On Mon, Jun 28, 2004 at 10:38:14PM +0100, Christophe Rhodes wrote:
> In addition, I don't suppose I can persuade you to add tests to your
> patch? There's a BACKTRACE-AS-LIST function lurking around which
> should be adequate to test robust backtraces in error handlers; it
> would be a great help to me as I contemplate examining the sparc
> runtime for these issues to have a set of tests that I know should
> work. :-)
Here is the patch again, with tests in debug.impure.lisp. I believe
that correct ports should pass all of these tests. The caveat is that
x86 currently does not -- it does not produce the "undefined function"
frame, and it sometimes does not have a complete backtrace at all.
I looked at fixing it, but got very confused by the x86 port.