DONT_ACCESS_GBLS_AFTER_EPILOGUE

[Mike Stump]

> Date: Sun, 13 Feb 2000 12:27:17 -0700
> From: Jeffrey A Law <law@cygnus.com>

> MRS didn't really understand the implications of trying to run the
> epilogue multiple times to pop frames for exception handling.  That
> scheme was fundamentally flawed

I beg to differ.  :-)  I knew how hard it would be.  In the end, I
would have have the whole epilogue and prologue generated machine
independently, with the md file describing the details of the epilogue
and prologue.  The EH subsystem would then use this information.

Also, I suspect the fatally flawed part comes from the hidden PIC
stuff.  My solution would be to expose more of the details of PIC to
the compiler.  I don't think that that would be a bad idea.

In 10 or 15 years, we can come back, and see if that isn't the
direction that we decided to take.  I predict it will be.  If I
continued doing what I wanted, we just would have been there sooner,
rather than later.

I think the way that epilogues and prologues are done could be
improved.