Hi Stuart, On Thu, Jan 16, 2025 at 09:47:44AM +1100, Stuart Prescott wrote: > Hi Julian > > On 16/01/2025 08:34, Julian Gilbey wrote: > > Hi Stuart, > > > > On Mon, Jan 06, 2025 at 10:53:52AM +1100, Stuart Prescott wrote: > > > [...] > > > I don't think an explicit dependency on only Qt5 is the right fix for > > > python3-qtconsole. > > > > > > - python3-qtconsole is a widget not an application > > > - python3-qtconsole doesn't ship any executables of its own > > > - python3-qtconsole works with QtPy5 QtPy6 and PySide6 > > > [...] > > > > I'm not sure about this. While it's good in theory, it means other > > packages that blindly depend on python3-qtconsole will possibly end up > > without the required dependencies, and I don't want to have to > > regularly check the rdeps of python3-qtconsole for this sort of bug. > > I think it makes more sense to have a default pyqt5, and then rdeps > > can also install other Qt packages if they prefer. > > I disagree strongly with that. There is no desire to have packages install > PyQt5 *and* PySide6 as that is perverse and painful - and this bug already > exists in the archive now. The whole point of this work is to fix this bug > in our packaging. > > The package is a version agnostic "qtconsole" not a version-specific > "qt5console" (or similar). This is a change that has to happen at some stage > as PyQt5 will be removed from the archive and then that work on each rdep > needs to be done. It's less work to do that now (when it is already scoped) > than later (when there will be even more rdeps). > > It will not be the job of the maintainer of python3-qtconsole to check rdeps > - that's the job of the maintainers of the rdeps. The basic problem is is > reliance on transitive dependencies, and we've known for a long time that > it's best to avoid transitive dependencies and instead be explicit about the > dependencies. > [...] > Only 4 more packages. Yep, definitely worth it :)
OK, I hear that argument. I'll file bugs against those 4 packages, then. Best wishes, Julian
