bug#5162: 23.1.90; display-time broken in non-X build

bug#5162: 23.1.90; display-time broken in non-X build

[Sven Joachim]

In non-X builds, the display-time command is broken, because the
find-image function is not defined:

,----
| Debugger entered--Lisp error: (void-function find-image)
|   (find-image (quote (... ...)))
|   (defvar display-time-mail-icon (find-image (quote ...)) ("/usr/local/src/emacs-23.1.90/lisp/time.elc" . 7689))
|   execute-extended-command(nil)
|   call-interactively(execute-extended-command nil nil)
`----

The following error messages from the build process (I used the 23.1.90
pretest tarball) might also be relevant:

,----
| EMACSLOADPATH=/usr/local/src/emacs-23.1.90/leim/../lisp LC_ALL=C ../src/emacs -batch --no-init-file --no-site-file --multibyte -l /usr/local/src/emacs-23.1.90/leim/../lisp/international/titdic-cnv \
| 	  -f batch-miscdic-convert -dir quail /usr/local/src/emacs-23.1.90/leim/MISC-DIC; \
| 	  echo "changed" > changed.misc
| Converting CTLau.html to CTLau.el...
| Converting CTLau.html to CTLau.el...done
| Converting cangjie-table.b5 to tsang-b5.el...
| File mode specification error: (wrong-type-argument stringp image-type-auto-detected-p)
| Converting cangjie-table.b5 to tsang-b5.el...done
| Converting cangjie-table.b5 to quick-b5.el...
| File mode specification error: (wrong-type-argument stringp image-type-auto-detected-p)
| Converting cangjie-table.b5 to quick-b5.el...done
| Converting pinyin.map to PY.el...
| File mode specification error: (wrong-type-argument stringp image-type-auto-detected-p)
| Converting pinyin.map to PY.el...done
| Converting cangjie-table.cns to tsang-cns.el...
| File mode specification error: (wrong-type-argument stringp image-type-auto-detected-p)
| Converting cangjie-table.cns to tsang-cns.el...done
| Converting cangjie-table.cns to quick-cns.el...
| File mode specification error: (wrong-type-argument stringp image-type-auto-detected-p)
| Converting cangjie-table.cns to quick-cns.el...done
| Converting CTLau-b5.html to CTLau-b5.el...
| Converting CTLau-b5.html to CTLau-b5.el...done
| Converting ziranma.cin to ZIRANMA.el...
| File mode specification error: (wrong-type-argument stringp image-type-auto-detected-p)
| Converting ziranma.cin to ZIRANMA.el...done
| Compiling quail/tsang-b5.el
`----

In GNU Emacs 23.1.90.1 (i486-pc-linux-gnu)
 of 2009-12-09 on turtle
configured using `configure  '--without-x' '--build=i486-pc-linux-gnu' 'build_alias=i486-pc-linux-gnu''

Important settings:
  value of $LC_ALL: nil
  value of $LC_COLLATE: C
  value of $LC_CTYPE: nil
  value of $LC_MESSAGES: nil
  value of $LC_MONETARY: nil
  value of $LC_NUMERIC: nil
  value of $LC_TIME: nil
  value of $LANG: de_DE.UTF-8
  value of $XMODIFIERS: nil
  locale-coding-system: utf-8-unix
  default enable-multibyte-characters: t

Major mode: Fundamental

Minor modes in effect:
  menu-bar-mode: t
  file-name-shadow-mode: t
  global-font-lock-mode: t
  global-auto-composition-mode: t
  auto-encryption-mode: t
  auto-compression-mode: t
  line-number-mode: t
  transient-mark-mode: t

Recent input:
ESC [ > 0 ; 2 5 1 ; 0 c ESC ] 1 1 ; r g b : f f f f 
/ f f f f / f f f f ESC \ ESC x d i s p l TAB t i m 
TAB RET ESC x r e p o r t - e m TAB RET

Recent messages:
("./emacs")
For information about GNU Emacs and the GNU system, type C-h C-a.
defvar: Symbol's function definition is void: find-image

Load-path shadows:
None found.

