Why is it so difficult to get a Lisp backtrace?

[Alan Mackenzie <acm@muc.de>]

Hello, Emacs.

(setq debug-on-error t) ought to be useful to get backtraces.  It is,
sometimes, but far from always.

There seem to be three use cases for debug-on-error:
(i) The vast bulk of users, who are barely, if at all, aware of its
  existence.  They will enable it on request from a maintainer to help
  debug a bug.
(ii) Hard-core maintainers, who run with it enabled constantly.  Here
  debug-ignored-errors helps suppress certain unhelpful errors which
  would otherwise occur too often and irritate.
(iii) Those maintainers who run with it disabled, but when an error
  occurs, want to be able to repeat that error and get a backtrace.

This third group of users is poorly catered for by the current collection
of mechanisms.  See also bug #56201, with thanks to Andreas and Lars who
helped me get to the bottom of it.  To be reasonably sure of getting a
backtrace, it seems one needs to do all of the following:
(i) (setq debug-on-error t).
(ii) (setq debug-on-signal t).
(iii) Bind debug-ignored-errors to nil.
(iv) Pray.
(v) Execute the erring command again.

I put it to people that we need a new command or variable not called
something like `backtrace-next-and-I-really-mean-it' which would allow
step (v) to generate the desired backtrace without having to go through
steps (i) to (iv) individually.  The lack of such a command/variable is
making (at least) my Emacs development less pleasant than it might
otherwise be.

-- 
Alan Mackenzie (Nuremberg, Germany).