Not loading the Registry settings

Not loading the Registry settings

[Juanma Barranquero]

I'd like to have a way to start Emacs with "emacs -Q" and have it not
to load the values from the Windows registry.

Adding a new command line option is easy, but I'm looking for the
best, non-Windows-specific way. How it's Emacs currently forced to
ignore X settings?

    Juanma

Re: Not loading the Registry settings

[David De La Harpe Golden]

Juanma Barranquero wrote:
> I'd like to have a way to start Emacs with "emacs -Q" and have it not
> to load the values from the Windows registry.
> 
> Adding a new command line option is easy, but I'm looking for the
> best, non-Windows-specific way. How it's Emacs currently forced to
> ignore X settings?
>

If you mean X resource database settings, I suspect they are always used 
unless you go all the way and -nw

_However_ you can override them on the command line with --xrm i.e even 
if I have 'Emacs.Fontbackend: xft' in my x resource database, I can still do
$ emacs -Q --xrm 'Emacs.FontBackend: x'
and enjoy awful font rendering.

looking at w32_make_rdb(), --xrm may work on w32 to override
registry settings? (untested).

Re: Not loading the Registry settings

[Adrian Robert]

David De La Harpe Golden <david@harpegolden.net> writes:

> Juanma Barranquero wrote:
>> I'd like to have a way to start Emacs with "emacs -Q" and have it not
>> to load the values from the Windows registry.
>>
>> Adding a new command line option is easy, but I'm looking for the
>> best, non-Windows-specific way. How it's Emacs currently forced to
>> ignore X settings?
>>
>
> If you mean X resource database settings, I suspect they are always
> used unless you go all the way and -nw

FWIW, under OS X the user consensus was that the semantics of -q/Q implied
that user customizations should not be loaded even if they come from the
resource database instead of .emacs.  Thus, there is the "ns_no_defaults"
hack in emacs.c.  It would be easier to do this cleanly if other platforms
choose to implement it as well.

Re: Not loading the Registry settings

[Juanma Barranquero]

On Sat, Sep 12, 2009 at 14:28, David De La Harpe Golden
<david@harpegolden.net> wrote:

> If you mean X resource database settings, I suspect they are always used
> unless you go all the way and -nw

Hmm. The trouble is, I want something to override all Registry
settings, precisely when I'm in window mode. So I can test some bug,
for example, without my Emacs.Background, Emacs.Foreground and
Emacs.Geometry settings getting in the way.

> looking at w32_make_rdb(), --xrm may work on w32 to override
> registry settings? (untested).

That works, but it is awfully ugly if you have many Registry settings.
I'd like a simple way to just say "no Registry settings AT ALL". I can
workaround now by running Emacs with a script doing

  emacs -Q -xrm Emacs.Geometry: -xrm Emacs.Foreground: -xrm Emacs.Background:

but it is a PITA.

    Juanma

Re: Not loading the Registry settings

[Stefan Monnier]

>> If you mean X resource database settings, I suspect they are always used
>> unless you go all the way and -nw

> Hmm. The trouble is, I want something to override all Registry
> settings, precisely when I'm in window mode. So I can test some bug,
> for example, without my Emacs.Background, Emacs.Foreground and
> Emacs.Geometry settings getting in the way.

Agreed.  Actually, I think that "-Q" should do that.  After all, to me
the main use of "-Q" is to have a clean start state that is independent
of any user settings.


        Stefan

Re: Not loading the Registry settings

[Miles Bader]

Stefan Monnier <monnier@iro.umontreal.ca> writes:
>> Hmm. The trouble is, I want something to override all Registry
>> settings, precisely when I'm in window mode. So I can test some bug,
>> for example, without my Emacs.Background, Emacs.Foreground and
>> Emacs.Geometry settings getting in the way.
>
> Agreed.  Actually, I think that "-Q" should do that.  After all, to me
> the main use of "-Q" is to have a clean start state that is independent
> of any user settings.

I agree.

[Currently now I do this by using "HOME=/tmp emacs -Q" (since I use
~/.Xdefaults, not the server's xrdb db), but it's a slight pain 'cause I
don't always remember...]

-Miles

-- 
Alone, adj. In bad company.

Re: Not loading the Registry settings

[Juanma Barranquero]

This a sample patch, not nearly ready (no docs, for example), just to
check whether I'm going in the right direction. It is the simplest
patch that works on Windows.

