Compilation auto-jump bug (was: Info bug)

[Juri Linkov]

>  > @@ -2271,8 +2262,7 @@
>  >    ;; Note that `Info-complete-menu-buffer' could be current already,
>  >    ;; so we want to save point.
>  > -  (save-excursion
>  > -    (set-buffer Info-complete-menu-buffer)
>  > +  (with-current-buffer Info-complete-menu-buffer
>  >      (let ((completion-ignore-case t)
>  >      (case-fold-search t)
>  >      (orignode Info-current-node)
>  >
>  > In theory, these constructs should be equivalent, but they are not.
>
> with-current-buffer is equivalent to
>
> (save-current-buffer
>   (set-buffer...

BTW, there is also the opposite bug: when `compilation-scroll-output' is
`first-error', but `compilation-auto-jump-to-first-error' is nil, when
the current buffer is not the *compilation* buffer, then it doesn't keep
point on the first error.  This means when I switch to the *compilation*
buffer, point is at the beginning of the *compilation* buffer instead of
the position of the first error.  But when the current buffer is the
*compilation* buffer at the time when compilation moves point, then it
keeps the correct position of the first error.  This is due to the code:

(defun compilation-auto-jump (buffer pos)
  (with-current-buffer buffer
    (goto-char pos)
    (if compilation-auto-jump-to-first-error
        (compile-goto-error))))

It doesn't use `save-excursion', so it is strange that it doesn't keep
point on the new position after `goto-char'.

But when `compilation-auto-jump-to-first-error' is t, then
after calling `compile-goto-error' and switching to the source buffer,
point keeps its new position on the first error in the *compilation* buffer.

-- 
Juri Linkov
http://www.jurta.org/emacs/