6:16 p.m.
This fixes the selection crash that people are seeing and makes my
selection optimisation work properly. The main problems were XtNameToWidget
not handling `.' and Fset_window_configuration nuking the subwindow
instance cache.
There is still a redisplay glitch but that appears to be something to do
with the actual gutter/glyph redisplay code rather than the glyph handling
changes I have made recently. I will fix when I can figure out what is
going wrong.
Martin - unless this leads to more crashes I would go ahead and release 21.2.36
Thanks
andy
2000-08-18 Andy Piper <andy(a)xemacs.org>
* frame.c (delete_frame_internal): use new
free_frame_subwindow_instances name.
* glyphs-msw.c (mswindows_tab_control_instantiate): verify index.
(add_tab_item): make return type correct.
(mswindows_tab_control_instantiate): assert index of tab.
(mswindows_tab_control_redisplay): Re-code to use
gui_item_equal_sans_selected and gui_item_list_find_selected.
* glyphs-widget.c (tab_control_update): Correct comment.
* window.c (window_unmap_subwindows): use new
unmap_subwindow_instance_cache_mapper.
(window_unmap_subwindows_cache_mapper): deleted.
(Fset_window_configuration): comparisons should now be with
EQ. Preserve the subwindow instance cache across configuration
changes.
(allocate_window): ditto.
(make_dummy_parent): ditto.
* glyphs.c (free_frame_subwindow_instances): rename from
free_frame_subwindow_instance_cache. finalize all instances rather
than just those in the display cache.
(finalize_all_subwindow_instances): walk windows unmapping and
finalizing subwindows.
(unmap_subwindow_instance_cache_mapper): moved from
window.c. Allow finalization as well as unmapping.
* gui.c (gui_item_list_find_selected): new function.
* gui.h (gui_item_list_find_selected): declare.
* glyphs-x.c (x_tab_control_redisplay): pick tab
explicitly. Re-code to use gui_item_equal_sans_selected and
gui_item_list_find_selected.
* glyphs-x.h: add lwlib-utils.h
* buffer.c (Frecord_buffer): undo previous change.
--------------------------------------------------------------
Dr Andy Piper
Principal Consultant, BEA Systems Ltd
5:06 p.m.
New subject: Tab selection crash fix ... NOT!! "Houston, we, uh, crashed."
21.2.35, cvs updated about three hours ago.
>>>> "Andy" == Andy Piper
<andy(a)xemacs.org> writes:
Andy> This fixes the selection crash that people are seeing and
Andy> makes my selection optimisation work properly. The main
Andy> problems were XtNameToWidget not handling `.' and
Andy> Fset_window_configuration nuking the subwindow instance
Andy> cache.
Andy> There is still a redisplay glitch but that appears to be
Andy> something to do with the actual gutter/glyph redisplay code
Andy> rather than the glyph handling changes I have made
Andy> recently. I will fix when I can figure out what is going
Andy> wrong.
OK, I haven't yet generated a crash. That's good.
Famous last words. I just got a complete lockup, had to kill xemacs
by hand. Sorry, Andy.
This is all still very very bogus. :(
**** What I remember about my first run:
Editres tells me I have two tabs widgets with the same buffer list.
The lists are not kept in synch when new buffer in a different mode is
added. Only one of the two is affected.
This may be associated with the fact that I started up XEmacs with
several file arguments, so I had a split frame to start with.
Tab controls now obey the default face (impressive) but they wait
until you select them. Evidently code needs to be added to notify
them about face changes. A separate issue is that this also may take
several attempts, I assume because of the multiple tab widget synching
problem.
The synching problem shows up as "received an unexpected
SelectionNotify event" messages rather than a crash. I also see a
garbled tab control.
I'm also getting Didier's (wasn't it?) BadFont and BadGC errors. This
seems to be associated with the garbled tab control.
Once I got the garbled tab control, the first attempt to switch
buffers via C-x b resulted in the lock-up after a few dozen more BadGC
errors.
**** Second try (recorded as I do it):
`src/xemacs -vanilla &', then `Get Tree' in editres generates
"received an unexpected SelectionNotify event". Split the window with
C-x 4 b RET, do editres `Refresh Tree', and get another "unexpected
SelectionNotify" message, and now have two buffers tabs (with the same
name, "Buffers"). Split the lower window with C-x 2, and get the same
result, ie, now I have three tab control widgets named "Buffers".
Oops, this time changing the default face (Options | Font Size | 16,
then Options | Fonts | Lucidatypewriter) only took effect when a new
window was created. After a few `find-file's, coalescing the
horizontally split windows into one and then splitting again with C-x
2, I now have 4 top gutters and four buffer tab controls, two with 2
and two with 3 buffer tabs. Urk.