Comments:

 - It piggybacks on -Q, as Stefan suggested. Personally I think (but
not very strongly) that a new option would be clearer.

 - It basically turns w32reg:x_get_string_resource() into a no-op.
Would the same approach for xrdb:x_get_string_resource() work? In
particular, in the ~/.Xdefaults case mentioned by Miles?

 - As defined, it is an irreversible operation. Once -Q is used,
x_get_string_resource is a no-op for the duration of the Emacs
instance. I've assumed that is the intended effect.

Thoughts?

    Juanma


diff --git a/src/emacs.c b/src/emacs.c
index 2c14be5..2f25248 100644
--- a/src/emacs.c
+++ b/src/emacs.c
@@ -239,6 +239,9 @@ int noninteractive;

 int noninteractive1;

+/* Nonzero means Emacs was started with --quick.  */
+int emacs_quick_startup;
+
 /* Name for the server started by the daemon.*/
 static char *daemon_name;

@@ -1472,6 +1475,16 @@ main (int argc, char **argv)
   no_loadup
     = argmatch (argv, argc, "-nl", "--no-loadup", 6, NULL, &skip_args);

+  /* Check the -quick switch, which means don't load init or site files, nor X
+     settings or Windows registry, and no splash; but pass it back
for startup.el.  */
+  if (argmatch (argv, argc, "-Q", "--quick", 5, NULL, &skip_args))
+    {
+      emacs_quick_startup = 1;
+      skip_args--;
+    }
+  else
+    emacs_quick_startup = 0;
+
 #ifdef HAVE_NS
   ns_alloc_autorelease_pool();
   if (!noninteractive)
diff --git a/src/lisp.h b/src/lisp.h
index 820c258..38e4854 100644
--- a/src/lisp.h
+++ b/src/lisp.h
@@ -3128,6 +3128,7 @@ extern Lisp_Object Vinvocation_name,
Vinvocation_directory;
 extern Lisp_Object Vbefore_init_time, Vafter_init_time;
 extern Lisp_Object Vinstallation_directory;
 extern Lisp_Object empty_unibyte_string, empty_multibyte_string;
+extern int emacs_quick_startup;
 EXFUN (Fkill_emacs, 1);
 #if HAVE_SETLOCALE
 void fixup_locale P_ ((void));
diff --git a/src/w32reg.c b/src/w32reg.c
index d2330e7..89294f8 100644
--- a/src/w32reg.c
+++ b/src/w32reg.c
@@ -150,6 +150,9 @@ x_get_string_resource (rdb, name, class)
      XrmDatabase rdb;
      char *name, *class;
 {
+  /* --quick was passed, so this is a no-op.  */
+  if (emacs_quick_startup) return NULL;
+
   if (rdb)
     {
       char *resource;

Re: Not loading the Registry settings

[Stefan Monnier]

>  - It basically turns w32reg:x_get_string_resource() into a no-op.

Interesting approach.

> Would the same approach for xrdb:x_get_string_resource() work? In
> particular, in the ~/.Xdefaults case mentioned by Miles?

I do not know, but I'd think so, yes.

>  - As defined, it is an irreversible operation. Once -Q is used,
> x_get_string_resource is a no-op for the duration of the Emacs
> instance. I've assumed that is the intended effect.

I think that's OK, but the double-parsing of -Q is not
very satisfactory.  I'd rather see a new variable Vinhibit_x_resources
instead, which the Lisp code could set to non-nil when it parses
the -Q argument.


        Stefan

Re: Not loading the Registry settings

[Juanma Barranquero]

On Mon, Sep 14, 2009 at 15:29, Stefan Monnier <monnier@iro.umontreal.ca> wrote:

> Interesting approach.

It's simple.

In my patch, "emacs -Q -xrm resource" won't obey -xrm either, but it
is trivially easy to change it to allow that, if desired (it's just
moving the "if() ... return" down a few lines in
w32reg:x_get_resource_string).

> I think that's OK, but the double-parsing of -Q is not
> very satisfactory.  I'd rather see a new variable Vinhibit_x_resources
> instead, which the Lisp code could set to non-nil when it parses
> the -Q argument.

