On Sep 25, 2006, at 11:08 PM, Russell E. Owen wrote:

In article <[EMAIL PROTECTED]>,
 Ronald Oussoren <[EMAIL PROTECTED]> wrote:

As I noted in a comment in that bugreport I'm not to keen on that. I
don't mind shipping the script ( or a GUI version of it) in the
application folder for users that want to switch to a newer Tk, or
even ship a minimal version of Tcl/Tk inside the Python.framework,
but running this script during installation means that (a) the script
must be 100% correct in all situations and (b) we'll end up with
Python installations that are slightly different which won't be fun
to investigate when someone reports a problem with tkinter.
Especially because users might not even realize they have a copy of
Tk in /Library/Frameworks.

(I've posted a slightly longer version of this to the bug report):

I see what you mean about different installations. I think the following
might work better:
- Always modify _tkinter.so to point to Tcl/Tk 8.4 in
   /Library/Frameworks.
This will fall back to the built in /System/Library/Frameworks if the
user has not installed an 8.4 of their own.

It avoids a few of the issues you bring up and is simpler and more
robust than what I originally suggested. Advantages:
- All installations would be the same.
- If the user installs a new Tcl/Tk after installing Python, it would be
used (unless it's 8.5, which would not be safe to try with Python).


It still does not address your concern than a user might accidentally
have a Tcl/Tk that they don't want to use. I'd personally be happier if users could easily upgrade their Tcl/Tk (since the installed one is so
bad), so I see this as more of an advantage than a disadvantage. Users
aren't going to typically install Tcl/Tk unless they want to use it, I
think. Still...I'm sure you've seen more requests for help than I have
over the years.

I'm not that much of an old hand support-wise ;-). I have seen multipe problem reports that turned out to be caused by software that a user had lying around and either didn't know about or completely forgot about. And that includes users that are technically savvy.


I'm not keen on including a Tcl/Tk for several reasons:
- Which version would you use? Even 8.4.11 has some important known
bugs, and 8.4.13 has different ones (at least one of which is very nasty
for my application, so I stick with 8.4.11 for now).
- If a user wanted to upgrade their Tcl/Tk, what would they do? The
answer is easy if we use the version of 8.4 found in /Library/ Framework
(if any).
- It can be tricky to add needed additions (my app uses the
"snack" sound library, for example). A standard Tcl/Tk makes this much
easier (and in fact  ActiveState Tcl/Tk already includes all additions
most folks would want).
- There is no universal Tcl/Tk yet (though one is planned). I personally
don't want to try to build one.

That's a shame. Including some build of Tk has one advantage: users wouldn't have to install Tk on Panther just to be able to use IDLE. If there is no universal build of Tcl/Tk available I'm definitely not going to spent time on trying to build one.

BTW. Versiontracker claims 8.4.13 is avaible as a universal binary although tcltkaqua doesn't mention the 8.4.13 release at all. The download for 8.4.13 seems to be from the tk-components project on SF.


So my personal suggestion is that we modify _tkinter.so using
Bob Ippolito's recipe unchanged (no fancy script that hunts for an
installed Tcl/Tk). It will be completely compatible with the built in
Tcl/Tk but gives any real users of Tcl/Tk (anyone who isn't just writing
"hello world") a trivial way to get a decent version.

What's the opionion of other Tk-using python users? I don't use tkinter on OSX because aquatk sucks in the L&F department and are hence happy with the current status quo (aka "IDLE works").


Ronald

Attachment: smime.p7s
Description: S/MIME cryptographic signature

_______________________________________________
Pythonmac-SIG maillist  -  Pythonmac-SIG@python.org
http://mail.python.org/mailman/listinfo/pythonmac-sig

Reply via email to