Re: [Pythonmac-SIG] Namespace conflict with WebWare and PyObjC / Import safety (Was: how do I use twisted cfreactor?)
On Jan 14, 2005, at 17:57, Bob Ippolito wrote: On Jan 14, 2005, at 17:14, Bob Ippolito wrote: On Jan 14, 2005, at 16:56, Kevin Dangoor wrote: Bob Ippolito wrote: (Kevin sent me the test off-list, and I took a look at it). I'm not sure exactly why your example crashes (somehow a retain message gets sent to a dead or non-object), however, the problem is that you are using an import statement from inside the implementation of the action. Don't do that. Do your imports in module level code. Wow. That was quick! I didn't realize that there was a gotcha with the imports. That was just a premature optimization, so I can easily avoid that :) Thanks for your help... that's certainly not the kind of thing I would have just guessed... I don't think there is typically a gotcha with imports, I've certainly never seen this happen before, and I have done imports from applicationDidFinishLaunching: (pygame, in particular) before. I have no idea if I should be blaming Cheetah, PyObjC or Python 2.3.0 (haven't tested with 2.4 or CVS), but I will try and remember to dig in later. I have traced the problem. It is a two-parter: (1) There is a module namespace conflict: WebWare has a package named WebKit PyObjC has a package named WebKit (2) Cheetah.Servlet checks to see if WebWare's WebKit is available, and ends up importing PyObjC's WebKit. (3) For whatever reason, it is not safe to import the WebKit wrapper from inside of an action (unless it's already imported, of course). Now this issue I will have to look further into. --- So apparently it's more or less undefined behavior if you have both WebWare and PyObjC installed! Fun :) I've filed a bug against WebWare http://sourceforge.net/tracker/index.php? func=detailaid=1102868group_id=4866atid=104866, hopefully they'll change the name of WebKit, I don't think we will do it in PyObjC -- because then we'd have to change the names of every wrapped framework for consistency. The actual crash is due to the fact that WebKit/__init__.py recycles the PyObjC global autorelease pool. God knows why. I've reverted this behavior in svn trunk r1337. -bob ___ Pythonmac-SIG maillist - Pythonmac-SIG@python.org http://mail.python.org/mailman/listinfo/pythonmac-sig
Re: [Pythonmac-SIG] Namespace conflict with WebWare and PyObjC / Import safety (Was: how do I use twisted cfreactor?)
On 14-jan-05, at 23:57, Bob Ippolito wrote: I have traced the problem. It is a two-parter: (1) There is a module namespace conflict: WebWare has a package named WebKit PyObjC has a package named WebKit As far as I know this is the first time this potential problem with python's naming convention is seen in the wild. Maybe it's worth it to report it to python-dev? -- Jack Jansen, [EMAIL PROTECTED], http://www.cwi.nl/~jack If I can't dance I don't want to be part of your revolution -- Emma Goldman ___ Pythonmac-SIG maillist - Pythonmac-SIG@python.org http://mail.python.org/mailman/listinfo/pythonmac-sig
Re: [Pyobjc-dev] Re: [Pythonmac-SIG] Namespace conflict with WebWare and PyObjC / Import safety (Was: how do I use twisted cfreactor?)
On Jan 15, 2005, at 18:26, Jack Jansen wrote: On 14-jan-05, at 23:57, Bob Ippolito wrote: I have traced the problem. It is a two-parter: (1) There is a module namespace conflict: WebWare has a package named WebKit PyObjC has a package named WebKit As far as I know this is the first time this potential problem with python's naming convention is seen in the wild. Maybe it's worth it to report it to python-dev? Well I filed a bug with Webware, let's wait a couple days and see if they say anything. -bob ___ Pythonmac-SIG maillist - Pythonmac-SIG@python.org http://mail.python.org/mailman/listinfo/pythonmac-sig
[Pythonmac-SIG] Namespace conflict with WebWare and PyObjC / Import safety (Was: how do I use twisted cfreactor?)
On Jan 14, 2005, at 17:14, Bob Ippolito wrote: On Jan 14, 2005, at 16:56, Kevin Dangoor wrote: Bob Ippolito wrote: (Kevin sent me the test off-list, and I took a look at it). I'm not sure exactly why your example crashes (somehow a retain message gets sent to a dead or non-object), however, the problem is that you are using an import statement from inside the implementation of the action. Don't do that. Do your imports in module level code. Wow. That was quick! I didn't realize that there was a gotcha with the imports. That was just a premature optimization, so I can easily avoid that :) Thanks for your help... that's certainly not the kind of thing I would have just guessed... I don't think there is typically a gotcha with imports, I've certainly never seen this happen before, and I have done imports from applicationDidFinishLaunching: (pygame, in particular) before. I have no idea if I should be blaming Cheetah, PyObjC or Python 2.3.0 (haven't tested with 2.4 or CVS), but I will try and remember to dig in later. I have traced the problem. It is a two-parter: (1) There is a module namespace conflict: WebWare has a package named WebKit PyObjC has a package named WebKit (2) Cheetah.Servlet checks to see if WebWare's WebKit is available, and ends up importing PyObjC's WebKit. (3) For whatever reason, it is not safe to import the WebKit wrapper from inside of an action (unless it's already imported, of course). Now this issue I will have to look further into. --- So apparently it's more or less undefined behavior if you have both WebWare and PyObjC installed! Fun :) -bob ___ Pythonmac-SIG maillist - Pythonmac-SIG@python.org http://mail.python.org/mailman/listinfo/pythonmac-sig