I had in fact done that before (though I reused instead the currently
undocumented emacs-quick-startup variable that startup.el uses now).
But I didn't want to mess with the processing of -Q/--quick in the
HAVE_NS block immediately following:

#ifdef HAVE_NS
  ns_alloc_autorelease_pool();
  if (!noninteractive)
    {
      char *tmp;
      display_arg = 4;
      if (argmatch (argv, argc, "-q", "--no-init-file", 6, NULL, &skip_args))
        {
          ns_no_defaults = 1;
          skip_args--;
        }
      if (argmatch (argv, argc, "-Q", "--quick", 5, NULL, &skip_args))
        {
          ns_no_defaults = 1;
          skip_args--;
        }
#ifdef NS_IMPL_COCOA
      if (skip_args < argc)
        {
          if (!strncmp(argv[skip_args], "-psn", 4))
            {
              skip_args += 1;
              chdir (getenv ("HOME"));
            }
          else if (skip_args+1 < argc && !strncmp(argv[skip_args+1], "-psn", 4))
            {
              skip_args += 2;
              chdir (getenv ("HOME"));
            }
        }
#endif
      /* This used for remote operation.. not fully implemented yet. */
      if (argmatch (argv, argc, "-_NSMachLaunch", 0, 3, &tmp, &skip_args))
          display_arg = 4;
      else if (argmatch (argv, argc, "-MachLaunch", 0, 3, &tmp, &skip_args))
          display_arg = 4;
      else if (argmatch (argv, argc, "-macosx", 0, 2, NULL, &skip_args))
          display_arg = 4;
      else if (argmatch (argv, argc, "-NSHost", 0, 3, &tmp, &skip_args))
          display_arg = 4;
    }
#endif /* HAVE_NS */

I think intermixing this with the new processing of -Q/--quick will
lead to uglier code...

    Juanma

Re: Not loading the Registry settings

[Juanma Barranquero]

On Mon, Sep 14, 2009 at 15:48, Juanma Barranquero <lekktu@gmail.com> wrote:

> I think that's OK, but the double-parsing of -Q is not
> very satisfactory.  I'd rather see a new variable Vinhibit_x_resources
> instead, which the Lisp code could set to non-nil when it parses
> the -Q argument.

I think I've misread this bit in my previous answer. Do you mean that
all the -Q processing would be done in startup.el? I think it is too
late then to affect to some initializations that read from X resources
/ Registry. What I did instead was to process -Q in emacs.c and set a
Lisp variable for startup.el.

    Juanma

Re: Not loading the Registry settings

[Adrian Robert]

Juanma Barranquero <lekktu@gmail.com> writes:

> On Mon, Sep 14, 2009 at 15:29, Stefan Monnier <monnier@iro.umontreal.ca> wrote:

>> I think that's OK, but the double-parsing of -Q is not
>> very satisfactory.  I'd rather see a new variable Vinhibit_x_resources
>> instead, which the Lisp code could set to non-nil when it parses
>> the -Q argument.
>
> I had in fact done that before (though I reused instead the currently
> undocumented emacs-quick-startup variable that startup.el uses now).
> But I didn't want to mess with the processing of -Q/--quick in the
> HAVE_NS block immediately following:

You can remove the -q/-Q part of that after this because it is just
doing what you are trying to do with the new patch.

(Vocal NS port user consensus a while back was that the semantics of -Q
implied that user customizations should not be loaded even if they
come from the resource database instead of .emacs.)

Re: Not loading the Registry settings

[Adrian Robert]

On Sep 14, 2009, at 3:14 PM, Stefan Monnier wrote:

> Something Emacs.app-related came up while discussing something else:

Ouch, I've tried twice to reply to this thread on emacs-devel, but  
something must be wrong with my GNUs configuration.  Let me try again  
below.



> Juanma Barranquero <lekktu@gmail.com> wrote:
>> I had in fact done that before (though I reused instead the currently
>> undocumented emacs-quick-startup variable that startup.el uses now).
>> But I didn't want to mess with the processing of -Q/--quick in the
>> HAVE_NS block immediately following:


The -Q/-q portion of this can be removed once you've implemented your  
change, since it implements the same thing.  I can clean up the rest  
in other files once this happens.

(A consensus of NS port users and developers decided a while ago that  
the semantics of -Q implied that user customizations should not be  
loaded, even if they come from the resource database instead of .emacs.)


-Adrian

Re: Not loading the Registry settings

[Stefan Monnier]

>> I think that's OK, but the double-parsing of -Q is not
>> very satisfactory.  I'd rather see a new variable Vinhibit_x_resources
>> instead, which the Lisp code could set to non-nil when it parses
>> the -Q argument.

> I think I've misread this bit in my previous answer. Do you mean that
> all the -Q processing would be done in startup.el?

