CVS HEAD still broken for OSX "make all install" for .app creation

CVS HEAD still broken for OSX "make all install" for .app creation

[Randal L. Schwartz]

The line

   Copying etc/DOC-23.0.60.2 to /Users/merlyn/MIRROR/emacs-CVS/nextstep/Emacs.app/Contents/Resources/share/emacs/23.0.60/etc ...

is broken, because there is no Resources/share in my app folder.

Thus, I'm still forced to say "make bootstrap install" and not "make all
install".  Otherwise, I get a DOC-23.0.60.2 "not found" on start up.  Oops.

Could some Smart Person please look at this and fix it (again?)?

-- 
Randal L. Schwartz - Stonehenge Consulting Services, Inc. - +1 503 777 0095
<merlyn@stonehenge.com> <URL:http://www.stonehenge.com/merlyn/>
Smalltalk/Perl/Unix consulting, Technical writing, Comedy, etc. etc.
See http://methodsandmessages.vox.com/ for Smalltalk and Seaside discussion

Re: CVS HEAD still broken for OSX

[Adrian Robert]

Randal L. Schwartz <merlyn <at> stonehenge.com> writes:

>    Copying etc/DOC-23.0.60.2 to
> /Users/merlyn/MIRROR/emacs-
CVS/nextstep/Emacs.app/Contents/Resources/share/emacs/23.0.60/etc ...
> 
> is broken, because there is no Resources/share in my app folder.
> 
> Thus, I'm still forced to say "make bootstrap install" and not "make all
> install".  Otherwise, I get a DOC-23.0.60.2 "not found" on start up.  Oops.
> 
> Could some Smart Person please look at this and fix it (again?)?


Hi,

I can't replicate it here.  Resources/share should be made by the
install-arch-indep process and would include all the lisp and
some other things besides just etc/DOC-*.
Could you post full details about your  configure invocation,
what cleanup steps you do before (make maintainer-clean, etc.),
and whether you have  any env variables set that might affect
emacs build or install?

Also, what is the history of the error -- from what you say
you got it before, then someone fixed it, now it has reappeared?

thanks,
Adrian

popup-menu with keyboard menu

[Stephen Leake]

I'm calling popup-menu with a keyboard menu, rather than from a mouse
action.

It's failing with "Wrong type argument: integerp, nil" if the mouse is
outside the Emacs frame (which it normally is, for me).

This patch fixes the problem:

===================================================================
RCS file: /sources/emacs/emacs/lisp/mouse.el,v
retrieving revision 1.347
diff -u -r1.347 mouse.el
--- lisp/mouse.el	11 Aug 2008 01:23:05 -0000	1.347
+++ lisp/mouse.el	27 Sep 2008 19:36:59 -0000
@@ -109,9 +109,7 @@
 				   (plist-get (get map 'menu-prop) :filter))))
 		    (if filter (funcall filter (symbol-function map)) map)))))
 	 event cmd)
-    (unless position
-      (let ((mp (mouse-pixel-position)))
-	(setq position (list (list (cadr mp) (cddr mp)) (car mp)))))
+    (unless position (setq position t))
     ;; The looping behavior was taken from lmenu's popup-menu-popup
     (while (and map (setq event
 			  ;; map could be a prefix key, in which case
===================================================================

'position' is later passed to x-popup-menu, which properly handles the
case of the mouse being outside the frame.

-- 
-- Stephe

Re: CVS HEAD still broken for OSX

[Randal L. Schwartz]

>>>>> "Adrian" == Adrian Robert <Adrian.B.Robert@gmail.com> writes:

Adrian> Also, what is the history of the error -- from what you say
Adrian> you got it before, then someone fixed it, now it has reappeared?

No, I haven't been able to do a "make all install" without "bootstrap"
on the new config yet.

Here's how I repeat the error, using the CVS view:

  make distclean
  ./configure --prefix=/opt/emacs --with-ns
  make bootstrap install

At this point, I have a working nextstep/Emacs.app with .1 of the DOC file
properly installed.  Now, pretending I've gone up to the next day, and have
pulled some new CVS updates, and want to do an incremental build...

  cvs up
  touch src/nsterm.m # stimulate a rebuild even if not needed for this test
  make all install

And in this pass, the .2 version of the DOC file is *not* installed properly
(although it's built properly), and thus nextstep/Emacs.app is no longer
functional.  A new "bootstrap install" works fine though, but takes a lot
longer since I'm throwing everything away every time.