Features:
(shadow mail-extr message ecomplete rfc822 mml easymenu mml-sec
password-cache mm-decode mm-bodies mm-encode mailcap mail-parse rfc2231
rfc2047 rfc2045 qp ietf-drums mailabbrev nnheader gnus-util netrc
time-date mm-util mail-prsvr gmm-utils wid-edit mailheader canlock sha1
hex-util hashcash mail-utils warnings emacsbug sendmail regexp-opt
ediff-hook vc-hooks lisp-float-type lisp-mode register page menu-bar
rfn-eshadow timer select mldrag mouse jit-lock font-lock syntax facemenu
font-core frame cham georgian utf-8-lang misc-lang vietnamese tibetan
thai tai-viet lao korean japanese hebrew greek romanian slovak czech
european ethiopic indian cyrillic chinese case-table epa-hook
jka-cmpr-hook help simple abbrev loaddefs button minibuffer faces
cus-face text-properties overlay md5 base64 format env code-pages mule
custom widget hashtable-print-readable backquote make-network-process
dbusbind multi-tty emacs)

bug#5162: 23.1.90; display-time broken in non-X build

[Glenn Morris]

Sven Joachim wrote:

> In non-X builds, the display-time command is broken, because the
> find-image function is not defined:

Works fine for me, in a clean build from the current CVS trunk.
I didn't try the actual 23.1.90 pretest.

Do you have find-image in your loaddefs.el? If not, I'd guess that you
did a with-x build first, then built without-x using the same
loaddefs.el file. The with and without-x loaddefs.el files are no
longer identical.

bug#5162: 23.1.90; display-time broken in non-X build

[Stefan Monnier]

> The with and without-x loaddefs.el files are no longer identical.

Then we need to detect this situation somehow and cause loaddefs.el to
be updated correspondingly.


        Stefan

bug#5162: 23.1.90; display-time broken in non-X build

[Glenn Morris]

Stefan Monnier wrote:

>> The with and without-x loaddefs.el files are no longer identical.
>
> Then we need to detect this situation somehow and cause loaddefs.el to
> be updated correspondingly.

The only idea I had was to regenerate loaddefs from scratch whenever
src/config.h changes.

bug#5162: 23.1.90; display-time broken in non-X build

[Stefan Monnier]

>>> The with and without-x loaddefs.el files are no longer identical.
>> Then we need to detect this situation somehow and cause loaddefs.el to
>> be updated correspondingly.
> The only idea I had was to regenerate loaddefs from scratch whenever
> src/config.h changes.

I think the way to make it work is to let autoload.el store the
autoload-excludes in the loaddefs.el file (as a special comment,
similar to the special comment listing the files that had no autoload
cookies).
Then autoload can compare the old list with the new one and force scan
any new unexcluded file, even if it's older than the timestamp.

Tho, maybe just not storing the auto-excludes in the "no-autoloads"
section might be enough to do the trick.  E.g. the patch below seems to
do the trick for me,


        Stefan


=== modified file 'lisp/emacs-lisp/autoload.el'
--- lisp/emacs-lisp/autoload.el	2009-11-07 01:55:16 +0000
+++ lisp/emacs-lisp/autoload.el	2009-12-10 07:52:26 +0000
@@ -457,8 +457,6 @@
           relfile
           ;; nil until we found a cookie.
           output-start)
-      (if (member absfile autoload-excludes)
-      	  (message "Generating autoloads for %s...skipped" file)
       (with-current-buffer (or visited
                                ;; It is faster to avoid visiting the file.
                                (autoload-find-file file))
@@ -572,7 +570,7 @@
           (message "Generating autoloads for %s...done" file))
         (or visited
             ;; We created this buffer, so we should kill it.
-            (kill-buffer (current-buffer)))))
+            (kill-buffer (current-buffer))))
       ;; If the entries were added to some other buffer, then the file
       ;; doesn't add entries to OUTFILE.
       (or (not output-start) otherbuf))))
@@ -729,8 +727,11 @@
 		  ((not (stringp file)))
 		  ((or (not (file-exists-p file))
                        ;; Remove duplicates as well, just in case.
-                       (member file done))
+                       (member file done)
+                       ;; If the file is actually excluded.
+                       (member (expand-file-name file) autoload-excludes))
                    ;; Remove the obsolete section.
+                   (setq changed t)
 		   (autoload-remove-section (match-beginning 0)))
 		  ((not (time-less-p (nth 4 form)
                                      (nth 5 (file-attributes file))))
@@ -746,9 +747,11 @@
 	    (setq files (delete file files)))))
       ;; Elements remaining in FILES have no existing autoload sections yet.
       (dolist (file files)
-        (if (autoload-generate-file-autoloads file nil buffer-file-name)
-            (push file no-autoloads)
-          (setq changed t)))
+        (cond
+         ((member (expand-file-name file) autoload-excludes) nil)
+         ((autoload-generate-file-autoloads file nil buffer-file-name)
+          (push file no-autoloads))
+         (t (setq changed t))))
 