Yes.

> I think it is too late then to affect to some initializations that
> read from X resources / Registry.  What I did instead was to
> process -Q in emacs.c and set a Lisp variable for startup.el.

I guess it'd be OK to move it from Lisp to C, but I'd first like to
understand which initializations take place too early to be affected by
Lisp code.


        Stefan

Re: Not loading the Registry settings

[Juanma Barranquero]

On Mon, Sep 14, 2009 at 23:53, Stefan Monnier <monnier@iro.umontreal.ca> wrote:

> but I'd first like to
> understand which initializations take place too early to be affected by
> Lisp code.

Attached is a backtrace of the first call to
w32reg.c:x_get_string_resource on an emacs -Q run.

As for the Lisp variable, `inhibit-x-resources' suggests that it can
be toggled by the user or lisp code. Is that the intended effect?

    Juanma



Breakpoint 3, x_get_string_resource (rdb=0x0, name=0x82ea80
"emacs.fontset-0", class=0x82ea60 "Emacs.Fontset-0") at w32reg.c:153
153       if (rdb)
(gdb) bt
#0  x_get_string_resource (rdb=0x0, name=0x82ea80 "emacs.fontset-0",
class=0x82ea60 "Emacs.Fontset-0") at w32reg.c:153
#1  0x0124d96a in xrdb_get_resource (rdb=0x0, attribute=49875811,
class=49875843, component=48252929, subclass=48252929) at frame.c:3858
#2  0x0124d9c5 in Fx_get_resource (attribute=49875811, class=49875843,
component=48252929, subclass=48252929) at frame.c:3884
#3  0x0103dc51 in Ffuncall (nargs=3, args=0x82ebf0) at eval.c:3056
#4  0x011df179 in Fbyte_code (bytestr=20903515, vector=20903580,
maxdepth=40) at bytecode.c:678
#5  0x0103e77d in funcall_lambda (fun=20903492, nargs=0,
arg_vector=0x82ef14) at eval.c:3233
#6  0x0103dfd9 in Ffuncall (nargs=1, args=0x82ef10) at eval.c:3092
#7  0x011df179 in Fbyte_code (bytestr=20914331, vector=20914564,
maxdepth=48) at bytecode.c:678
#8  0x0103e77d in funcall_lambda (fun=20914300, nargs=0,
arg_vector=0x82f234) at eval.c:3233
#9  0x0103dfd9 in Ffuncall (nargs=1, args=0x82f230) at eval.c:3092
#10 0x011df179 in Fbyte_code (bytestr=20369283, vector=20369364,
maxdepth=32) at bytecode.c:678
#11 0x0103be89 in Feval (form=20369269) at eval.c:2383
#12 0x01039b9a in internal_lisp_condition_case (var=48316609,
bodyform=20369269, handlers=20369461) at eval.c:1458
#13 0x011dfe32 in Fbyte_code (bytestr=20367203, vector=20368132,
maxdepth=56) at bytecode.c:868
#14 0x0103e77d in funcall_lambda (fun=20367180, nargs=0,
arg_vector=0x82f924) at eval.c:3233
#15 0x0103dfd9 in Ffuncall (nargs=1, args=0x82f920) at eval.c:3092
#16 0x011df179 in Fbyte_code (bytestr=20363323, vector=20363540,
maxdepth=48) at bytecode.c:678
#17 0x0103e77d in funcall_lambda (fun=20363300, nargs=0,
arg_vector=0x82fbb0) at eval.c:3233
#18 0x0103e21a in apply_lambda (fun=20363300, args=48252929,
eval_flag=1) at eval.c:3157
#19 0x0103c268 in Feval (form=48920549) at eval.c:2419
#20 0x01006e92 in top_level_2 () at keyboard.c:1366
#21 0x01039c96 in internal_condition_case (bfun=0x1006e7e
<top_level_2>, handlers=48316609, hfun=0x10069d0 <cmd_error>) at
eval.c:1513
#22 0x01006ec7 in top_level_1 () at keyboard.c:1374
#23 0x010396ef in internal_catch (tag=48312729, func=0x1006e97
<top_level_1>, arg=48252929) at eval.c:1249
#24 0x01006e02 in command_loop () at keyboard.c:1329
#25 0x01006127 in recursive_edit_1 () at keyboard.c:951
#26 0x010065fc in Frecursive_edit () at keyboard.c:1013
#27 0x01002a8d in main (argc=2, argv=0xa92748) at emacs.c:1849

Lisp Backtrace:
"x-get-resource" (0x82ebf4)
"create-fontset-from-x-resource" (0x82ef14)
"w32-initialize-window-system" (0x82f234)
"byte-code" (0x82f4b0)
"command-line" (0x82f924)
"normal-top-level" (0x82fbb0)

Re: Not loading the Registry settings

[Stefan Monnier]

>> but I'd first like to understand which initializations take place too
>> early to be affected by Lisp code.

