Tk Source Code

View Ticket
Login
Ticket UUID: 572995
Title: Fix/test wm geom command
Type: Patch Version: None
Submitter: mdejong Created on: 2002-06-24 07:26:00
Subsystem: 65. Generic Window Operations Assigned To: hobbs
Priority: 5 Medium Severity:
Status: Pending Last Modified: 2026-10-04 07:56:20
Resolution: None Closed By:
    Closed on:
Description:
This patch adds tests to the cross platform wm.test
file while removing them from unixWm.test. These
tests would have made it easier to catch bug
571600, for example.

This patch does make one change, in the sense
that something that did not raise an error in
the past will raise an error in 8.4.

A geom command like:

wm geom . +1+-10

Is not valid. The implementation of ParseGeometry
would incorrectly parse this y coord as a negative
number even though the y coord sign (saved in
the wmPtr->flags as WM_NEGATIVE_Y) would not
be set.

There were other variations like "+1--10" that
would not raise an error while parsing but would
not act correctly either. The docs for the geom
command state that only a single - or + character
can come before a x or y value.
User Comments: serhiy.storchaka added on 2026-10-04 07:56:20:

XParseGeometry accepts the same syntax: the offset after + or - can itself be signed, as in +1+-10 or +1--10, with the same meaning as in Tk. It also accepts + as the sign of the offset (+1++10) and X as the size separator, which Tk rejects. The wm manual page does not document signed offsets.


dkf added on 2002-06-24 19:19:58:
Logged In: YES 
user_id=79902

With experience, [wm geom . +1+-10] is necessary for those
(hopefully rare) times when a window needs to be placed off
one of the edges.  The problem comes when you have some code
to center a window, and the window is wider (or taller) than
the screen; the obvious way of doing it (see below) fails.

  set x [expr {([winfo screenwidth .]-[winfo width .])/2}]
  # Ditto for y...
  wm geometry . [format {+%d+%d} $x $y]

I've had applications that failed on laptops (embarrasingly,
in front of a demo to my boss) because Tk threw an error,
and I don't want to go through *that* loop again.

mdejong added on 2002-06-24 15:18:24:
Logged In: YES 
user_id=90858

Here is what the current docs have to say about it:

wm geometry window ?newGeometry?
If newGeometry is specified, then the  geometry  of
window  is changed and an empty string is returned.
Otherwise  the  current  geometry  for  window   is
returned  (this  is the most recent geometry speci-
fied either by manual resizing or in a wm  geometry
command).    NewGeometry   has  the  form  =widthx-
height+-x+-y, where  any  of  =,  widthxheight,  or
+-x+-y  may  be  omitted.   Width  and  height  are
positive integers specifying the desired dimensions
of window.

The older docs read "±x±y" (I don't know how that will
show up on your machine, but on mine it looks like a plus
above a minus). At any rate, the point is that the
x and y values should be positive. If you want to declare
a negative y value it means start from the bottom of the
screen (as in "-10-10"). The code in ParseGeometry was
parsing off the initial + or - and then sending the
rest to strtoul on unix and strtol on Win and Mac.
That is not correct. Only a single + or - char should
be allowed. The old code let junk like "+1--2" through.
That contridicts the docs and I can't figure what
it would be good for.

hobbs added on 2002-06-24 14:34:56:
Logged In: YES 
user_id=72656

In what way is +1+-10 invalid?  It seems to do the expected 
thing (10 pixels about the top border), although I'm not sure 
what that is worth.

mdejong added on 2002-06-24 14:26:00:

File Added - 25684: wmgeom.patch2

Attachments: