Re: DWARF-2 EH for MinGW gcj?

[Andrew Haley <aph at redhat dot com>]

Mohan Embar writes:
 > Hi Ranjit,
 > 
 > >One of the approaches could be to see if the Win32 GUI
 > >callback code has frame pointers and then walk the chain
 > >of FPs back up the stack till we find a call frame that
 > >has associated DW2 unwind info. This could involve ugly
 > >code like what you can see in the implementation of
 > >backtrace for SJLJ i386 targets in libgcj.
 > 
 > Interesting. You would think that one of the MinGW folks
 > would have thought of this, though....

My understanding was that you have to use SEH data to unwind frames.
DWARF exception try{} regions don't save registers, and thus the
handlers rely on the unwinder to restore register contents.

 > >Another approach could be to simply declare that MinGW
 > >GCC would not support exception propagation across
 > >foreign function calls and move on to better things in
 > >life. :-/
 > 
 > Nice one :)

I have some sympathy with this argument.

Andrew.