> Attached is a backtrace of the first call to
> w32reg.c:x_get_string_resource on an emacs -Q run.

Thanks.  The patch below should address this issue (it just moves code
around).

> As for the Lisp variable, `inhibit-x-resources' suggests that it can
> be toggled by the user or lisp code.  Is that the intended effect?

Well, I'm not sure if it would be useful, but the name seems to reflect
more precisely the effect of the variable, and indeed it can later on
be changed from Lisp, tho it's not clear whether that would be useful.


        Stefan


=== modified file 'lisp/startup.el'
--- lisp/startup.el	2009-09-13 04:07:30 +0000
+++ lisp/startup.el	2009-09-15 01:28:52 +0000
@@ -727,44 +727,6 @@
     (setq eol-mnemonic-dos  "(DOS)"
           eol-mnemonic-mac  "(Mac)")))
 
-  ;; Make sure window system's init file was loaded in loadup.el if
-  ;; using a window system.
-  (condition-case error
-    (unless noninteractive
-      (if (and initial-window-system
-	       (not (featurep
-		     (intern
-		      (concat (symbol-name initial-window-system) "-win")))))
-	  (error "Unsupported window system `%s'" initial-window-system))
-      ;; Process window-system specific command line parameters.
-      (setq command-line-args
-	    (funcall
-	     (or (cdr (assq initial-window-system handle-args-function-alist))
-		 (error "Unsupported window system `%s'" initial-window-system))
-	     command-line-args))
-      ;; Initialize the window system. (Open connection, etc.)
-      (funcall
-       (or (cdr (assq initial-window-system window-system-initialization-alist))
-	   (error "Unsupported window system `%s'" initial-window-system))))
-    ;; If there was an error, print the error message and exit.
-    (error
-     (princ
-      (if (eq (car error) 'error)
-	  (apply 'concat (cdr error))
-	(if (memq 'file-error (get (car error) 'error-conditions))
-	    (format "%s: %s"
-                    (nth 1 error)
-                    (mapconcat (lambda (obj) (prin1-to-string obj t))
-                               (cdr (cdr error)) ", "))
-	  (format "%s: %s"
-                  (get (car error) 'error-message)
-                  (mapconcat (lambda (obj) (prin1-to-string obj t))
-                             (cdr error) ", "))))
-      'external-debugging-output)
-     (terpri 'external-debugging-output)
-     (setq initial-window-system nil)
-     (kill-emacs)))
-
   (set-locale-environment nil)
 
   ;; Convert preloaded file names in load-history to absolute.
@@ -869,6 +831,44 @@
     (and command-line-args
          (setcdr command-line-args args)))
 
+  ;; Make sure window system's init file was loaded in loadup.el if
+  ;; using a window system.
+  (condition-case error
+    (unless noninteractive
+      (if (and initial-window-system
+	       (not (featurep
+		     (intern
+		      (concat (symbol-name initial-window-system) "-win")))))
+	  (error "Unsupported window system `%s'" initial-window-system))
+      ;; Process window-system specific command line parameters.
+      (setq command-line-args
+	    (funcall
+	     (or (cdr (assq initial-window-system handle-args-function-alist))
+		 (error "Unsupported window system `%s'" initial-window-system))
+	     command-line-args))
+      ;; Initialize the window system. (Open connection, etc.)
+      (funcall
+       (or (cdr (assq initial-window-system window-system-initialization-alist))
+	   (error "Unsupported window system `%s'" initial-window-system))))
+    ;; If there was an error, print the error message and exit.
+    (error
+     (princ
+      (if (eq (car error) 'error)
+	  (apply 'concat (cdr error))
+	(if (memq 'file-error (get (car error) 'error-conditions))
+	    (format "%s: %s"
+                    (nth 1 error)
+                    (mapconcat (lambda (obj) (prin1-to-string obj t))
+                               (cdr (cdr error)) ", "))
+	  (format "%s: %s"
+                  (get (car error) 'error-message)
+                  (mapconcat (lambda (obj) (prin1-to-string obj t))
+                             (cdr error) ", "))))
+      'external-debugging-output)
+     (terpri 'external-debugging-output)
+     (setq initial-window-system nil)
+     (kill-emacs)))
+
   (run-hooks 'before-init-hook)
 
