12:19 p.m.
Version: all known, including 21.4, 21.5.
Frames without minibuffer are not displayed correctly on X device.
The lowest line of non-minibuffer window in the frame gets overlapped
with modeline as shown in an attached window dump.
Redisplay related variables are as follows.
(specifier-spec-list menubar-visible-p)
((global (nil)))
(default-toolbar-position)
top
(specifier-spec-list default-toolbar-visible-p)
((global (nil)))
(specifier-spec-list toolbar-buttons-captioned-p)
((global (nil . t)))
7:32 p.m.
New subject: redisplay w/o minibuffer broken - 2
Version: 21.4
Window system: X
When `make-frame' is called as in `ediff-setup-control-frame' with
`nil' value of `minibuffer' property, xemacs may calculate `f->height'
incorrectly.
With `modeline-shadow-thickness' globally 0 and default face font
height 13 one newly created frame may have `f->pixheight' 21 and
`f->height' 2. That many characters do not fit in frame pixel height.
So either text area or modeline of frame root window is (nearly)
invisible.
Sometimes it even calculates heights reasonably: pixheight = 34 and
height = 2. But echo area in that frame gets displayed and overlaps
modeline text.
8:05 p.m.
New subject: redisplay w/o minibuffer broken - 3
Version: 21.4
Window system: X
One obviously broken thing: in the following call sequence:
#0 change_frame_size_1 (f=0x8ae1d60, newheight=2, newwidth=3)
at src/frame.c:3070
#1 0x08156789 in change_frame_size (f=0x8ae1d60, newheight=2, newwidth=3,
delay=0) at src/frame.c:3146
#2 0x08230d9f in EmacsFrameResize (widget=0x853f670)
at src/EmacsFrame.c:387
#3 0x401339ec in ResizeWrapper () from /usr/X11R6/lib/libXm.so.2
#4 0x401337d8 in ResizeWrapper4 () from /usr/X11R6/lib/libXm.so.2
#5 0x402afba7 in XtSetValues () from /usr/X11R6/lib/libXt.so.6
#6 0x0823153d in EmacsFrameSetCharSize (widget=0x853f670, columns=1, rows=1)
at src/EmacsFrame.c:673
#7 0x082087c1 in x_set_initial_frame_size (f=0x8ae1d60, flags=15, x=1025,
y=769, w=1, h=1) at src/frame-x.c:715
#8 0x08209634 in x_set_frame_properties (f=0x8ae1d60, plist=145947700)
at src/frame-x.c:927
#9 0x08151cd8 in Fset_frame_properties (frame=145628512, plist=145947700)
at src/frame.c:2375
#10 0x0814d171 in Fmake_frame (props=145991868, device=1081545268)
at src/frame.c:480
`change_frame_size_1' makes `f->height' equal to `newheight', but
`f->pixheight' never gets updated, leaving frame sizes inconsistent as
described previously.
Cannot reproduce that when catching frame heights changes with
debugger. Apparently there is some sort of race condition.
4:36 p.m.
Version: 21.4
Window system: X
Adding
{ int ph; char_to_pixel_size (f, f->width, f->height, NULL, &ph); assert
(f->pixheight >= ph); /* have enough space for all chars */ }
after resetting `f->size_slipped' in `redisplay_frame' shows that
ill-formed frames really get into `redisplay_frame'. They have
`pixheight' and `height' inconsistent and require size adjusting, but
`size_slipped' is not set. Setting `f->size_slipped' in them by
debugger make them look normally.
The trouble is that when debugger printing between creating frame and
redisplaying it is on, obtaining such a frames can not be reproduced.
What functions are intended to change frame sizes and set
`f->size_slipped'? In particular, where must such a setting occur
during `Fmake_frame' when frame height property is specified? XEmacs
experts?
8:35 p.m.
Broken `height' value of 26 gets there through `XtConfigureWidget'
call inside `x_layout_widgets', which call is really caused by
`sit-for' with some arbitrary timeout value (so race conditions are
likely). The backtrace follows.
Just after `change_frame_size' call in `EmacsFrameResize' was added
the following:
{
int ph;
char_to_pixel_size (f, f->width, f->height, NULL, &ph);
assert (f->pixheight >= ph || f->size_slipped);
}
so that xemacs does not try to set inconsistent frame measurements
unless `f->size_slipped' ensures that the next `redisplay_frame' will
fix them.
Is `x_layout_widgets' really supposed to leave frame measurements
inconsistent without scheduling to fix them on redisplay?
