When a QApplication project is compiled and linked on a compatible OS (e.g. RHEL 7.4) it uses version of libxcb library provided by that system. Unfortunately, when starting such an application, the DISPLAY environment variable can still point to a different system, with various X display technologies, e.g. Xvnc, NoMachine, Exceed onDemand, not necessarily compatible with the X version and libraries used on the original system.
This results in a segfault, without any chance to guess whether it is a client application bug or X compatibility issue, e.g.:
There was already a patch discussed preventing such nullptr dereference, but it was abandoned (see
QTBUG-64928 for details):
Here is the same try on such incompatible DISPLAY, but with the patch applied:
Please reconsider patching the 5.12 LTS release with it.
In our case, we have to ship a commercial app without debugging symbols as a binary package to our customers. Therefore, we often have no way to reproduce or debug such problems on-site. Any extra messaging might help identifying the issue, contrary to the generic "Memory fault(coredump)" current behavior.