have written some data movers in Perl, accessing a 7.2 LPAR which also worked for years, without issue. There is a non-zero chance that this error is only showing up in this particular case: With .NET, maybe only on Linux. I don't know for sure, but evidence suggests it's not a generic issue.

One of the things in regular use is my Cisco device management. https://github.com/PoC-dev/cisco-erfassung

REST services may be better suited for this type of environment so there are fewer layers on the Linux server.

Well, you move some layers to IBM i, but the overall layering is still required.

Thanks for reporting back.

You're welcome!

:wq! PoC