tags 32014 + patch
quit

João Távora <joaotavora@gmail.com> writes:

> The error's backtrace isn't shown even with debug-on-error set to t.

If you (setq debug-ignored-errors nil) first, then the backtrace is

Debugger entered--Lisp error: (text-read-only)
  indent-line-to(7)
  lisp-indent-line()
  funcall-interactively(lisp-indent-line)
  call-interactively(lisp-indent-line record nil)
  command-execute(lisp-indent-line record)
  execute-extended-command(nil "lisp-indent-line" "lisp-indent-line")
  funcall-interactively(execute-extended-command nil "lisp-indent-line" "lisp-indent-line")
  call-interactively(execute-extended-command nil nil)
  command-execute(execute-extended-command)

> Is this the intended behaviour?  In 25.2 the string foo was correctly
> indented back one character, so this seems like a regression.

No, it's an accident.  In lisp-indent-line, I simplified 

	(setq shift-amt (- indent (current-column)))
	(if (zerop shift-amt)
	    nil
	  (delete-region beg (point))
	  (indent-to indent)))

into

    (indent-line-to indent)

but it turns out not be equivalent in this case.  indent-line-to doesn't
respect the prompt's field property.

I propose this for emacs-26: