On 29-Jan-2009, Ben Abbott wrote:
|
| On Jan 28, 2009, at 9:57 PM, John W. Eaton wrote:
|
| > Even without a version of gnuplot that supports the X11 window
| > position feature, I see a difference with this patch, and I wasn't
| > expecting that. Given the following sequence
| >
| > xlabel ("foo");
| > ## move window with mouse
| > ylabel ("bar");
| >
| > Without the patch, the window stays put after moving it with the
| > mouse.
| >
| > With the patch, and a version of gnuplot that does not support the
| > position feature, the window moves back to its original position after
| > the ylabel call.
|
| I think you got that backwards (or my changeset is improperly
| implemented).
|
| With or without my changeset , and without the gnuplot patch
| supporting the position feature, the window should stay put.
It doesn't, at least for me. When I apply the patch and use gnuplot
4.2.2, and execute the sequence above, the window moves back to the
original position after the call to ylabel. If I remove the patch,
the window stays put.
| If the event it is sensible to have gnuplot pass back each figure's
| window x11 id (is it called that?), it should be possible for to
| detect movements of the windows by the use and to reflect that in the
| properties.
|
| Is that a sensible thing to do?
|
| ... and does anyone know enough about x11 to explain how that should
| be implemented?
I think the simplest thing would be to query the window's geometry.
Maybe that could be limited to any time we need to get the position
property for the figure window. But first you need to be able to get
the window id, or whatever is needed to get the geometry of the window
back from gnuplot. Is that possible, or will it require changes to
gnuplot?