roystgnr

> Maybe create a new function like reinit that can be used by solve functions. IMHO this may be the right thing to do medium-term. In hindsight the right thing...

That's definitely an actual failure in misc_ex5, but I'm still trying to replicate it.

Never mind; it's easy to replicate in parallel.

And easy to fix; it wasn't an actual failure, just an assertion getting called in parallel that should have only been called on proc 0.

That MOOSE thing is python acting up again, but the complex configuration's an actual bug in the new unit tests.

Well, that was scary. After I restored the regular test instead of the debugging test, it started failing on Hexes. Turned out to be a bug in the test rather...

It looks like `SYSTEM_INFO` is a Win32 API struct, defined in sysinfoapi.h? You might try adding an include for that to memio.c. But if you're hitting this problem at all...

Okay - the problem is now fixed for TypeVector and Point, but it's not fixed in general: the C++ standard doesn't mandate that `is_trivially_copyable::value == true`, and that's too important...

Perhaps I should give it a different name? "MemCopyable"? But this just seems like a question of vocabulary - that link goes on to say, "Even if it knew that...

Exactly. I really do want to scream if someone ends up trying to memcpy() into an object for which just copying bytes is not a valid way to copy an...