homepage

(sent to me instead of to the bug manager; note that 
the tkFont naming approach is also used for widgets and 
callbacks /F)
The class tkFont has a problem name it chooses if no 
name is
given. The choosen method 'name = "font" + str(id
(self))' isn't safe
because the name may exists longer than the Font 
object and so another
Font object may be created with a different font but the 
same
id() and so the same name. id() says
 Return the identity of an object. This is guaranteed 
to be unique
 among simultaneously existing objects.
 ^^^^^^^^^^^^^^^^^^^^^^^
Note that this bug doesn't strike if you hold a reference 
to the Font
object and so the GC can't reclaim the space.
To reproduce run the attached program. It produces 
something like:
previously: ('Times', 15, 'bold')
now: ('Times', 18, 'bold')
iterations: 8
mapping: {'font136494612': ('Times', 
14, 'bold'), 'font136499772': ('Times', 
15, 'bold'), 'font136545468': ('Times', 
17, 'bold'), 'font136483156': ('Times', 
11, 'bold'), 'font136502700': ('Times', 
12, 'bold'), 'font136510460': ('Times', 
13, 'bold'), 'font136515228': ('Times', 
16, 'bold'), 'font136365348': ('Times', 10, 'bold')}
Traceback (most recent call last):
 File "tkfont_bug.py", line 20, in ?
 assert 0
AssertionError
Raimar

This bug has bitten me too, in a real program.

This still happens, but around 2000 iterations here on newer Python
versions (trunk, 3alpha, 2.5.x, 2.4.x). Using Python 1.6 I've got it at
2 iterations (min), but.. is this going to be fixed or given the amount
of iterations needed nowadays this is no longer a problem ? If it should
be fixed, are there any suggestions on how to proceed ?

Uhm, now I'm getting it at around 3 iterations with python-trunk. So,
can't we just use a simple generator for this ? Patch attached.
The same could be done for widget and callback naming.

Patch is really simple and looks ok to me, can we get this committed please.

New changeset a77e23135675 by Andrew Svetlov in branch 'default':
Issue #802310: Generate always unique tkinter font names if not directly passed
http://hg.python.org/cpython/rev/a77e23135675 

I've pushed fix inspired by Guilherme's suggestion.
Fix has been applied to 3.3 only because:
1. It changes font name generation schema
2. It's definitelly minor issue as exists starting from 2003.
Thanks.