Re: current compile.el issues

[Stefan Monnier]

>  * Since (I guess) there's no way to limit the region parsed for
>    errors now, C-x ` doesn't work the way it used to in Lisp
>    *Compile-Log* buffers -- you get the first error in the buffer, not
>    the first from the last compilation.

Actually what you describe was not the behavior I was seeing yesterday.
But the behavior of yesterday had the problem that even though C-x `
correctly jumped to the first message of last compilation (assuming no
comint-trancte-buffer was at play), M-x first-error still jumped to the
beginning of the buffer.
This was actually the behavior obtained with sml-mode+oldcompile.el.

But I just installed a change that introduces a new variable
`compilation-messages-start' variable, automatically set when calling
compilation-forget-errors (which python.el incidentally didn't do, even it
should have, even with oldcompile.el) which implements the "proper"
behavior (as provided by tex-mode.el in Emacs-20) where even M-x
first-error only jumps to the first error of the latest compilation.


        Stefan