bug#57674: 29.0.50; `backtrace-goto-source' function broken.

[Lars Ingebrigtsen <larsi@gnus.org>]

Michelangelo Rodriguez <michelangelo.rodriguez@gmail.com> writes:

> M-x debug-on-entry <RET> backtrace-goto-source <RET>
> M-x backtrace-goto-source <RET>
> Select a frame.
> Press "s"
> press "c" in the new backtrace.
> It should appears Source code location not known in the echo area.

I can reproduce this problem (but adding the debug-on-entry to
backtrace-goto-source just makes things more confusing).

The problem seems to be that

 (nth (backtrace-get-index) backtrace-frames)

returns a frame without a :source-available entry here:

(defun backtrace-goto-source ()
  "If its location is known, jump to the source code for the frame at point."
  (interactive)
  (let* ((index (or (backtrace-get-index) (user-error "Not in a stack frame")))
         (frame (nth index backtrace-frames))
         (source-available (plist-get (backtrace-frame-flags frame)
                                      :source-available)))
    (unless (and source-available
                 (run-hook-with-args-until-success
                  'backtrace-goto-source-functions))
      (user-error "Source code location not known"))))

I'm pretty unfamiliar with backtrace.el, but I can't see that slot being
added anywhere there?  edebug does add it, but isn't involved in any way
here.  So it looks like something has gotten lost over the years?

However, `RET' on a symbol finds the sounds just fine, which is probably
why nobody has noticed this problem.  I.e., nobody's using the `s'
command in *Backtrace* buffers.