Somewhere, for the incremental build, the paths are getting mangled.

-- 
Randal L. Schwartz - Stonehenge Consulting Services, Inc. - +1 503 777 0095
<merlyn@stonehenge.com> <URL:http://www.stonehenge.com/merlyn/>
Smalltalk/Perl/Unix consulting, Technical writing, Comedy, etc. etc.
See http://methodsandmessages.vox.com/ for Smalltalk and Seaside discussion

Re: popup-menu with keyboard menu

[Glenn Morris]

Stephen Leake wrote:

> I'm calling popup-menu with a keyboard menu, rather than from a mouse
> action.
>
> It's failing with "Wrong type argument: integerp, nil" if the mouse is
> outside the Emacs frame (which it normally is, for me).

Can you give a recipe starting from emacs -Q showing how to get this
error?

I couldn't reproduce it, and x-popup-menu has code to handle this
case.

Perhaps you are on Windows, since the version of x-popup-menu in
w32menu.c seems to be missing the relevant part from the xmenu.c
version.

          /* If a click happens in an external tool bar or a detached
             tool bar, x and y is NIL.  In that case, use the current
             mouse position.  This happens for the help button in the
             tool bar.  Ideally popup-menu should pass NIL to
             this function, but it doesn't.  */
          if (NILP (x) && NILP (y))
            get_current_pos_p = 1;

Re: CVS HEAD still broken for OSX

[Adrian Robert]

Randal L. Schwartz <merlyn <at> stonehenge.com> writes:

> Here's how I repeat the error, using the CVS view:
> 
>   make distclean
>   ./configure --prefix=/opt/emacs --with-ns
>   make bootstrap install

OK, thanks. I'm not sure what the behavior is with --prefix specified
but without --disable-self-contained.  Where do the lisp and the
Emacs.app end up after the make invocation?

Could you try either with no --prefix or
adding --disable-ns-self-contained
to the configure invocation?

If this fixes it we should probably auto-disable
self-contained if a prefix is set.


thanks,
Adrian

Re: CVS HEAD still broken for OSX

[Randal L. Schwartz]

>>>>> "Adrian" == Adrian Robert <Adrian.B.Robert@gmail.com> writes:

Adrian> Randal L. Schwartz <merlyn <at> stonehenge.com> writes:
>> Here's how I repeat the error, using the CVS view:
>> 
>> make distclean
>> ./configure --prefix=/opt/emacs --with-ns
>> make bootstrap install

Adrian> OK, thanks. I'm not sure what the behavior is with --prefix specified
Adrian> but without --disable-self-contained.  Where do the lisp and the
Adrian> Emacs.app end up after the make invocation?

They end up inside the app.  I just wanted to make sure it *wasn't* going to
install it in some wonky /usr/local location otherwise.  And it doesn't even
try to touch /opt/emacs, so this is already good.

Adrian> Could you try either with no --prefix or
Adrian> adding --disable-ns-self-contained
Adrian> to the configure invocation?

Will do.

-- 
Randal L. Schwartz - Stonehenge Consulting Services, Inc. - +1 503 777 0095
<merlyn@stonehenge.com> <URL:http://www.stonehenge.com/merlyn/>
Smalltalk/Perl/Unix consulting, Technical writing, Comedy, etc. etc.
See http://methodsandmessages.vox.com/ for Smalltalk and Seaside discussion

Re: CVS HEAD still broken for OSX

[Randal L. Schwartz]

>>>>> "Randal" == Randal L Schwartz <merlyn@stonehenge.com> writes:

Adrian> Could you try either with no --prefix or
Adrian> adding --disable-ns-self-contained
Adrian> to the configure invocation?

Randal> Will do.

No, same response, still installs (or tries to) into the wrong dir:

   Copying etc/DOC-23.0.60.2 to /Users/merlyn/MIRROR/emacs-CVS/nextstep/Emacs.app/Contents/Resources/share/emacs/23.0.60/etc

That's the wrong path for the self-contained app.  So leaving out --prefix
didn't matter.  This path is wrong when I do "make all install".

