Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
    From IRC (bridge bot)
    @FromIRC
    <crmaurei> I see
    From IRC (bridge bot)
    @FromIRC
    <ctismer> I thought this was already repaired, but obviously not. The names should be in the same order as in the dependency. It is hard to sort according to the dependency, but it would be easy to check that the subset is in the wrong order.
    <crmaurei> yeah, but since the workaround is so simple, I don't see it as a critical aspect, just a nice-to-have
    <ctismer> since the workaround is simple, the check should be added. Also, you get a late crash when using lower case names. Should also be easy.
    <crmaurei> yeah
    Alexis Jeandet
    @jeandet
    @cmaureir , I did a last test this morning, as long as I use packaged numpy I get no segfaults, then with or without debug I get the segfault with numpy from PyPI. Thank you for your time.
    Cristián Maureira-Fredes
    @cmaureir
    still a weird issue
    in any case, it could be that we are doing something with a refcount when building without --debug that is triggering something on numpy
    So in any case, do not discard the idea that could be a bug generated by pyside
    From IRC (bridge bot)
    @FromIRC
    <crmaurei> sifalt: ping
    <crmaurei> https://bugreports.qt.io/browse/PYSIDE-1066 is it possible to update the macOS sdk version we have in the provisioning that generate the wheels?
    From IRC (bridge bot)
    @FromIRC
    <sifalt> crmaurei: I think so, while I believe the same shit will hit the Qt apps if not updated.
    <sifalt> crmaurei: actually i think there is some actions ongoing alreaydy...
    <alcroito> hmm?
    <alcroito> i thought we use 10.13 for packaging, and that 10.14 is not ready due tests failing and whatnot
    From IRC (bridge bot)
    @FromIRC
    <sifalt> alcroito: we use 10.13 for packaging... but the SDK can be something different..right?
    <alcroito> sifalt sdk is then either 10.13 or 10.14
    <alcroito> sifalt i see MacOSX10.14.sdk in dev branch
    <sifalt> alcroito: you are right. And while checking 5.13.0 Qt bins, the otool claims sdk 10.14 also
    <alcroito> must be some other issue
    <sifalt> alcroito: makes me wonder where did this bugreport come from...
    <alcroito> sifalt the bug report is legit, i had the same myself, i just don't know why it happens
    <sifalt> alcroito: from pypi bins? which version?
    <alcroito> sifalt the one they reported in the bug report
    From IRC (bridge bot)
    @FromIRC
    <crmaurei> it's weird, since the wheel generation happens on the same machine, I though the sdk was explicitely passed as an argument
    <crmaurei> but I didn't find it on the build_instructions
    <crmaurei> :⁠P
    From IRC (bridge bot)
    @FromIRC
    <sifalt> crmaurei: I remember that we used to have that, but it was dropped for some reason...
    Alexis Jeandet
    @jeandet
    it looks like I can't add ownership tag on a function argument, like it only works on methods.
    I get (bindings) [8s] Running Source generator... (bindings) ASSERT: "scope" in file /home/jeandet/Documents/prog/pyside-setup/sources/shiboken2/generator/shiboken2/shibokengenerator.cpp, line 2393
    the xml typesystem snippet is
        <function signature="register_provider(PyDataProvider*)">
            <modify-function  signature="register_provider(PyDataProvider*)">
              <modify-argument index="1">
                  <define-ownership owner="c++" />
              </modify-argument>
            </modify-function>
        </function>
    Cristián Maureira-Fredes
    @cmaureir
    This tag has some limitations, it doesn’t support function modifications, besides you can’t add a function overload using add-function tag to an existent function. These limitation will be addressed in future versions of ApiExtractor.
    could it be possible for you to declare it as a object-type ?
    Alexis Jeandet
    @jeandet
    @cmaureir , yeah I can use a fake object for this, thank you.
    So in the future it will be possible to change ownership of function args?
    Alexis Jeandet
    @jeandet
    I've used pythonqt and Pybind11 in the past, until now my experience with PySide is not bad specially with Qt integration. I only miss the easy cpp side customisation and modern cpp/stl support from Pybind11. Also it's quite easy to embed an IPython console in your app with PySide which is really nice :).
    Cristián Maureira-Fredes
    @cmaureir
    @jeandet that would be the idea, yes, but since there are so many things to improve, we could lost track if it. If you encounter the same issue again, would you be so kind of openning a Suggestion in JIRA? that will help us prioritize the features.
    I'm glad the experience was good at the moment, and we are aware of other really nice solutions out there like pybind11, or even cppyy
    so that's why we need your feedback to know where to focus the future of shiboken :)
    Ogi Moore
    @j9ac9k
    @cmaureir just saw you elevating the priority on my bug report (PYSIDE-1066) thanks for taking a look at it! Not sure what I can do on my end, but if there is anything I can do please don't hesitate to ping me.
    From IRC (bridge bot)
    @FromIRC
    <crmaurei> yay! thanks for telling us xrogaan :⁠D
    From IRC (bridge bot)
    @FromIRC
    <ctismer> xrogaan_: Whow, what a nice article! I don't speak french, but Google Translate did a good job, when I was pasting the paragraphs in.
    Aens
    @Aens
    cool
    btw @ctismer, I saw that you patched the qt namespace issue. Will it release for 5.14 or another version?
    Christian Tismer
    @ctismer
    @Aens yeah, I had to change quite a lot. It will probably be in 5.13.1
    and merged in every 5.14 release as well.
    Aens
    @Aens
    cool
    thank you