Re: [lmi] PATCH: Build fixes after unwind.cpp addition

[Greg Chicares]

On 3/5/21 5:05 PM, Vadim Zeitlin wrote:

[...wxStackWalker is not a "silver bullet" solution for lmi...]

>  So maybe I should provide an alternative version of unwind.cpp for MSW?
> This would, of course, be much more work (I'd have to redo everything that
> wxStackWalker already does), but looks like a better alternative. What do
> you think?
I agree with this point (and probably all the others that I've
elided for the moment because I have a pressing problem:

  https://git.savannah.nongnu.org/cgit/lmi.git/commit/?h=odd/eraseme

that I need to solve first). For now, let me just ask whether
your yet-unwritten msw unwinder would
 - work with i686 as well as with x86_64
 - require '-fno-omit-frame-pointer'
In the very short term, either of those restrictions would
make this msw unwinder less immediately useful to us, meaning
only that it would be a less urgent priority. But it's
important anyway, and well worth doing, because
 - we can use x86_64 today, even though we still release
   i686 builds, so an x86_64-only unwinder would allow us
   to explore any "who ordered that?" exception (hence, if
   i686 is much extra work, we should skip that architecture);
 - we can hope that '-fno-omit-frame-pointer' will become
   usable with debian's MinGW-w64 package ere long, although
   any hope for native MinGW-w64 binaries seems forlorn.
And, having probably worked on wxStackWalker yourself, you're
much better able than I to write this, so please go right ahead.