       (when no-autoloads
 	;; Sort them for better readability.

bug#5162: 23.1.90; display-time broken in non-X build

[Andreas Schwab]

Glenn Morris <rgm@gnu.org> writes:

> The with and without-x loaddefs.el files are no longer identical.

If that is true then it is a major bug.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#5162: 23.1.90; display-time broken in non-X build

[Sven Joachim]

On 2009-12-10 08:58 +0100, Stefan Monnier wrote:

>>>> The with and without-x loaddefs.el files are no longer identical.
>>> Then we need to detect this situation somehow and cause loaddefs.el to
>>> be updated correspondingly.
>> The only idea I had was to regenerate loaddefs from scratch whenever
>> src/config.h changes.
>
> I think the way to make it work is to let autoload.el store the
> autoload-excludes in the loaddefs.el file (as a special comment,
> similar to the special comment listing the files that had no autoload
> cookies).
> Then autoload can compare the old list with the new one and force scan
> any new unexcluded file, even if it's older than the timestamp.
>
> Tho, maybe just not storing the auto-excludes in the "no-autoloads"
> section might be enough to do the trick.  E.g. the patch below seems to
> do the trick for me,

I see that you have already committed it, and it has definitely improved
the situation, but it is still necessary to run "make clean" when
re-configuring "--without-x" (before your patch, only the brute
"make maintainer-clean" would have done the trick).

Sven

bug#5162: 23.1.90; display-time broken in non-X build

[Stefan Monnier]

>> The with and without-x loaddefs.el files are no longer identical.
> If that is true then it is a major bug.

I thought so too at first, but I think it'll work.  Of course, thre may
still be scenario where it screws up somehow, but I can't hink of any:
basically loaddefs.el is only used during the dump process, so it
doesn't matter if it's build-specific.


        Stefan

bug#5162: 23.1.90; display-time broken in non-X build

[Andreas Schwab]

Stefan Monnier <monnier@IRO.UMontreal.CA> writes:

>>> The with and without-x loaddefs.el files are no longer identical.
>> If that is true then it is a major bug.
>
> I thought so too at first, but I think it'll work.  Of course, thre may
> still be scenario where it screws up somehow, but I can't hink of any:
> basically loaddefs.el is only used during the dump process, so it
> doesn't matter if it's build-specific.

Then it should be moved out of the source directory (and not be
distributed).

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#5162: 23.1.90; display-time broken in non-X build

[Eli Zaretskii]

> From: Andreas Schwab <schwab@linux-m68k.org>
> Date: Thu, 10 Dec 2009 10:22:19 +0100
> Cc: 5162@emacsbugs.donarmstrong.com, Sven Joachim <svenjoac@gmx.de>
> 
> Glenn Morris <rgm@gnu.org> writes:
> 
> > The with and without-x loaddefs.el files are no longer identical.
> 
> If that is true then it is a major bug.

I agree.  Does anyone know why, and which change(s) caused this?

bug#5162: 23.1.90; display-time broken in non-X build

[Glenn Morris]

Eli Zaretskii wrote:

> I agree.  Does anyone know why, and which change(s) caused this?

http://lists.gnu.org/archive/html/emacs-devel/2009-10/msg00117.html

A discussion that most of the commenters on this bug were involved in,
but this consequence was not appreciated till now.

