@mflatt: @spall and I are looking at thread/atomic.rkt (since we have atomicity issues in our implementation)

is that a stub for future implementation, or does that in fact force atomicity?

That’s atomic only for the Racket-thread implementation (i.e., not for OS-level threads = Chez threads)

right, so the bug we’re having is atomicity in the equivalent of thread-group-add!

how does the atomic.rkt code force the engines you’ve implemented to be atomic?

In swap-in-thread in “schedule.rkt”, the out-of-ticks handler just resumes the same engine if it’s in atomic mode

ok, thanks, that’s helpful