bug#16749: More informative error when restoring frameset with killed buffer

bug#16749: More informative error when restoring frameset with killed buffer

[Glenn Morris]

Package: emacs
Version: 24.3.50
Severity: minor

Current trunk on GNU/Linux under X.

emacs -Q
C-x C-f 1
C-x r f a
C-x k 1 RET
C-x r j a
   -> jump-to-register: Wrong type argument: stringp, nil

I see that the commentary of frameset.el says it does not restore killed
buffers. That's fine, but a more informative error message would be
better ("Buffer 1 has been killed" or somesuch).

bug#16749: More informative error when restoring frameset with killed buffer

[Glenn Morris]

PS Perhaps it should not stop with an error on encountering a dead
buffer, but should continue restoring as much as it can, then give a
summary message/error about dead buffer(s) at the end.

bug#16749: More informative error when restoring frameset with killed buffer

[Andreas Röhler]

Am 14.02.2014 08:01, schrieb Glenn Morris:
>
> PS Perhaps it should not stop with an error on encountering a dead
> buffer, but should continue restoring as much as it can,

Suggest an ask before: maybe the killing was deliberately but the jump-key a mistake.

then give a
> summary message/error about dead buffer(s) at the end.
>
>
>
>

bug#16749: More informative error when restoring frameset with killed buffer

[Juanma Barranquero]

On Fri, Feb 14, 2014 at 7:52 AM, Glenn Morris <rgm@gnu.org> wrote:

> I see that the commentary of frameset.el says it does not restore killed
> buffers.

That comment refers to window-state-put's functionality, but this bug
is just that I forgot to check that the buffer was live before trying
to restore the point. Fixed now.

> That's fine, but a more informative error message would be
> better ("Buffer 1 has been killed" or somesuch).

At that point, the saved state does not contain the buffer name, just
a marker that points nowhere, so we cannot say "Buffer XXX was killed"
because we don't know that it was called XXX. (Well, strictly speaking
the buffer name is probably somewhere inside frameset-states, but I
don't think digging that info is worth the effort.)

> PS Perhaps it should not stop with an error on encountering a dead
> buffer, but should continue restoring as much as it can, then give a
> summary message/error about dead buffer(s) at the end.

It already restores as much as it can, because that's what
window-state-put does. Dead buffers are ignored and their windows get
another buffer. The current bug does not happen when restoring the
window/buffer, just the selected buffer and point at the end of
jump-to-register.

Also, when using frame-configuration-to-register, jumping to a saved
frame configuration for a deleted frame fails silently. I don't think
jump-to-frameset should be noisier.

   J