subr.el change breaks bootstrapping

subr.el change breaks bootstrapping

[Juanma Barranquero]

Something in this change breaks bootstrapping.

2006-05-24  Alan Mackenzie  <acm@muc.de>

	* startup.el (command-line): For names of preloaded files, don't
	append ".elc" (now done in Fload), and call file-truename on the
	lisp directory.

	* subr.el (eval-after-load): Fix the doc-string.  Allow FILE to
	match ANY loaded file with the right name, not just those in
	load-path.  Put a regexp matching the file name into
	after-load-alist, rather than the name itself.

	* subr.el: New functions load-history-regexp,
	load-history-filename-element, do-after-load-evaluation.

	* international/mule.el (load-with-code-conversion): Do the
	eval-after-load stuff by calling do-after-load-evaluation.


-- 
                    /L/e/k/t/u

Re: subr.el change breaks bootstrapping

[Chong Yidong]

"Juanma Barranquero" <lekktu@gmail.com> writes:

> Something in this change breaks bootstrapping.

Also, the changes to subr.el cause a pure space overflow.

> 2006-05-24  Alan Mackenzie  <acm@muc.de>
>
> 	* startup.el (command-line): For names of preloaded files, don't
> 	append ".elc" (now done in Fload), and call file-truename on the
> 	lisp directory.
>
> 	* subr.el (eval-after-load): Fix the doc-string.  Allow FILE to
> 	match ANY loaded file with the right name, not just those in
> 	load-path.  Put a regexp matching the file name into
> 	after-load-alist, rather than the name itself.
>
> 	* subr.el: New functions load-history-regexp,
> 	load-history-filename-element, do-after-load-evaluation.
>
> 	* international/mule.el (load-with-code-conversion): Do the
> 	eval-after-load stuff by calling do-after-load-evaluation.

Re: subr.el change breaks bootstrapping

[Alan Mackenzie]

On Wed, 24 May 2006, Juanma Barranquero wrote:

>Something in this change breaks bootstrapping.

The other half of the change in in .../src/lread.c.  Did you compile that
in too?

>2006-05-24  Alan Mackenzie  <acm@muc.de>
>
>	* startup.el (command-line): For names of preloaded files, don't
>	append ".elc" (now done in Fload), and call file-truename on the
>	lisp directory.
>
>	* subr.el (eval-after-load): Fix the doc-string.  Allow FILE to
>	match ANY loaded file with the right name, not just those in
>	load-path.  Put a regexp matching the file name into
>	after-load-alist, rather than the name itself.
>
>	* subr.el: New functions load-history-regexp,
>	load-history-filename-element, do-after-load-evaluation.
>
>	* international/mule.el (load-with-code-conversion): Do the
>	eval-after-load stuff by calling do-after-load-evaluation.

-- 
Alan.

Re: subr.el change breaks bootstrapping

[Luc Teirlinck]

Alan Mackenzie wrote:

   The other half of the change in in .../src/lread.c.  Did you compile that
   in too?

I believe that after your changes jka-cmpr-hook.el was used before it
was loaded.  I loaded it earlier and increased BASE_PURESIZE.

I believe that this is the right fix.  If not, we can undo my changes
and fix it differently.

Sincerely,

Luc.

Re: subr.el change breaks bootstrapping

[Luc Teirlinck]

>From my previous reply:

   I believe that after your changes jka-cmpr-hook.el was used before it
   was loaded.

More precisely, the new function load-history-regexp in subr.el uses
jka-compr-load-suffixes and then your changes called that function
before jka-cmpr-hook.el was loaded.

Sincerely,

Luc.

Re: subr.el change breaks bootstrapping

[Alan Mackenzie]

Hi, Luc.

On Wed, 24 May 2006, Luc Teirlinck wrote:

>>>From my previous reply:

>   I believe that after your changes jka-cmpr-hook.el was used before it
>   was loaded.

>More precisely, the new function load-history-regexp in subr.el uses
>jka-compr-load-suffixes and then your changes called that function
>before jka-cmpr-hook.el was loaded.

OK.  This sounds right.

If there were to be any problem with the fix you've made, another fix
would be to amend load-history-regexp so that jka-compr-load-suffixes is
only used inside an (unless purify-flag ....) form.

>Sincerely,

>Luc.

-- 
Alan.

Re: subr.el change breaks bootstrapping

[Stefan Monnier]

> If there were to be any problem with the fix you've made, another fix
> would be to amend load-history-regexp so that jka-compr-load-suffixes is
> only used inside an (unless purify-flag ....) form.

