Re: [libunwind] Unwinding using the context passed to a signal handler (doesn't work)

[David Mosberger]

>>>>> On Tue, 30 Mar 2004 19:33:16 -0800, "Young, Mark" <address@hidden> said:

  Mark> Sorry if I seem to be flogging a dead horse...

No, that's quite alright.  Open discussions is what this mailing list
is for! ;-)

  Mark> I think the critical knowledge of whether or not a sigcontext
  Mark> can be used to initialize an unwind cursor should be hidden in
  Mark> libunwind, and I would like to see libunwind provide a
  Mark> portable interface for signal handlers to initialize an unwind
  Mark> cursor positioned at the signal frame.

It seems we agree that such a facility wouldn't qualify as being
admitted into the libunwind API on the grounds of being a primitive,
since it's obviously possible to implement it portably with the
existing libunwind API.

So, the only other reason that would justify its addition would be if
there were a demonstrated and overwhelming performance advantage to
doing so and that the performance delta could not be fixed by
improving the existing libunwind implementation.

I certainly agree that there could be _some_ performance advantage to
unwinding directly from the signal context (where this is possible),
but I'm not sure the difference is significant enough to complicate
the API.  If you have numbers to the contrary, I'd certainly be
interested in seeing them.

        --david