Re: Geometry issues ...

[Michael Meeks <michael ximian com>]

Hi Frank,

	Ok - it seems I have to reply to you :-)

On Thu, 2002-07-04 at 03:43, Frank Worsley wrote:
> The problem is that window size changes on a location basis.

	I agree with you for the 'Open In New Window' (OINW) mode, fine - I'm
just arguing about the 'browse' mode.

> Even if the
> the WM would restore the windows to a correct initial size, this still
> wouldn't solve the problem with Nautilus. In Nautilus I want my window
> to have one size when I open my 'Music' folder from the desktop and
> another size when I open my 'Work' folder from the desktop.

	Fair enough, then store the size, but not the position - that's a
fairly simple solution. Only store the size if we're not in OINW mode,
it doesn't worry me. Or - store both, but certainly ignore the position
in browse mode, or we get all windows piled up in an unusable mess.

> > 	Nevertheless - it's quite possible to serialise the last closed window
> > size for the non 'new-window' mode - but not in the meta-data, perhaps a
> > GConf key.
> 
> That seems like a bad hack to me when the real thing that needs fixing
> is the metadata system.

	Why ? why fix the metadata system ? it's not reasonable to complain
about the underlying system when it is being deliberately misused. It's
less simple and generic to fix the metadata system than to stop
hammering the API needlessly, - read the metadata code:

nautilus-directory-metafile.c (nautilus_directory_metadata_monitor),
nautilus-directory-metafile-monitor.c (corba_metafile_changed):
nautilus-directory.h etc. etc.

	Etc. you want to not emit changes when the "geometry" property is
changed ? do you want the same for any other properties ? where do you
propose to do the string compares ? is the 'geometry' property that
special that it can be constantly changed and ignored ?

> Storing the window size for a location really should be done in the
> metadata. I think fixing this to work right shouldn't be difficult at
> all, I will try and make a patch this weekend.

	I look forward to seeing it; it'd be great to fix the other
pre-existing bugs in the geometry serialisation code as well.

	Regards,

		Michael.

-- 
 mmeeks gnu org  <><, Pseudo Engineer, itinerant idiot