If no good solution can be found, it can be reverted at no great loss.

bug#5162: 23.1.90; display-time broken in non-X build

[Eli Zaretskii]

> From: Glenn Morris <rgm@gnu.org>
> Cc: Andreas Schwab <schwab@linux-m68k.org>,  5162@emacsbugs.donarmstrong.com,  svenjoac@gmx.de
> Date: Thu, 10 Dec 2009 13:59:19 -0500
> 
> Eli Zaretskii wrote:
> 
> > I agree.  Does anyone know why, and which change(s) caused this?
> 
> http://lists.gnu.org/archive/html/emacs-devel/2009-10/msg00117.html

Thanks for the reminder.

> If no good solution can be found, it can be reverted at no great loss.

Maybe we should.  Having loaddefs.el depend on the particulars of the
build will definitely open a can of worms, because *.el files are
generally considered architecture and build independent.  For example,
does it still work to invoke Emacs built with X with the -nw option?

In general, if some build-dependent set of autoloads is skipped, then
how can we be sure that a binary which supports two different
configurations will work?

bug#5162: 23.1.90; display-time broken in non-X build

[Stefan Monnier]

> I see that you have already committed it, and it has definitely improved
> the situation, but it is still necessary to run "make clean" when
> re-configuring "--without-x" (before your patch, only the brute
> "make maintainer-clean" would have done the trick).

Hmm... indeed it seems that the normal build procedure runs
`cd lisp; make autoloads' only when lisp/loaddefs.el is missing and
doesn't otherwise update the file.

The patch below should fix that, tho at the cost of making the autoloads
twice in the bootstrap case.  Can you confirm it fixes the problem?
Assuming it does, I'm not sure whether we should use such a patch as is,
or whether we should "do it right"


        Stefan


--- Makefile.in.~1.213.~	2009-12-01 12:51:45.000000000 -0500
+++ Makefile.in	2009-12-10 13:26:49.000000000 -0500
@@ -109,7 +109,7 @@
 # with "make -j" and results in more deterministic compilation warnings.
 # cus-load and finder-inf are not explicitly requested by anything, so
 # we add them here to make sure they get built.
-all: compile-main $(lisp)/cus-load.el $(lisp)/finder-inf.el
+all: autoloads compile-main $(lisp)/cus-load.el $(lisp)/finder-inf.el
 	@: Let us check that we byte-compiled all the files.
 	$(MAKE) $(MFLAGS) compile-last EMACS=$(EMACS)
 

bug#5162: 23.1.90; display-time broken in non-X build

[Sven Joachim]

On 2009-12-10 20:09 +0100, Stefan Monnier wrote:

>> I see that you have already committed it, and it has definitely improved
>> the situation, but it is still necessary to run "make clean" when
>> re-configuring "--without-x" (before your patch, only the brute
>> "make maintainer-clean" would have done the trick).
>
> Hmm... indeed it seems that the normal build procedure runs
> `cd lisp; make autoloads' only when lisp/loaddefs.el is missing and
> doesn't otherwise update the file.
>
> The patch below should fix that, tho at the cost of making the autoloads
> twice in the bootstrap case.  Can you confirm it fixes the problem?

It doesn't, because Emacs is rebuilt first, using the outdated
loaddefs.el, and the autoloads are only updated later.

Sven

bug#5162: 23.1.90; display-time broken in non-X build

[Glenn Morris]

This change is pertinent to this issue, but does not seem to have been
mentioned here:

2009-12-17  Andreas Schwab  <schwab at linux-m68k.org>

    * emacs-lisp/autoload.el (batch-update-autoloads): Only
    exclude unconditionally preloaded files.

bug#5162: 23.1.90; display-time broken in non-X build

[Chong Yidong]

> This change is pertinent to this issue, but does not seem to have been
> mentioned here:
>
> 2009-12-17  Andreas Schwab  <schwab at linux-m68k.org>
>
>     * emacs-lisp/autoload.el (batch-update-autoloads): Only
>     exclude unconditionally preloaded files.

Looks like the with-x and without-x builds now have the same
loaddefs.el.