[LLVMdev] Is LLVM expressive enough to represent asynchronous exceptions?

[Sohail Somani]

On 11-06-12 2:54 AM, Rafael Ávila de Espíndola wrote:
> On 11-06-12 12:01 AM, Sohail Somani wrote:
>> Is LLVM expressive enough to represent asynchronous exceptions?
>> ---------------------------------------------------------------
> 
> If you search the list you will find some discussion on better ways to 
> represent EH. The focus has been on C++ for now. You may also want to see
> 
> http://www.nondot.org/sabre/LLVMNotes/ExceptionHandlingChanges.txt

That's encouraging, both people responding pointed to the same thing. So
it's a known problem with a "preferred" solution.

I also prefer the "attach unwind information to a block" method but I'd
do it in a manner that also supports resuming exceptions (so as to
support Windows SEH as well.) I guess that was option 2.

So would it be acceptable to implement something like in the above
linked notes with an eye towards async exceptions? If so, I'll write up
how I'll do it, at least in terms of IR-related changes.

Thanks a lot!