Bryan Donlan wrote:
> Are you referring to the "Unifying events and threads" paper?
Yes, but more specifically to the addendum to the thesis proposal.
> I'll
> take a look a bit later tonight. Is there a recommended place to
> contact this GHC HQ other than this cvs-ghc list?
No, I just wanted to make sure you had a more complete context. It's
not much fun to start into a hacking project only to find that someone
else is halfway through something similar.
> I don't see why it won't work; if epoll_create fails with ENOSYS, we
> can just switch over to the select() loop.
If you've successfully linked GHC against a glibc that has bindings for
epoll_create, then you're going to be running on a kernel that has
epoll_create. You won't be able to pick that binary up and run it on an
older glibc that is missing epoll_create bindings, and if you were to
try such a thing, a glibc built against a modern kernel is not likely to
run successfully on a really old kernel. In either case, the
probability of getting a program to run successfully at all is extremely
low, and the number of people who would in any case benefit from the
extra complexity and indirection is about zero.
> That said, I suppose kernels older than 2.5.44 are rather deprecated
> now, so it might not be necessary.
Indeed.
> I don't know whether other
> platforms than linux might need this sort of support though.
None does, as far as I know.
<b
_______________________________________________
Cvs-ghc mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/cvs-ghc