   (if (daemonp)

Re: Not loading the Registry settings

[Juanma Barranquero]

On Tue, Sep 15, 2009 at 03:31, Stefan Monnier <monnier@iro.umontreal.ca> wrote:

> Thanks.  The patch below should address this issue (it just moves code
> around).

After your patch, setting the variable in Lisp woks (see code below;
it does not include X changes, only Windows).

The only drawback is that, on Windows, environment variables (HOME,
SHELL, TERM, LANG, PRELOAD_WINSOCK and a few others) can also be
defined in the registry, and they are loaded in
w32.c:init_environment(), so -Q will not affect them. That is bad,
because it negates one of the advantages of this change, i.e., making
emacs -Q more useful for testing.

> Well, I'm not sure if it would be useful, but the name seems to reflect
> more precisely the effect of the variable, and indeed it can later on
> be changed from Lisp, tho it's not clear whether that would be useful.

BTW, the question of whether "emacs -Q -xrm myresouce" should obey
myresource or not is still open.

    Juanma


diff --git a/lisp/startup.el b/lisp/startup.el
index 72c185c..3ffc4b8 100644
--- a/lisp/startup.el
+++ b/lisp/startup.el
@@ -366,8 +366,6 @@ from being initialized."
 		 string)
   :group 'auto-save)

-(defvar emacs-quick-startup nil)
-
 (defvar emacs-basic-display nil)

 (defvar init-file-debug nil)
@@ -799,7 +797,7 @@ opening the first frame (e.g. open a connection to
an X server).")
 	 ((member argi '("-Q" "-quick"))
 	  (setq init-file-user nil
 		site-run-file nil
-		emacs-quick-startup t))
+		inhibit-x-resources t))
 	 ((member argi '("-D" "-basic-display"))
 	  (setq no-blinking-cursor t
 		emacs-basic-display t)
@@ -2272,7 +2270,7 @@ A fancy display is used on graphic displays,
normal otherwise."
     (if (or inhibit-startup-screen
 	    initial-buffer-choice
 	    noninteractive
-	    emacs-quick-startup)
+	    inhibit-x-resources)

 	;; Not displaying a startup screen.  If 3 or more files
 	;; visited, and not all visible, show user what they all are.
diff --git a/src/emacs.c b/src/emacs.c
index 2c14be5..9538e3c 100644
--- a/src/emacs.c
+++ b/src/emacs.c
@@ -239,6 +239,9 @@ int noninteractive;

 int noninteractive1;

+/* Nonzero means Emacs was run in --quick mode.  */
+int inhibit_x_resources;
+
 /* Name for the server started by the daemon.*/
 static char *daemon_name;

@@ -1483,11 +1486,6 @@ main (int argc, char **argv)
           ns_no_defaults = 1;
           skip_args--;
         }
-      if (argmatch (argv, argc, "-Q", "--quick", 5, NULL, &skip_args))
-        {
-          ns_no_defaults = 1;
-          skip_args--;
-        }
 #ifdef NS_IMPL_COCOA
       if (skip_args < argc)
         {
@@ -2680,6 +2678,10 @@ was found.  */);
 This is nil during initialization.  */);
   Vafter_init_time = Qnil;

+  DEFVAR_BOOL ("inhibit-x-resources", &inhibit_x_resources,
+	       doc: /* If non-nil, X resources and Windows Registry settings
are not used.  */);
+  inhibit_x_resources = 0;
+
   /* Make sure IS_DAEMON starts up as false.  */
   daemon_pipe[1] = 0;
 }
diff --git a/src/lisp.h b/src/lisp.h
index 820c258..f837708 100644
--- a/src/lisp.h
+++ b/src/lisp.h
@@ -3143,6 +3143,9 @@ void shut_down_emacs P_ ((int, int, Lisp_Object));
 /* Nonzero means don't do interactive redisplay and don't change tty modes */
 extern int noninteractive;

