[PATCH] D133376: Allow epilogue_begin to be emitted when generating DWARF

[John Reagan via Phabricator via llvm-commits]

JohnReagan added a comment.

In D133376#3824277 <https://reviews.llvm.org/D133376#3824277>, @dblaikie wrote:

> In D133376#3822690 <https://reviews.llvm.org/D133376#3822690>, @JohnReagan wrote:
>
>> Thanks for this.  On OpenVMS, our calling convention wants to know about both prologue and epilogue sections.  For asych-error handling, we won't call a routine's handler (ie, personality-routine) if it is in its prologue or epilogue since the routine's frame is no longer active.  In those cases, we walk back up the stack one frame to begin our handler search.  Today, we just look at the instruction sequences and trying to guess as best we can.  Having something in the debug_line table will be a big help.
>
> that seems concerning - sounds like the correctness/behavior of the platform depends on the presence/contents of .debug_line? That doesn't sound right. I thought anything relevant to that should be encoded in .debug_frame/.eh_frame, no?

Sorry for the delay, been on the road and only catching up.  Yes my bad, we have that in eh_frame, not in the debug_line.  The weakness of epilogue support has always been a concern of mine and I'm happy that things are catching up.  The debug_line indicators of epilogue will help our debugger and performance analysis tools.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D133376/new/

https://reviews.llvm.org/D133376