This time the "unexpected SelectionNotify" messages all seemed to be
associated with editres `Refresh Tree' commands.
I've managed to get garbled tab control again, but no BadFont/BadGC,
and no lockup. So apparently it's a different issue. Far too much
tab and buffer activity to make a recipe :(
I've noticed editres will never admit that any of the tabs themselves
are mapped, only the tab control parents, although it does show the
tabs in the widget tree.
Third try (recorded as I do it):
OK, maybe I've screwed up something. Everything I can see says my
build is fine, it's XEmacs that's sick, but.... cvs status said my
(locally autoconf'd) configure is different from the repository, rm it
and cvs up; everything else is Up-To-Date. config.status --recheck;
make beta; src/xemacs & ... yup, splitting windows still instantiates
a tab control every time. Yeech.
At this point, I don't have a recipe to reproduce the crash, but I'll
try and come up with one.
--------------------------------------------------------------------------------
Here's the visual evidence of garbled tab control:
--------------------------------------------------------------------------------
Here's my Installation:
uname -a: Linux tleepslib 2.2.10 #2 Thu Jun 24 19:22:23 JST 1999 i686 unknown
/coda/Projects/XEmacs/21.2-HEAD/configure
'--srcdir=/coda/Projects/XEmacs/21.2-HEAD' '--with-widgets=athena'
'--with-athena=3d' '--with-sound=none,nas' '--with-mule'
'--with-xim=xlib' '--with-xfs' '--use-union-type'
'--pdump'
XEmacs 21.2-b35 "Nike" configured for `i686-pc-linux'.
Compilation / Installation:
Source code location: /coda/Projects/XEmacs/21.2-HEAD
Installation prefix: /usr/local
Operating system description file: `s/linux.h'
Machine description file: `m/intel386.h'
Compiler: gcc -g -O3 -Wall -Wno-switch -Wpointer-arith -Winline
-Wmissing-prototypes -Wshadow
Relocating allocator for buffers: no
GNU version of malloc: yes
- Using Doug Lea's new malloc from the GNU C Library.
Window System:
Compiling in support for the X window system:
- X Windows headers location: /usr/X11R6/include
- X Windows libraries location: /usr/X11R6/lib
- Handling WM_COMMAND properly.
Compiling in support for the Athena widget set:
- Athena headers location: X11/Xaw3d
- Athena library to link: Xaw3d
Using Lucid menubars.
Using Lucid scrollbars.
Using Athena dialog boxes.
Using Athena native widgets.
TTY:
Compiling in support for ncurses.
Images:
Compiling in support for GIF images (builtin).
Compiling in support for XPM images.
Compiling in support for PNG images.
Compiling in support for JPEG images.
Compiling in support for TIFF images.
Sound:
Compiling in support for NAS (network audio system).
Databases:
Compiling in support for Berkeley database.
Compiling in support for PostgreSQL.
- Using PostgreSQL header file: postgresql/libpq-fe.h
- Using PostgreSQL V7 bindings.
Internationalization:
Compiling in support for Mule (multi-lingual Emacs).
Compiling in support for XIM (X11R5+ I18N input method).
- Using raw Xlib to provide XIM support.
- Using XFontSet to provide bilingual menubar.
Compiling in support for Canna on Mule.
Compiling in support for the WNN input method on Mule.
Mail:
Compiling in support for "dot-locking" mail spool file locking method.
Other Features:
Compiling in support for dynamic shared object modules.
Using the union type for Lisp_Objects.
Using the new portable dumper.
Compiling in support for extra debugging code.
WARNING: ---------------------------------------------------------
WARNING: Compiling in support for runtime error checking.
WARNING: XEmacs will run noticeably more slowly as a result.
WARNING: Error checking is on by default for XEmacs beta releases.
WARNING: ---------------------------------------------------------
--
University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Institute of Policy and Planning Sciences Tel/fax: +81 (298) 53-5091
_________________ _________________ _________________ _________________
What are those straight lines for? "XEmacs rules."
10:44 p.m.
New subject: Tab selection crash fix ... NOT!! "Houston, we, uh, crashed."
I have some more fixes, but I''m not sure that they will fix all these. The
SelectionNotify thing looks very weird.
andy
At 01:06 AM 8/23/00 +0900, Stephen J. Turnbull wrote:
21.2.35, cvs updated about three hours ago.
>>>>> "Andy" == Andy Piper <andy(a)xemacs.org> writes:
Andy> This fixes the selection crash that people are seeing and
Andy> makes my selection optimisation work properly. The main
Andy> problems were XtNameToWidget not handling `.' and
Andy> Fset_window_configuration nuking the subwindow instance
Andy> cache.
