Summary

As part of WT-8198, I've been trying to figure out why a pointer is null. Unfortunately, none of the spawn hosts have a core dump available because the UBSAN flags use `halt_on_error` (for stopping the program) instead of `abort_on_error` (for generating a core dump), and manually running `format.sh` seems to only very rarely hit the issue. I've been running this test on all cores on two machines for an entire day and never saw it.

The goal is that, by allowing a core dump, we can diagnose this properly.

Motivation

Does this affect any team outside of WT?

No

Is this issue urgent?

About as urgent as WT-8198.

Acceptance Criteria (Definition of Done)

When UBSAN failures generate core dumps (instead of immediately halting).

Testing

N/A

Documentation update

No