A couple of methods not wrapped properly?

[spnw]

Hi, unless I'm missing something I seem to have stumbled across a bug in the binding generation.

It pertains to two functions in particular:

• Based on its signature, q_stackedlayout_widget appears to wrap QStackedLayout::Widget, but the underlying QStackedLayout_Widget wraps QWidget::widget.
• Similarly, q_gestureevent_set_accepted appears to wrap QGestureEvent::setAccepted while the backend wraps QEvent::setAccepted.

The upshot is that I get a link error when building libqt6c, as the C++ wrappers used by these C functions do not exist. The library builds fine if I comment them out. I should mention I'm using CMake instead of Zig, but this appears to be an implementation issue as best as I can tell.

Thanks a lot for this excellent library by the way!

[rcalixte]

Hi! Thanks for the interest!

• q_stackedlayout_widget implements the function signature for QStackedLayout::Widget, but the underlying QStackedLayout_Widget calls QWidget::widget.

• Similarly, q_gestureevent_set_accepted appears to wrap QGestureEvent::setAccepted while the backend is calling QEvent::setAccepted.

I checked the history for both and they appear to have been functioning the same since inception (minus a few tangential bugs, sorry). The current version appears to be constructed as expected for both the QStackedLayout and QGestureEvent (both method variants). I think maybe what you are encountering is the method inheritance that has been simulated to support Qt's heavy usage of it? I can't be certain for sure but typically, including a class requires linking to any parents classes as well. QStackedLayout's headers include the QWidget binding header so that suggests it should be linked too. If you have an error output or stack trace, I would love to see it.

The upshot is that I get a link error when building libqt6c, which is fixed by commenting out the two C functions. I should mention I'm using CMake instead of Zig, but this appears to be an implementation issue as best as I can tell.

I've suspected some folks from the C crowd would make an effort for CMake. I had some challenges so I hope it's working out well for you! I'd be curious to hear about it even if I don't think the library's toolchain will change any time soon.

Thanks a lot for this excellent library by the way!

Thanks for giving it a try!

[spnw]

I can't be certain for sure but typically, including a class requires linking to any parents classes as well. QStackedLayout's headers include the QWidget binding header so that suggests it should be linked too. If you have an error output or stack trace, I would love to see it.

Here's the linker error. The missing functions appears to be the libqt6c wrappers rather than the actual Qt methods:

  "_QGestureEvent_SetAccepted", referenced from:
      _q_gestureevent_set_accepted in libqgesture.c.o
  "_QStackedLayout_Widget", referenced from:
      _q_stackedlayout_widget in libqstackedlayout.c.o
ld: symbol(s) not found for architecture x86_64

The current version appears to be constructed as expected for both the QStackedLayout and QGestureEvent (both method variants).

QGestureEvent_SetAccepted is defined as a function of two arguments (QGestureEvent* self, bool accepted). At the call site in q_gestureevent_set_accepted, it is given three arguments (GestureEvent*, Gesture*, bool). This is indeed how it's declared, but it does not correspond to the definition.

I've suspected some folks from the C crowd would make an effort for CMake. I had some challenges so I hope it's working out well for you! I'd be curious to hear about it even if I don't think the library's toolchain will change any time soon.

I'm a CMake noob but it was basically as simple as dropping the libqt6c files straight into my existing Qt project!

[rcalixte]

"_QGestureEvent_SetAccepted", referenced from:
_q_gestureevent_set_accepted in libqgesture.c.o
"_QStackedLayout_Widget", referenced from:
_q_stackedlayout_widget in libqstackedlayout.c.o
ld: symbol(s) not found for architecture x86_64

It seems like the library build isn't configured correctly. The wrappers and the bindings need to be compiled into the static library together. It looks like the symbols from the bindings are missing.

QGestureEvent_SetAccepted is defined as a function of two arguments (QGestureEvent* self, bool accepted). At the call site in q_gestureevent_set_accepted, it is given three arguments (GestureEvent*, Gesture*, bool). This is indeed how it's declared, but it does not correspond to the definition.

I see... I'll take a look at what's going on here. In theory, it shouldn't even compile with the mismatch but I'll dig deeper on the actual details and get back to you.

[spnw]

The wrappers and the bindings need to be compiled into the static library together.

They are. These two functions are the only ones that don't link, the other thousands work fine. :)

Thank you for your quick response on this.

[rcalixte]

Feel free to give it another go. It should be fixed in #40. I'm seeing a few other things I may want to change or revisit but shouldn't be buggy at the moment.

[spnw]

I can confirm it's fixed. I expect to be using this library pretty heavily in the near future, so I'll let you know if I stumble across anything else.

[rcalixte]

I can confirm it's fixed. I expect to be using this library pretty heavily in the near future, so I'll let you know if I stumble across anything else.

That's exciting! I can't wait to see what you come up with!

[rcalixte]

Fun fact: the total number of methods that were impacted like this (outside of QScintilla which is third-party), is exactly 4:

• QGestureEvent::setAccepted
• QStackedLayout::widget
• QTcpSocket::bind
• QUdpSocket::bind

[spnw]

A ha! I figured I may have missed some as I'm only using the core modules right now. Still a very small number, so it seems like it was a pretty weird edge case.