This affects only py3k, but apparently tkinter has been in thissituation for more than 9 months now.I see these lines at _tkinter.c:const char *s = Tcl_GetStringResult(self->interp);const char *p = s;res = PyUnicode_FromStringAndSize(s, (int)(p-s));and I was wondering how could res not end up being an empty stringalways ? Then I did some quick tests here and the return is always anempty string, when wantobjects is set to False. Maybe the second lineshould be:const char *p = strchr(s, '\0');I've attached this correction. Not sure if it is the best way to solvethe problem.

I think in Python 3, the whole wantobjects=False case should go. It wasa compatibility measure to support applications that didn't expect Tclobjects; for Python 3, only a single case should be supported.

I will agree that wantobjects=False should go, it is also discouraged touse Tcl_GetStringResult because it may lose information. But something should be done at FromObj since it accesses Tcl_Obj fieldsdirectly, and it may end up not using the current object representationwhich may cause weird things like inissue1447222. Said that, I'm notsure how to fix it for now.

I've removed wantobjects and substituted some internalRep usage too.

Looks asissue18877 is related to this. I have proposed for it a little different patch than first Guilherme's patch.As for removing wantobjects, we perhaps can do this only in 3.4.

New changeset08c45e18e7b2 by Serhiy Storchaka in branch '3.3':Issue#3015: Fixed tkinter with wantobject=False.  Any Tcl command callhttp://hg.python.org/cpython/rev/08c45e18e7b2New changeset65dd0de6b4a5 by Serhiy Storchaka in branch 'default':Issue#3015: Fixed tkinter with wantobject=False.  Any Tcl command callhttp://hg.python.org/cpython/rev/65dd0de6b4a5

Here is a patch which change the default value for the wantobjects parameter of _tkinter.create() and adds deprecation warnings.

> Here is a patchYou again forgot to attach a patch :) .

Oh, thanks Arfrever.

I am in the process of reviewing this patch, but I don't know what "wantobjects" does. I can make a guess, I think it is a hack to make tcl objects work in Python. I am guessing this is less needed in Python 3.4, but still has some dependencies. If we can add in the documentation of what "wantobjects" is and why it is being depreciated somewhere, that would be great. I would do it myself, but I do need some direction.I ran the tests, and it seems to run fine. The patch didn't have any new tests. We should add one that checks if the depreciation is working.

@Lita.Cho: (I'd like to this on the web but the links are still broken after the website port) if you search for `wantobjects` in `Doc/whatsnew/2.3.rst`, there's a description of why the parameter was added.

Now I think we shouldn't deprecate and remove wantobjects=False. Tkinter was partially broken with wantobjects=False, on other hand, it was partially broken with wantobjects=True. Many bugs was fixed last months, all tests now work with wantobjects=False, and I hope Tkinter now correctly supports both mode.Besides using in old third-party code, this parameter is very helpful for testing. Tcl/Tk changes internal details from release to release, so when in one release Tk function returns list or integer, in other release it can return special object ('dict' or 'pixel'). Tkinter should be more stable against changing of return type, and wantobjects=False helps in detecting some such cases.I propose to close this issue as fixed.

Serhiy, does that mean this is fixed the way it is? Do I need to do anything else to close out this issue? It looks like wantobjects is set to False, currently.

New changesetc69e8ea3bf10 by Serhiy Storchaka in branch 'default':Issue#3015: _tkinter.create() now creates tkapp object with wantobject=1 byhttp://hg.python.org/cpython/rev/c69e8ea3bf10

Lita Cho, originally reported issue is fixed.I asked Martin (and Guilherme if he is here). Are you agree to close this issue? Seeissue21585 about extending testing.

That's perfect. I agree that this issue is closed! :)