Earlier this year, we floating around a spec for extending the compact
unwind format to allow for better asynchronous support for prologue and
epilogues on OpenVMS. We need to be able to properly unwind and find
handlers even when the asynchronous event occurs during the prologue or
Given the good feedback that we received here, we went back and refined
our document. I'll post it to the list this week but I'd also love to
sit and chat with interesting parties at next week's LLVM conference.
I don't think there is a specific table "sign" for EH/unwind support so
in lieu of that, I'll be at the debug table (unless there is "EH/unwind
info" table?). So if you like EH and .cfi directives, some join me.
I'm sure there is more that I need to know and I can use your help.
If interested, I can also highlight the few additions we have over the
default AMD64 ABI document that let us work in our hybrid 32/64-bit
environment like our linker making stub routines that reside in 32-bit
address space so that the address of any routine will fit in 32-bits but
the code (and GOT) can actually reside in a 64-bit address space; and
how we use GOTPCRELs to access static data so it can remain in a 32-bit
address space; etc.