bug#24449: Emacs 25.1 RC2: Byte compiler reports error in wrong place.

[Alan Mackenzie <acm@muc.de>]

Hello, Eli.

On Sat, Sep 17, 2016 at 12:35:05PM +0300, Eli Zaretskii wrote:
> > Date: Sat, 17 Sep 2016 08:29:52 +0000
> > Cc: 24449@debbugs.gnu.org
> > From: Alan Mackenzie <acm@muc.de>

> > 5. The change I made yesterday appears not to have made anything any
> >   worse.

> Thanks, I guess that means you should push it.

Done.  I took the liberty of amending the comment before
`byte-compile-set-symbol-position'.

> > What I would suggest should get done: we should make the column numbers
> > 0-based

> This should be a separate change, and before doing it, we should make
> sure to fix code that assumes the columns to be 1-based (e.g, what
> does "C-x `" do?).

Ah.  OK.

> > suitable places to call `byte-compile-set-symbol-position' for
> > the error messages we see should be identified, the calls inserted, and
> > another bootstrap build done to see how much this helps.

> Sounds a good idea, thanks.

Done this too, in one place, which causes the "obsolete function"
messages to get the correct position.

Messages generated by `macroexp--warn-and-return' continue to have wrong
positions.  I'm not sure it's possible to fix this, and my intellect
isn't up to working out how it works, at least not today.

The messages about unknown functions, or not known to be defined at
runtime functions continue to give EOF as their position.  This was the
use case for the parameter `allow-previous' in
`byte-compile-set-symbol-position', but it didn't work before, and it
continues not to work just as well now.  It would be possible to fix
this by not deleting elements from `read-symbol-positions-list', but
this would slow down compilation (even if only a little), and generally
seems not to be worth the trouble.

-- 
Alan Mackenzie (Nuremberg, Germany).