+/* Nonzero means don't load X resources or Windows Registry settings.  */
+extern int inhibit_x_resources;
+
 /* Pipe used to send exit notification to the daemon parent at
    startup.  */
 extern int daemon_pipe[2];
diff --git a/src/w32reg.c b/src/w32reg.c
index d2330e7..e31fba0 100644
--- a/src/w32reg.c
+++ b/src/w32reg.c
@@ -150,6 +150,10 @@ x_get_string_resource (rdb, name, class)
      XrmDatabase rdb;
      char *name, *class;
 {
+  if (inhibit_x_resources)
+    /* --quick was passed, so this is a no-op.  */
+    return NULL;
+
   if (rdb)
     {
       char *resource;

Re: Not loading the Registry settings

[Stefan Monnier]

>> Thanks.  The patch below should address this issue (it just moves code
>> around).

> After your patch, setting the variable in Lisp works (see code below;
> it does not include X changes, only Windows).

Good, thanks.

> The only drawback is that, on Windows, environment variables (HOME,
> SHELL, TERM, LANG, PRELOAD_WINSOCK and a few others) can also be
> defined in the registry, and they are loaded in
> w32.c:init_environment(), so -Q will not affect them. That is bad,
> because it negates one of the advantages of this change, i.e., making
> emacs -Q more useful for testing.

OTOH, it's just the same behavior as under POSIX where environment
variables are not provided via xrm but directly (and -Q does not
override them).  For HOME settings, I think it's TRT.  For other
variables, it's less clear, but I think it's OK for now.

>> Well, I'm not sure if it would be useful, but the name seems to reflect
>> more precisely the effect of the variable, and indeed it can later on
>> be changed from Lisp, tho it's not clear whether that would be useful.

> BTW, the question of whether "emacs -Q -xrm myresouce" should obey
> myresource or not is still open.

I think it *should* obey myresource.  I don't know how to implement that
behavior, tho.


        Stefan

Re: Not loading the Registry settings

[Juanma Barranquero]

On Tue, Sep 15, 2009 at 15:34, Stefan Monnier <monnier@iro.umontreal.ca> wrote:

> OTOH, it's just the same behavior as under POSIX where environment
> variables are not provided via xrm but directly (and -Q does not
> override them).  For HOME settings, I think it's TRT.  For other
> variables, it's less clear, but I think it's OK for now.

OK

> I think it *should* obey myresource.  I don't know how to implement that
> behavior, tho.

On Windows it is trivial, so I'll add it to my patch. OK, there it
goes. I've added the "if (...) return" code to
xrdb.c:x_get_string_resource, so I'd be glad if someone on GNU/Linux
can test that it works as expected (or at least that it does nothing
bad).

Once that is verified and you've installed your startup.el patch, I
can go on installing this and then Adrian can tweak for NextStep and
the X integration can be ironed out.

    Juanma


	* NEWS: Mention new behavior of -Q and new variable
	`inhibit-x-resources'.

	* startup.el (emacs-quick-startup): Remove variable and all uses.
	(command-line): Set `inhibit-x-resources' instead.
	(command-line-1): Use `inhibit-x-resources' instead.

	* emacs.c (inhibit_x_resources): New variable.
	(main) [HAVE_NS]: Don't process --quick command line option.
	(syms_of_emacs) <inhibit-x-resources>: DEFVAR_BOOL it.

	* lisp.h (inhibit_x_resources): Declare it extern.

	* w32reg.c (x_get_string_resource):
	* xrdb.c (x_get_string_resource): Obey inhibit_x_resources.


diff --git a/etc/NEWS b/etc/NEWS
index 3db3f53..9e3a2f8 100644
--- a/etc/NEWS
+++ b/etc/NEWS
@@ -38,6 +38,11 @@ world-readable install.

 * Changes in Emacs 23.2

+** Command-line option -Q (--quick) now also disables loading X resources.
+On Windows, Registry settings are ignored, though environment variables set
+on the Registry are still honored.  The new variable `inhibit-x-resources'
+shows whether X resources were loaded or not.
+
 ** New completion-style `initials' to complete M-x lch to list-command-history.