#1 0x082315ac in EmacsFrameResize (widget=0x8967c10)
at src/EmacsFrame.c:391
#2 0x401339ec in ResizeWrapper () from /usr/X11R6/lib/libXm.so.2
#3 0x401337d8 in ResizeWrapper4 () from /usr/X11R6/lib/libXm.so.2
#4 0x4029e57c in XtConfigureWidget () from /usr/X11R6/lib/libXt.so.6
#5 0x0820b5a8 in x_layout_widgets (w=0x8cc4f50, client_data=0x8e8b2c8,
call_data=0xbfffe124)
at src/frame-x.c:1814
#6 0x4028f716 in XtCallCallbackList () from /usr/X11R6/lib/libXt.so.6
#7 0x08232759 in Resize (w=0x8cc4f50)
at src/EmacsManager.c:176
#8 0x401339ec in ResizeWrapper () from /usr/X11R6/lib/libXm.so.2
#9 0x40133800 in ResizeWrapper6 () from /usr/X11R6/lib/libXm.so.2
#10 0x4029e57c in XtConfigureWidget () from /usr/X11R6/lib/libXt.so.6
#11 0x4009ab81 in XmeConfigureObject () from /usr/X11R6/lib/libXm.so.2
#12 0x40142d1d in Resize () from /usr/X11R6/lib/libXm.so.2
#13 0x401339ec in ResizeWrapper () from /usr/X11R6/lib/libXm.so.2
#14 0x40133828 in ResizeWrapper8 () from /usr/X11R6/lib/libXm.so.2
#15 0x40140d58 in StructureNotifyHandler () from /usr/X11R6/lib/libXm.so.2
#16 0x4029ad68 in XtDispatchEventToWidget () from /usr/X11R6/lib/libXt.so.6
#17 0x4029b730 in _XtOnGrabList () from /usr/X11R6/lib/libXt.so.6
#18 0x4029baa7 in XtDispatchEvent () from /usr/X11R6/lib/libXt.so.6
#19 0x402a692e in XtAppProcessEvent () from /usr/X11R6/lib/libXt.so.6
#20 0x082067c4 in drain_X_queue ()
at src/event-Xt.c:2985
#21 0x08206929 in emacs_Xt_event_pending_p (user_p=1)
at src/event-Xt.c:3110
#22 0x0811ed01 in Fsit_for (seconds=139618876, nodisplay=1081545268)
at src/event-stream.c:438
# (unwind-protect ...)
sit-for(0.1)
# bind (count window-min-height ctl-frame-iconified-p dont-iconify-ctl-frame
deiconify-ctl-frame ctl-frame old-ctl-frame lines fheight fwidth adjusted-parameters
designated-minibuffer-frame ctl-buffer)
ediff-setup-control-frame(#<buffer "*Ediff Control Panel*"> #<x-frame
"emacs" 0x82b>)
# bind (done-C done-B done-A designated-minibuffer-frame wind-B-start wind-A-start
use-same-frame orig-wind three-way-comparison split-window-function force-one-frame
ctl-frame-exists-p frame-C frame-B frame-A wind-C wind-B wind-A window-min-height
control-buf buf-C buf-B buf-A)
ediff-setup-windows-multiframe-compare(#<buffer "scratch-db">
#<buffer ".gdbinit"> nil #<buffer "*Ediff Control
Panel*">)
# bind (control-buf buf-C buf-B buf-A)
ediff-setup-windows-multiframe(#<buffer "scratch-db"> #<buffer
".gdbinit"> nil #<buffer "*Ediff Control Panel*">)
# bind (control-buffer buffer-C buffer-B buffer-A)
ediff-setup-windows(#<buffer "scratch-db"> #<buffer
".gdbinit"> nil #<buffer "*Ediff Control Panel*">)
# (unwind-protect ...)
# bind (control-buffer control-buffer-name merge-buffer-file setup-parameters
startup-hooks file-C buffer-C file-B buffer-B file-A buffer-A)
ediff-setup(#<buffer "scratch-db"> "/tmp/gin/scratch-db1VA7vn"
#<buffer ".gdbinit"> "/tmp/gin/.gdbinit1VAI6t" nil nil ((lambda
nil (delete-file "/tmp/gin/scratch-db1VA7vn") (delete-file
"/tmp/gin/.gdbinit1VAI6t") (if ... ...))) ((ediff-job-name . ediff-buffers))
nil)
# bind (file-C file-B file-A buf-C-file-name buf-C-is-alive buf-B-file-name
buf-A-file-name merge-buffer-file job-name startup-hooks buf-C buf-B buf-A)
ediff-buffers-internal("scratch-db" ".gdbinit" nil nil
ediff-buffers)
# bind (job-name startup-hooks buffer-B buffer-A)
#<compiled-function (buffer-A buffer-B &optional startup-hooks job-name)
"...(13)" [startup-hooks buffer-B buffer-A job-name ediff-buffers
ediff-buffers-internal nil] 6
("/home/gin/xemacs-21.4-l/xemacs-packages/lisp/ediff/ediff.elc" . 5607) (let
(bf) (list ... ...))>("scratch-db" ".gdbinit")
call-interactively(ediff-buffers)
command-execute(ediff-buffers t)
# bind (_execute_command_keys_ _execute_command_name_ prefix-arg)
execute-extended-command(nil)
# bind (command-debug-status)
call-interactively(execute-extended-command)
# (condition-case ... . error)
# (catch top-level ...)
12:05 p.m.
>>>> "Ilya" == Ilya N Golubev
<gin(a)mo.msk.ru> writes:
Ilya> Is `x_layout_widgets' really supposed to leave frame
Ilya> measurements inconsistent without scheduling to fix them on
Ilya> redisplay?
I doubt it, but that's probably originally Andy Piper code. I'll try
and take a look at it soon, if Andy doesn't say something, or Mike
Sperber.
--
Institute of Policy and Planning Sciences http://turnbull.sk.tsukuba.ac.jp
University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Ask not how you can "do" free software business;
ask what your business can "do for" free software.
7791
days inactive
7802
days old
5 comments
2 participants
participants (2)
-
Ilya N. Golubev -
Stephen J. Turnbull