Experimental Physics and
Industrial Control System

On Jul 7, 2020, at 8:53 AM, Ralph Lange via Core-talk <core-talk at aps.anl.gov> wrote:

In a module (currently against the 3.15 branch on Debian Linux) I have defined a local record type. What I am seeing:

If the first record instance that my IOC loads is of that new record type, the database gets corrupted.

The IOC commands (dbl, dbDumpRecord) show a fake bi record that either has a weird name (arbitrary bytes), shows "(null)" or an empty string. Almost every time, the IOC segfaults while doing the 'dbl' or 'dbDumpRecord'. Stray pointer, obviously.

As soon as the first record being loaded is not an instance of my locally defined type, everything is fine.

I will certainly dive in and find what's causing this, but if someone has seen this before, I would appreciate a heads-up to shorten that procedure.

Not specifically, but do the same problems exist if you stop it before iocInit? Try to determine whether this could be caused by something being corrupted during record initialization or if it’s an issue with loading the record instance. Does it happen if the only record loaded is that one?

That could also imply something is wrong with the recordSizeOffset data, are you sure you completely rebuilt the record.o file after the last change to its record.dbd file?

The iocsh command dba might be helpful, and there are one or two others that delve inside the data structures.

- Andrew