Rather than an "unless purify-flag" I'd recommend an "when (fboundp ...)"


        Stefan

Re: subr.el change breaks bootstrapping

[Luc Teirlinck]

Alan Mackenzie wrote:

   If there were to be any problem with the fix you've made,

The only drawback I see from loading "jka-cmpr-hook" this early is
that it will not be possible to load multilingual text from that file.
This is not a problem now and very unlikely to ever be.  You can not
load multilingual text from several other files that are loaded even
earlier either: subr, widget, custom, files, simple and others.

   another fix would be to amend load-history-regexp so that
   jka-compr-load-suffixes is only used inside an (unless purify-flag
   ....) form

Is the correct value of jka-compr-load-suffixes ever needed at the
early stage where it caused trouble?  Autoloading
jka-compr-load-suffixes might also be a solution, if allowing for
future multilingual text in "jka-cmpr-hook" would be a real issue.

Sincerely,

Luc.

Re: subr.el change breaks bootstrapping

[Juanma Barranquero]

On 5/24/06, Alan Mackenzie <acm@muc.de> wrote:

> The other half of the change in in .../src/lread.c.  Did you compile that
> in too?

I was bootstraping from a fresh "cvs -q update -C".

-- 
                    /L/e/k/t/u

Re: subr.el change breaks bootstrapping

[Juanma Barranquero]

After the latest batch of changes, I'm still unable to bootstrap.

 cvs -q update -C
 cd nt
 nmake realclean
 configure
 nmake bootstrap

brings

 [...]
 Saving file c:/bin/emacs/HEAD/lisp/loaddefs.el...
 Loading c:/bin/emacs/HEAD/lisp/vc-cvs.el (source)...
 Symbol's value as variable is void: dired-mode-map
 make[1]: *** [autoloads] Error -1
 make[1]: Leaving directory `C:/bin/emacs/HEAD/lisp'
 make: *** [bootstrap-gmake] Error 2

-- 
                    /L/e/k/t/u

Re: subr.el change breaks bootstrapping

[Nick Roberts]

 >  [...]
 >  Saving file c:/bin/emacs/HEAD/lisp/loaddefs.el...
 >  Loading c:/bin/emacs/HEAD/lisp/vc-cvs.el (source)...
 >  Symbol's value as variable is void: dired-mode-map
 >  make[1]: *** [autoloads] Error -1
 >  make[1]: Leaving directory `C:/bin/emacs/HEAD/lisp'
 >  make: *** [bootstrap-gmake] Error 2

Should be fixed now.

-- 
Nick                                           http://www.inet.net.nz/~nickrob

Re: subr.el change breaks bootstrapping

[Juanma Barranquero]

On 5/24/06, Nick Roberts <nickrob@snap.net.nz> wrote:

> Should be fixed now.

Yes, thanks.
                    /L/e/k/t/u

Re: subr.el change breaks bootstrapping

[Alan Mackenzie]

Morning, Luc!

On Wed, 24 May 2006, Luc Teirlinck wrote:

>Alan Mackenzie wrote:
>
>   If there were to be any problem with the fix you've made,

>The only drawback I see from loading "jka-cmpr-hook" this early is
>that it will not be possible to load multilingual text from that file.
>This is not a problem now and very unlikely to ever be.  You can not
>load multilingual text from several other files that are loaded even
>earlier either: subr, widget, custom, files, simple and others.

Sounds cheerful!

>   another fix would be to amend load-history-regexp so that
>   jka-compr-load-suffixes is only used inside an (unless purify-flag
>   ....) form

>Is the correct value of jka-compr-load-suffixes ever needed at the
>early stage where it caused trouble?

I doubt it very much.  The trouble only happens at preloading time, and
nobody is ever going to be compressing the files.elc that get preloaded.
And if they did want to do this, there would be other workarounds.  In
fact, there are already several (many?) workarounds in place for
preloading files, so another one wouldn't make much odds.

>Autoloading jka-compr-load-suffixes might also be a solution, if
>allowing for future multilingual text in "jka-cmpr-hook" would be a real
>issue.

Just as a side issue, what does the prefix "jka-" signify?  Is that the
author's initials?

>Sincerely,

>Luc.

-- 
Alan.

Re: subr.el change breaks bootstrapping

[Luc Teirlinck]

Alan Mackenzie wrote:

   Just as a side issue, what does the prefix "jka-" signify?  Is that the
   author's initials?

In as far as I know, it indeed seems to be just the author's initials.

Sincerely,

Luc.