 ** Unibyte sessions are declared obsolete.
diff --git a/lisp/startup.el b/lisp/startup.el
index 72c185c..3ffc4b8 100644
--- a/lisp/startup.el
+++ b/lisp/startup.el
@@ -366,8 +366,6 @@ from being initialized."
 		 string)
   :group 'auto-save)

-(defvar emacs-quick-startup nil)
-
 (defvar emacs-basic-display nil)

 (defvar init-file-debug nil)
@@ -799,7 +797,7 @@ opening the first frame (e.g. open a connection to
an X server).")
 	 ((member argi '("-Q" "-quick"))
 	  (setq init-file-user nil
 		site-run-file nil
-		emacs-quick-startup t))
+		inhibit-x-resources t))
 	 ((member argi '("-D" "-basic-display"))
 	  (setq no-blinking-cursor t
 		emacs-basic-display t)
@@ -2272,7 +2270,7 @@ A fancy display is used on graphic displays,
normal otherwise."
     (if (or inhibit-startup-screen
 	    initial-buffer-choice
 	    noninteractive
-	    emacs-quick-startup)
+	    inhibit-x-resources)

 	;; Not displaying a startup screen.  If 3 or more files
 	;; visited, and not all visible, show user what they all are.
diff --git a/src/emacs.c b/src/emacs.c
index 2c14be5..9538e3c 100644
--- a/src/emacs.c
+++ b/src/emacs.c
@@ -239,6 +239,9 @@ int noninteractive;

 int noninteractive1;

+/* Nonzero means Emacs was run in --quick mode.  */
+int inhibit_x_resources;
+
 /* Name for the server started by the daemon.*/
 static char *daemon_name;

@@ -1483,11 +1486,6 @@ main (int argc, char **argv)
           ns_no_defaults = 1;
           skip_args--;
         }
-      if (argmatch (argv, argc, "-Q", "--quick", 5, NULL, &skip_args))
-        {
-          ns_no_defaults = 1;
-          skip_args--;
-        }
 #ifdef NS_IMPL_COCOA
       if (skip_args < argc)
         {
@@ -2680,6 +2678,10 @@ was found.  */);
 This is nil during initialization.  */);
   Vafter_init_time = Qnil;

+  DEFVAR_BOOL ("inhibit-x-resources", &inhibit_x_resources,
+	       doc: /* If non-nil, X resources and Windows Registry settings
are not used.  */);
+  inhibit_x_resources = 0;
+
   /* Make sure IS_DAEMON starts up as false.  */
   daemon_pipe[1] = 0;
 }
diff --git a/src/lisp.h b/src/lisp.h
index 820c258..f837708 100644
--- a/src/lisp.h
+++ b/src/lisp.h
@@ -3143,6 +3143,9 @@ void shut_down_emacs P_ ((int, int, Lisp_Object));
 /* Nonzero means don't do interactive redisplay and don't change tty modes */
 extern int noninteractive;

+/* Nonzero means don't load X resources or Windows Registry settings.  */
+extern int inhibit_x_resources;
+
 /* Pipe used to send exit notification to the daemon parent at
    startup.  */
 extern int daemon_pipe[2];
diff --git a/src/w32reg.c b/src/w32reg.c
index d2330e7..2b5b352 100644
--- a/src/w32reg.c
+++ b/src/w32reg.c
@@ -76,7 +76,7 @@ w32_get_rdb_resource (rdb, resource)
   return NULL;
 }

-LPBYTE
+static LPBYTE
 w32_get_string_resource (name, class, dwexptype)
      char *name, *class;
      DWORD dwexptype;
@@ -160,6 +160,10 @@ x_get_string_resource (rdb, name, class)
         return resource;
     }

+  if (inhibit_x_resources)
+    /* --quick was passed, so this is a no-op.  */
+    return NULL;
+
   return (w32_get_string_resource (name, class, REG_SZ));
 }

diff --git a/src/xrdb.c b/src/xrdb.c
index 0a74c08..fac97e4 100644
--- a/src/xrdb.c
+++ b/src/xrdb.c
@@ -693,6 +693,10 @@ x_get_string_resource (rdb, name, class)
 {
   XrmValue value;

+  if (inhibit_x_resources)
+    /* --quick was passed, so this is a no-op.  */
+    return NULL;
+
   if (x_get_resource (rdb, name, class, x_rm_string, &value))
     return (char *) value.addr;

Re: Not loading the Registry settings

[Juanma Barranquero]

On Mon, Sep 14, 2009 at 21:39, Adrian Robert <adrian.b.robert@gmail.com> wrote:

> The -Q/-q portion of this can be removed once you've implemented your
> change, since it implements the same thing.  I can clean up the rest in
> other files once this happens.

OK, I've installed it, removing the -Q part from inside the #idfef
HAVE_NS. Now please change (or fix) to your heart's content.

    Juanma