CVS Emacs doesn't build

CVS Emacs doesn't build

[Paul Pogonyshev]

Current CVS fails to build here:

paul@localhost:~/emacs$ make distclean && ./configure --with-x' && make bootstrap

...
Loading vc-cvs...
Generating autoloads for fringe.el...
Generating autoloads for fringe.el...done
Generating autoloads for pcvs.el...
Generating autoloads for pcvs.el...done
Generating autoloads for pcvs-defs.el...
Generating autoloads for pcvs-defs.el...done
Saving file /home/paul/emacs/lisp/loaddefs.el...
Symbol's function definition is void: coding-system-equal
make[1]: *** [autoloads] Error 255
make[1]: Leaving directory `/home/paul/emacs/lisp'
make: *** [bootstrap] Error 2

Re: CVS Emacs doesn't build

[Paul Pogonyshev]

I wrote:

> Current CVS fails to build here:

Apparently no one cares...

So, the cause is this change:

2004-10-12  Kenichi Handa  <handa@m17n.org>

	* international/mule.el (coding-system-equal): Move from mule-util.el.

	* international/mule-util.el (coding-system-equal): Move to mule.el.

I managed to build Emacs by reverting this patch and removing
`lisp/loaddefs.el'.  Presumably, the following one-liner should
solve the problem (i.e. instead of reverting the change of 12
October.)  I didn't verify though.

Paul


--- mule.el.~1.204.~	2004-10-18 01:41:16.000000000 +0300
+++ mule.el	2004-10-18 01:41:53.000000000 +0300
@@ -535,6 +535,7 @@ coding system whose eol-type is N."
 		 (and (not (> (downcase c1) (downcase c2)))
 		      (< c1 c2)))))))
 
+;;;###autoload
 (defun coding-system-equal (coding-system-1 coding-system-2)
   "Return t if and only if CODING-SYSTEM-1 and CODING-SYSTEM-2 are identical.
 Two coding systems are identical if two symbols are equal

Re: CVS Emacs doesn't build

[Luc Teirlinck]

Paul Pogonyshev wrote:

   I wrote:

   > Current CVS fails to build here:

   Apparently no one cares...

   So, the cause is this change:

   2004-10-12  Kenichi Handa  <handa@m17n.org>

	   * international/mule.el (coding-system-equal): Move from mule-util.el.

	   * international/mule-util.el (coding-system-equal): Move to mule.el.

I built Emacs CVS without problems several times after that.

It used to be that one had to do `make maintainer-clean' before doing
`make-bootstrap' to guarantee a successful build.  Maybe it still is.
There was talk about changing that, but I do not know whether
something has actually been done in the meantime.  The script I use
personally still does `make maintainer-clean'.

Sincerely,

Luc.

Re: CVS Emacs doesn't build

[Richard Stallman]

    I managed to build Emacs by reverting this patch and removing
    `lisp/loaddefs.el'.

We don't want to do that.

    +;;;###autoload
     (defun coding-system-equal (coding-system-1 coding-system-2)
       "Return t if and only if CODING-SYSTEM-1 and CODING-SYSTEM-2 are identical.
     Two coding systems are identical if two symbols are equal

    --- mule.el.~1.204.~	2004-10-18 01:41:16.000000000 +0300
    +++ mule.el	2004-10-18 01:41:53.000000000 +0300
    @@ -535,6 +535,7 @@ coding system whose eol-type is N."
		     (and (not (> (downcase c1) (downcase c2)))
			  (< c1 c2)))))))

    +;;;###autoload

mule.el is preloaded, so this autolod cookie ought to have no effect.

Can you debug why the build fails?

Re: CVS Emacs doesn't build

[Paul Pogonyshev]

Richard Stallman wrote:
>     I managed to build Emacs by reverting this patch and removing
>     `lisp/loaddefs.el'.
> 
> We don't want to do that.
> 
>     +;;;###autoload
>      (defun coding-system-equal (coding-system-1 coding-system-2)
>        "Return t if and only if CODING-SYSTEM-1 and CODING-SYSTEM-2 are identical.
>      Two coding systems are identical if two symbols are equal
> 
>     --- mule.el.~1.204.~ 2004-10-18 01:41:16.000000000 +0300
>     +++ mule.el 2004-10-18 01:41:53.000000000 +0300
>     @@ -535,6 +535,7 @@ coding system whose eol-type is N."
>        (and (not (> (downcase c1) (downcase c2)))
>      (< c1 c2)))))))
> 
>     +;;;###autoload
> 
> mule.el is preloaded, so this autolod cookie ought to have no effect.
> 
> Can you debug why the build fails?

I finally got around to it.  Change in autoloading wasn't necessary, just
removing `loaddefs.el' or `make maintainer-clean' was enough.  Sorry for
the false alarm.

Paul