On 20.04.2010 22:33, ext Lauro Moura wrote:
The API change here would be making some Qt methods increment the refcount of objects what aren't owned but just required to be alive. In this case, QHttp().get(...,qfile_instance) requires the qfile_instance to be alive until the (asynchronous) fetch process finishes. So, the bug author was talking about making calls like this storing a reference to the object, almost like the following pseudo code:class QHttp(QObject): def get(self, ..., fileobject): self.qfile = fileobject It may not look like a API matter in terms of signature but changes the behavior of these calls. Anyway, this seems to be related to the "Pythonic API" but can be put in a proper PSEP in the future.
I'd regard the current behaviour just a bug: a Python programmer shouldn't need to be concerned about object lifetimes, and any unexpected behaviour definitely shouldn't result in a segfault. Am I correct that fixing this wouldn't actually change the behaviour of any existing programs?
Unless anyone objects, we could treat this just like a regular bug and prioritize and fix it as usual.
Cheers, ma. _______________________________________________ PySide mailing list [email protected] http://lists.openbossa.org/listinfo/pyside
