Re: current compile.el issues

[Dave Love]

address@hidden (Daniel Pfeiffer) writes:

> I only have the current Emacs uninstalled, but I can't verify that, with or
> without -q.  In fact the two variables have the same value by default for me.

Presumably srcdir == builddir in that case.

emacs-version
  => "21.3.50.14"
data-directory
  => "/dl/sr/homes/px/fx/esrc/etc/"
doc-directory
  => "/scr/emacs-build/etc/"

> I think doing this the same way as scrolling is a good idea, but I
> can always set this option if it defaults to 0.  This is however the
> first request since the fringe arrow appeared, so I don't know if
> many people still want this.

I don't find the fringe arrow very helpful, especially if there's no
fringe!  It's also not the behaviour I've been used to for many years.

> But I think he means the additional decorative stuff that
> compilation-mode-font-lock-keywords gives.

Yes.

> I don't know if these should be customizable,

I don't understand the purpose of that pattern.  It doesn't seem
useful and it is distracting when scanning the output -- typically
every line has a highlighted item.  I think the rest of the patterns
are OK.  (I know developers other than gerd don't seem to get the same
cognitive effects as I do, but I don't find I'm that different from
most people.)

> I'm not sure that can be avoided.  One minor point is multiple
> messages on a line ? if you didn't have the cursor on one, you
> couldn't see, which message you are visiting.

I don't know when that's an issue.  Not, as far as I know, with the
systems I use.  Unfortunately it seems only Stefan and I understand
this mode of working.

> A major concern would be that the point may be so far from the
> message that you can't display both in one window at the same time.

In that case, point probably has to move, but it usually won't be an
issue.  One could also limit the output from the subprocess if
necessary.

> Do you want the comint window to then get split?

I guess that would be a reasonable solution if it's necessary.

> And are you only concerned
> with this for comint, i.e. have a different behaviour for that?

I was talking specifically about compilation-shell-minor-mode.

> This could be done with the new next-error-function, which you could
> set to a save-excursion`ed variant in comint-buffers.

I've yet to understand what to do with that stuff, now it's changed
under me again.  If you're supposed to define your own function and
the minor mode is supposed to redefine it, there's a problem.

>> >  * I still don't know what the proper definition of `informational'
>> >    messages is, since messages can be highlighted with
>> >    `compilation-info-face' when compilation-skip-threshold's value is 1.
>> 
>> I don't understand this.  Maybe Daniel can help.
>
> compilation-skip-threshold is just that, i.e. saying what (not) to skip with
> C-x ` et al.  It has nothing to do with display, i.e. faces.  You may well
> wan't to normally skip these, but occasionally mouse-2 click on one, to visit
> it anyways.  You wouldn't know you could, if it weren't highlighted.

I'm asking what the _definition_ of such messages is.  What the
patterns highlight isn't consistent with the story was about what
informational messages are.  It isn't internally consistent either
since the same sort of warning message from gcc will apparently be
treated differently from other compilers.  Compare

foo.c(2:0) : informational EDC0804: Function foo is not referenced.
/tmp/a.c:2: warning: `foo' defined but not used