Re: Compile issues

[Daniel Pfeiffer]

Saluton, Moin,

Dave Love <address@hidden> skribis:

> address@hidden (Daniel Pfeiffer) writes:
> 
> > The easy way out, (setq compilation-environment '("LANG=C")), is
> > hardly compatible with the goals of internationalization.
> 
> Setting LANG is wrong -- use LC_MESSAGES.  In the current situation,
> `compilation-environment' should be customized with a default that
> sets LC_MESSAGES.

True.  But, when I lately converted from a Latin-9 locale to a UTF-8 one, I
noticed that the semantics of the various LC_* vars are not homogenously
implemented in various programs.  So I stopped using them, but the working
solution would have been LANGUAGE.  Though this might again not be implemented
on all systems...

Anyway I was saying all of this is _not_ a correct solution.  For quite a few
of the regexps and associated data, we'd need national variants.  And these
would have to be multilingual, because not all programs are translated yet.
I'm really unsure what is the lesser evil, take away localization from the
compiler output or miss out on some of the messages.

> > I myself have verified and completed maybe at most one third of what
> > is there.  For the rest I just believed the old mode and its
> > comments.
> 
> Well I commented on (the English versions of) the cases I noticed.  I
> don't know what `verified' might mean here.

I tried to get many programs I have access to, to spit out some messages.

> >> >> If you tell us what the user does,
> >> > Types C-c C-k and then C-x `.
> >
> > Which is fixed, so how long are you going to discuss last weeks
> > errors.
> 
> I said I used `next-error' with the development code.  I see no
> evidence this was fixed last week, e.g. what I just sent to Stefan.
> I'm fed up with this sort of treatment when reporting Emacs problems,
> and I've had enough of this exchange.

Sorry!  On march 26th you had reported that "you get `Moved past last error'
on the first C-x `".  I replied that this was due to the point being used as
the search-next-error starting position.  On April 1st I checked in a fix for
this and copied you in on the announcement, which you apparently didn't get,
if I understand your enigmatic "??" and explanation thereof correctly.

So what you report here must be a different, much more limited mistake.  Since
you like to make incomplete statements, I can only guess that you are alluding
to this:

(save-window-excursion
  (save-excursion
    (global-font-lock-mode 1)
    (comint-run "bash")
    (compilation-minor-mode)
    font-lock-keywords))

But if you use (compilation-shell-minor-mode) instead, you can issue a command
(e.g. cat compilation.txt).  Then C-x ` works, correctly trying to jump to the
first error, even when the point is at the end.  So I still don't get your
bug.  Could we please have a new concrete, reproducible error report against a
current-at-time-of-writing version of compile?

> > Luckily only power users need to get into the internals.
> 
> That probably fits with how I view people who call themselves `power
> users'.
> 
> On the other hand, I want to be able to use `compilation-minor-mode'
> and customize its behaviour through a documented interface.

I agree, and from the above this seems to be given.

> I want to
> be able to do this in code which can cleanly use the released versions
> of compile.el too.

This is always difficult, albeit necessary for externally distributed modules.
 A test for the new mode might be (fboundp 'compilation-start)

coralament / best Grötens / liebe Grüße / best regards / elkorajn salutojn
Daniel Pfeiffer

-- 
lerne / learn / apprends / läramå / ucz się    Esperanto:
                              http://lernu.net/