re: https://github.com/chaincodelabs/libmultiprocess/pull/133#discussion_r1928300799
As a local fixup, this variable shouldn’t pollute the global namespace.
Maybe unset(FOUND_LIBATOMIC) after find_dependency(CapnProto)?
You would know more than me, but it seems like that change could make things worse. If we just unset the variable, wouldn’t that just be overwriting it as unset instead of true and polluting the namespace in a different way?
I think you are right that there may be a less invasive way to work around the debian bug though. Maybe checking to see if the variable is set before setting it, and restoring the previous value after finding capnproto. I would have to refresh my memory of cmake variable scopes before doing this though, and I would want he other code in compat_find.cmake from #119 to be consistent so both workarounds are identical.
Could be a good followup but I think current change is probably the most direct fix for the current issue.