-- 
Randal L. Schwartz - Stonehenge Consulting Services, Inc. - +1 503 777 0095
<merlyn@stonehenge.com> <URL:http://www.stonehenge.com/merlyn/>
Smalltalk/Perl/Unix consulting, Technical writing, Comedy, etc. etc.
See http://methodsandmessages.vox.com/ for Smalltalk and Seaside discussion

Re: popup-menu with keyboard menu

[Stephen Leake]

Glenn Morris <rgm@gnu.org> writes:

> Stephen Leake wrote:
>
>> I'm calling popup-menu with a keyboard menu, rather than from a mouse
>> action.
>>
>> It's failing with "Wrong type argument: integerp, nil" if the mouse is
>> outside the Emacs frame (which it normally is, for me).
>
> Can you give a recipe starting from emacs -Q showing how to get this
> error?

Interestingly, I can't.

> I couldn't reproduce it, and x-popup-menu has code to handle this
> case.

The bug was in popup-menu, not x-popup-menu; the fix was to rely on
x-popup-menu to do the right thing.

But it turns out popup-menu doesn't do what I want anyway; it executes
the menu keymap as a GUI menu, not a keyboard menu. Which makes sense,
given the name :).

How do I invoke a keyboard menu from lisp?

-- 
-- Stephe

Re: popup-menu with keyboard menu

[Glenn Morris]

Stephen Leake wrote:

> The bug was in popup-menu, not x-popup-menu; the fix was to rely on
> x-popup-menu to do the right thing.

But popup-menu does not use `position' for anything, it just passes it
to x-popup-menu. So I don't so how you can get an error from a bad
position element in popup-menu itself.

Re: popup-menu with keyboard menu

[Stephen Leake]

Glenn Morris <rgm@gnu.org> writes:

> Stephen Leake wrote:
>
>> The bug was in popup-menu, not x-popup-menu; the fix was to rely on
>> x-popup-menu to do the right thing.
>
> But popup-menu does not use `position' for anything, it just passes it
> to x-popup-menu. So I don't so how you can get an error from a bad
> position element in popup-menu itself.

'position' is passed to x-popup-menu; it was x-popup-menu that was
raising the error.

The problem is that when the mouse is outside the window, 'position'
would have nil in it for the actual position. Then x-popup-window
would try to convert that to an integer.

But since it's not failing with emacs -Q, I must have something that's
redefining mouse-position and returning incorrect results, that
x-popup-menu is then getting confused by.

-- 
-- Stephe

Re: CVS HEAD still broken for OSX

[Adrian Robert]

Randal L. Schwartz <merlyn <at> stonehenge.com> writes:

> 
> >>>>> "Randal" == Randal L Schwartz <merlyn <at> stonehenge.com> writes:
> 
> Adrian> Could you try either with no --prefix or
> Adrian> adding --disable-ns-self-contained
> Adrian> to the configure invocation?
> 
> Randal> Will do.
> 
> No, same response, still installs (or tries to) into the wrong dir:
> 
>    Copying etc/DOC-23.0.60.2 to /Users/merlyn/MIRROR/emacs-
CVS/nextstep/Emacs.app/Contents/Resources/share/emacs/23.0.60/etc
> 
> That's the wrong path for the self-contained app.  So leaving out --prefix
> didn't matter.  This path is wrong when I do "make all install".

OK, I'd say the path is correct but the directory is not being made because
the "install" target is for some reason deciding it doesn't need to be.
(The entire subtree under Resources gets rearranged by NS's install-arch-dep,
but the subsequent "install" somehow doesn't see this.)

For now, you can use my approach to update-installs which is:

#!/bin/sh
cd lisp
make -j3
cd ..
rm -f etc/DOC-* src/emacs-*
make -j3

This suppresses the strange, linux-kernel-like versioning added to
DOC and emacs by resetting it.  (Who here is making use of the
historic copies and for what?)  DOC in the app package might not
actually get updated though, so a better solution should be implemented.

I guess the make-lisp step could be added to the remake script..

Re: popup-menu with keyboard menu

[Glenn Morris]

Stephen Leake wrote:

> 'position' is passed to x-popup-menu; it was x-popup-menu that was
> raising the error.
>
> The problem is that when the mouse is outside the window, 'position'
> would have nil in it for the actual position.

I can only say again that x-popup-menu seems to have code to handle
the case where X and Y are nil, and that I don't get any error in such
a case. Again, the relevant code seems to be missing from the version
in w32menu.c.