Re: [Caml-list] Threads and "transaction isolation" in OCaml

[Xavier Leroy <Xavier.Leroy@inria.fr>]

On 2013-08-16 18:07, Markus Mottl wrote:
> I just wondered how much we can rely on current OCaml-semantics where
> context-switches are impossible as long as there are no allocations.
> [...]
> But, surprisingly to me, I did manage to get a failure with byte code
> + POSIX threads when running this example on Linux (Centos 6.4) via
> VMware Fusion (Mac OS X).

Bytecode and native code poll for asynchronous events (context
switches, pending signals, etc) at different program points:

- Bytecode:
    . at every function call
    . at the beginning of every iteration of a while or for loop
    . at the end of "b" in "try b with ...".

- Native:
    . at every allocation in the minor heap.

Your example has a function call in the "transaction", hence the
behavior you observe.

> Anyway, is it considered reasonably future-safe to write code of that
> sort?  I'd rather not have new compiler optimizations, etc., interfere
> with these assumptions in the near future.

There is no short-term plan to change this behavior, but no guarantee
that it will be here forever.  Caveat emptor.

- Xavier Leroy