Steve Youngs <[EMAIL PROTECTED]> writes:
> There will need to be at least some separation because the current
> system is completely broken for XEmacs auto-autoloads.
What exactly is "broken" ?
> > (GNU Emacs would use separate compilation whereas XEmacs would
> > compile all files in batch, ...).
>
> My version is far more efficient, I only have the overhead of loading
> XEmacs once, you have that overhead multiplied by the number of files
> to be built.
Yes, but the overhead is _really_ small :
$ time xemacs -batch -no-autoloads -l bytecomp \
-eval '(princ "hello world")' hello world
real 0m0.048s
user 0m0.020s
sys 0m0.030s
and the possibility to have separate compilation (modify only
xtla-browse.el, then type "make") seems more important to me than
those 0m0.048s.
Using virtual targets, typing "make" on an already built project will
re-evaluate the action for those virtual targets anyway.
Also, if you have an intermediate file, it's more in the make
philosophy to have a separate target for this file. For example, I
strongly prefer
xtla-version.elc: xtla-version.el
how to byte-compile xtla-version.el
xtla-version.el:
how to make xtla-version.el
to
xtla-version.elc:
how to make xtla-version.el
how to byte-compile xtla-version.el
--
Matthieu