Andy> There is still a redisplay glitch but that appears to be
Andy> something to do with the actual gutter/glyph redisplay code
Andy> rather than the glyph handling changes I have made
Andy> recently. I will fix when I can figure out what is going
Andy> wrong.
OK, I haven't yet generated a crash. That's good.
Famous last words. I just got a complete lockup, had to kill xemacs
by hand. Sorry, Andy.
This is all still very very bogus. :(
**** What I remember about my first run:
Editres tells me I have two tabs widgets with the same buffer list.
The lists are not kept in synch when new buffer in a different mode is
added. Only one of the two is affected.
This may be associated with the fact that I started up XEmacs with
several file arguments, so I had a split frame to start with.
Tab controls now obey the default face (impressive) but they wait
until you select them. Evidently code needs to be added to notify
them about face changes. A separate issue is that this also may take
several attempts, I assume because of the multiple tab widget synching
problem.
The synching problem shows up as "received an unexpected
SelectionNotify event" messages rather than a crash. I also see a
garbled tab control.
I'm also getting Didier's (wasn't it?) BadFont and BadGC errors. This
seems to be associated with the garbled tab control.
Once I got the garbled tab control, the first attempt to switch
buffers via C-x b resulted in the lock-up after a few dozen more BadGC
errors.
**** Second try (recorded as I do it):
`src/xemacs -vanilla &', then `Get Tree' in editres generates
"received an unexpected SelectionNotify event". Split the window with
C-x 4 b RET, do editres `Refresh Tree', and get another "unexpected
SelectionNotify" message, and now have two buffers tabs (with the same
name, "Buffers"). Split the lower window with C-x 2, and get the same
result, ie, now I have three tab control widgets named "Buffers".
Oops, this time changing the default face (Options | Font Size | 16,
then Options | Fonts | Lucidatypewriter) only took effect when a new
window was created. After a few `find-file's, coalescing the
horizontally split windows into one and then splitting again with C-x
2, I now have 4 top gutters and four buffer tab controls, two with 2
and two with 3 buffer tabs. Urk.
This time the "unexpected SelectionNotify" messages all seemed to be
associated with editres `Refresh Tree' commands.
I've managed to get garbled tab control again, but no BadFont/BadGC,
and no lockup. So apparently it's a different issue. Far too much
tab and buffer activity to make a recipe :(
I've noticed editres will never admit that any of the tabs themselves
are mapped, only the tab control parents, although it does show the
tabs in the widget tree.
Third try (recorded as I do it):
OK, maybe I've screwed up something. Everything I can see says my
build is fine, it's XEmacs that's sick, but.... cvs status said my
(locally autoconf'd) configure is different from the repository, rm it
and cvs up; everything else is Up-To-Date. config.status --recheck;
make beta; src/xemacs & ... yup, splitting windows still instantiates
a tab control every time. Yeech.
At this point, I don't have a recipe to reproduce the crash, but I'll
try and come up with one.
--------------------------------------------------------------------------------
Here's the visual evidence of garbled tab control:
--------------------------------------------------------------------------------
Here's my Installation:
uname -a: Linux tleepslib 2.2.10 #2 Thu Jun 24 19:22:23 JST 1999 i686 unknown
/coda/Projects/XEmacs/21.2-HEAD/configure
'--srcdir=/coda/Projects/XEmacs/21.2-HEAD' '--with-widgets=athena'
'--with-athena=3d' '--with-sound=none,nas' '--with-mule'
'--with-xim=xlib'
'--with-xfs' '--use-union-type' '--pdump'
XEmacs 21.2-b35 "Nike" configured for `i686-pc-linux'.
Compilation / Installation:
Source code location: /coda/Projects/XEmacs/21.2-HEAD
Installation prefix: /usr/local
Operating system description file: `s/linux.h'
Machine description file: `m/intel386.h'
Compiler: gcc -g -O3 -Wall -Wno-switch
-Wpointer-arith -Winline -Wmissing-prototypes -Wshadow
Relocating allocator for buffers: no
GNU version of malloc: yes
- Using Doug Lea's new malloc from the GNU C Library.
Window System:
Compiling in support for the X window system:
- X Windows headers location: /usr/X11R6/include
- X Windows libraries location: /usr/X11R6/lib
- Handling WM_COMMAND properly.
Compiling in support for the Athena widget set:
- Athena headers location: X11/Xaw3d
- Athena library to link: Xaw3d
Using Lucid menubars.
