On 01/31/2013 03:42 PM, Patrick Walton wrote:
On 1/31/13 3:41 PM, Brian Anderson wrote:
There's also the problem with failure after switching stacks. Right now
there is a guard in the stack switch that catches exceptions thrown by
foreign code and aborts the process, which makes this bogus:
do stackswitch {
fail;
}
We could remove that guard and leave the behavior undefined ...
Could we implement unsafe `catch` to make this work? Then the stack
switching code could catch the failure and abort.
The scheduler probably needs the ability to catch anyway, right?
I think that's not exactly the issue. If we catch the failure we could
be catching a legitimate Rust failure that shouldn't abort. We won't
know if we're catching a Rust failure or a C++ failure. The more I think
about it though the more I think it's not a major obstacle.
_______________________________________________
Rust-dev mailing list
[email protected]
https://mail.mozilla.org/listinfo/rust-dev
