[DGD] error context frames versus stack frames

[Carter Cheng]

Thanks Felix and Raymond as well. I probably need to do some more browsing
of the dgd source code to see if I can understand how best to implement
this using llvm.

On Fri, Apr 14, 2017 at 2:34 AM, Raymond Jennings <shentino at gmail.com>
wrote:

> Just remember, function calls, catches, and rlimits can all nest
> separately.
>
> Each function can itself have zero or more each of catch and rlimits
> blocks nested pretty much arbitrarily.
>
> catch {
>    rlimits (50; 50000) {
>       rlmits (20; 200) {
>           catch(error("Foobar"));
>           error("Baz");
>       }
>    }
> }
>
> Also, don't forget, that if you're in a catch handler (or anywhere
> else for that matter) inside a tick limiting rlimits, and you are out
> of ticks, you will immediately encounter a fresh "Out of ticks" error,
> even in a catch handler.
>
> On Thu, Apr 13, 2017 at 11:01 AM, Felix A. Croes <felix at dworkin.nl> wrote:
> > Carter Cheng <cartercheng at gmail.com> wrote:
> >
> >> Is this basically used to locate the correct exception frame on the
> stack
> >> corresponding to the function activation? In something like C++ one has
> to
> >> unwind until one reaches a handler if it exists. I suspect in dgd's
> version
> >> of lpc one has two unwind until one gets to the "function activation"
> that
> >> originates the catch/rlimit blocks the resume execution after the
> >> catch/rlimit
> >
> > Yes, there is an exception stack and a rlimits stack, both independent
> > of the function stack.  When handling an exception DGD actually takes
> > the exception stack as its lead, and unwinds function calls and rlimits
> > contexts as needed.
> >
> > Regards,
> > Felix Croes
> > ____________________________________________
> > https://mail.dworkin.nl/mailman/listinfo/dgd
> ____________________________________________
> https://mail.dworkin.nl/mailman/listinfo/dgd