Hacker News Reader

This sounds exactly like the sort of behavior that should be specified/committed in the form of a unit test.

That would've caused this code change to fail, unless I'm missing something.

No-- the cloning operation didn't preserve the partition information. When I first bisected/found the issue I thought it was as you describe, but it's really just that cloning 0'd it out in the new copy and everything else was proper.

> The field was set to 0 by a function it was passed to previously.

The entire structure was just memset to 0, it's not like the struct was carefully initialised.