[GRASS-dev] debugging nviz with TclTk 8.5 aqua

[Michael Barton]

On Jul 27, 2008, at 10:12 AM, Glynn Clements wrote:

>
> William Kyngesburye wrote:
>
>> Since the app startup doesn't have access to the user's startup
>> environment, they can't set it before running GRASS.
>
> We could do with having Init.sh explicitly source a file containing
> variable settings.
>
>>> Why not just include 2 versions of wish: 8.5 aqua and 8.5 x11.
>>> Switching GRASS_WISH will make the GRASS aqua or x11 compatible.
>>> These are small files and then binary users would not need to worry
>>> about having TclTk installed or which version they have installed.
>>
>> I noticed something that could cause trouble in freely switching
>> between aqua and X11 tcltk at runtime.  There is this 'osxaqua' env
>> var (I think it came from Lorenzo a long time ago).  This needs to be
>> set when running the GUI with tcltk aqua.
>>
>> In tcl scripts it's used to set keycontrol to the Mac command key and
>> execom to spawn.
>>
>> There should be a way to detect the graphics system in tcltk so the
>> user doesn't have to figure it out.  I poked around a bit and found  
>> that
>>
>> [tk windowingsystem]
>>
>> returns 'x11', 'aqua' and 'win32', independent of the platform.  So a
>> combination of $tcl_platform(platform) and [tk windowingsystem] could
>> be used to identify Mac X11 or Aqua.  I did find that [tk
>> windowingsystem] opens a small empty tk window, but it's a place to
>> start and maybe you know how to execute it properly.
>
> Tk itself creates the top-level window.
>
> I have changed gis.m to use "tk windowingsystem" instead of
> $env(osxaqua).
>
>> 'osxaqua' is also used once outside of tcl scripts, in init.sh.  It's
>> used to decide whether to pipe the gis.m startup command to "sh &" or
>> not.  I don't know if this really needs to be there, or if a  
>> different
>> test could be used.
>
> That code is complete gibberish, and potentially quite dangerous (if
> wish writes anything to stdout, the shell will attempt to execute it),
> so I have removed it.
>
> ISTR that it's a hack to deal with the case where Init.sh doesn't have
> a stdout (i.e. if it's run directly from the desktop rather than from
> a terminal).
>
> It may no longer be required with more recent versions of Tcl/Tk (or
> more recent versions of OSX). If wish still needs a stdout, it's
> probably better to just run the grass70 script with its stdout
> redirected to a file.

I also think that these are left over ways to deal with issues that  
are no longer issues. I'll try to test today and look at the tcltk  
code. Thanks for working on this.


Michael