Using Lucid scrollbars.
Using Athena dialog boxes.
Using Athena native widgets.
TTY:
Compiling in support for ncurses.
Images:
Compiling in support for GIF images (builtin).
Compiling in support for XPM images.
Compiling in support for PNG images.
Compiling in support for JPEG images.
Compiling in support for TIFF images.
Sound:
Compiling in support for NAS (network audio system).
Databases:
Compiling in support for Berkeley database.
Compiling in support for PostgreSQL.
- Using PostgreSQL header file: postgresql/libpq-fe.h
- Using PostgreSQL V7 bindings.
Internationalization:
Compiling in support for Mule (multi-lingual Emacs).
Compiling in support for XIM (X11R5+ I18N input method).
- Using raw Xlib to provide XIM support.
- Using XFontSet to provide bilingual menubar.
Compiling in support for Canna on Mule.
Compiling in support for the WNN input method on Mule.
Mail:
Compiling in support for "dot-locking" mail spool file locking method.
Other Features:
Compiling in support for dynamic shared object modules.
Using the union type for Lisp_Objects.
Using the new portable dumper.
Compiling in support for extra debugging code.
WARNING: ---------------------------------------------------------
WARNING: Compiling in support for runtime error checking.
WARNING: XEmacs will run noticeably more slowly as a result.
WARNING: Error checking is on by default for XEmacs beta releases.
WARNING: ---------------------------------------------------------
--
University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Institute of Policy and Planning Sciences Tel/fax: +81 (298) 53-5091
_________________ _________________ _________________ _________________
What are those straight lines for? "XEmacs rules."
--------------------------------------------------------------
Dr Andy Piper
Principal Consultant, BEA Systems Ltd
11:22 a.m.
New subject: Tab selection crash fix ... NOT!! "Houston, we, uh, crashed."
At 01:06 AM 8/23/00 +0900, Stephen J. Turnbull wrote:
Editres tells me I have two tabs widgets with the same buffer list.
The lists are not kept in synch when new buffer in a different mode is
added. Only one of the two is affected.
This is correct since you can only ever see one.
This may be associated with the fact that I started up XEmacs with
several file arguments, so I had a split frame to start with.
Right - you only get more than 1 when you have more than one domain (window).
Tab controls now obey the default face (impressive) but they wait
until you select them. Evidently code needs to be added to notify
them about face changes. A separate issue is that this also may take
several attempts, I assume because of the multiple tab widget synching
problem.
The synching problem shows up as "received an unexpected
SelectionNotify event" messages rather than a crash. I also see a
garbled tab control.
Please try the attached patch which fixes all sync problems for me. The
basic problem is that by selecting a tab you can change its displayed state
but not its stored state if in the process you switch windows and hence get
another tab.
This was very difficult to debug .....
I'm also getting Didier's (wasn't it?) BadFont and BadGC
errors. This
seems to be associated with the garbled tab control.
Please try the patch, but I have never seen these problems so I am not hopeful.
andy
2000-08-23 Andy Piper <andy(a)xemacs.org>
* gui-msw.c (mswindows_handle_gui_wm_command): set
action_occurred.
* gui-x.c (popup_selection_callback): ditto.
* glyphs.h (IMAGE_INSTANCE_WIDGET_ACTION_OCCURRED): new accessor.
(XIMAGE_INSTANCE_WIDGET_ACTION_OCCURRED): ditto.
(struct Lisp_Image_Instance): add action_occurred flag.
* glyphs.c (redisplay_subwindow): use action_occurred flag.
(image_instance_changed): ditto.
(reset_frame_subwindow_instance_cache): only unmap windows - do
not remove them from the cache also.
* glyphs-widget.c (tab_control_update): better debug.
(progress_gauge_update): ditto.
(layout_update): ditto.
(layout_instantiate): ditto.
(tab_control_order_only_changed): cope with null pending items.
* glyphs-msw.c (mswindows_tab_control_redisplay): add better
debug. Force selection of an item when an action occurred. Cope
with null pending_items.
(mswindows_progress_gauge_redisplay): better debug.
* glyphs-x.c (x_tab_control_redisplay): ditto.
* redisplay.c (redisplay_frame): reset the frame cache if the
frame is garbaged.
* window.c (Fset_window_configuration): potentially re-enable
frame cache reset.
(window_unmap_subwindows): need to finalize instances here since
it is only used in mark_window_as_deleted.
--------------------------------------------------------------
Dr Andy Piper
Principal Consultant, BEA Systems Ltd
8858
days inactive
8863
days old
3 comments
2 participants
participants (2)
-
Andy Piper -
Stephen J. Turnbull