Arch Linux User Repository

Thanks!

By "minimally required" I didn't mean by the actual program, but by the individual packages.

I suspect the bug in paru is that it installs the sum of all the child packages depends arrays, and then calls makepkg. But since no child package requires qt5-datavis3d or qt5-multimedia (and maybe others), these end up missing.

I agree that it's a deviation from the specs for paru, and it should just look at the global depends array to know what to install. But it is a bit weird to specify something in the global depends that is not actually required by any child package. If things are optionally required by child packages, it could be omitted from the top-level depends?

Anyway this would just be a way to avoid this bug in paru. I understand this package is not technically at fault (save for possibly failing to include python-numpy in the global makedepends, but this doesn't seem to actually cause any problem).

yes, there are more dependencies listed than are minimally required, but that is normal fair for Arch packages - which generally enable as many useful features as the maintainer sees fit.

The issue is that for whatever reason, your paru command did not download the qt5-datavis3d package from AUR, and expected that pacman would handle it. I think this may be related to that package being dropped from [extra] into AUR some weeks ago - but troubleshooting helpers is outside of the scope of AUR maintenance.

The 'missing packages' in your previous logs are simply the packages that did not auto install due to a failed dependency in the chain (qt5-datavis3d)

You're right, with qt5-datavis3d installed, makepkg -s works fine. It does look like an issue in the build system I'm using: it doesn't like that the global depends array has more constraints than the sum of all the individual depends (in this case, qt5-datavis3d and a few others are never required by any package, but are required by the top-level depends).

Is that really spec-compliant? The doc says depends should be at least everything needed to build the package. I guess it doesn't say it has to be exactly what the packages require.

@gyscos: all good, some packages are more enjoyable to maintain than others - I cringe every time I get a pyside2 notification as it usually means more patching - so I probably didn't come off at my best either.

I've built in a clean chroot with only qt5-datavis3d installed - can't reproduce any dependency/build issues. I haven't tried with paru but I can't see why it would have an issue with the current build. Can you confirm that it builds ok on your side with datavis already installed?

I'm sorry if I sounded aggressive. I was merely investigating this build failure, and before filing a ticket with the build system, wanted to make sure this package was conformant. Thank you for your work maintaining it!

$ namcap *zst | grep -i datavis3d
pyside2 W: Dependency qt5-datavis3d detected but optional (libraries ['usr/lib/libQt5DataVisualization.so.5'] needed in files ['usr/lib/python3.13/site-packages/PySide2/QtDataVisualization.cpython-313-x86_64-linux-gnu.so'])
$ namcap *zst | grep -i qt5-multimedia
pyside2 W: Dependency qt5-multimedia detected and implicitly satisfied but optional (libraries ['usr/lib/libQt5Multimedia.so.5', 'usr/lib/libQt5MultimediaWidgets.so.5'] needed in files ['usr/lib/python3.13/site-packages/PySide2/QtMultimedia.cpython-313-x86_64-linux-gnu.so', 'usr/lib/python3.13/site-packages/PySide2/QtMultimediaWidgets.cpython-313-x86_64-linux-gnu.so'])

I'm not sure why you have an issue with the dependency declarations. Can they be improved? Definitely. Are they causing issues? Not from what I can see. This is a legacy package that has changed hands more than once. Unless there is a specific issue that needs addressing, please avoid the excess commenting.

For example, which package needs qt5-multimedia? It's not a make-depend, yet none of the resulting package depend on it.

But which package depends on qt5-datavis3d? I see pyside2 has it as opt-depend, but is it really required for building the package if no package requires it, and it's not a make-depend?

Or more generally, how do you use the global depends and makedepends differently in this case where all split packages override the depends array?

@gyscos: I seem to have missed a question in your many statements. When you're able to report an issue using makepkg -si and not a helper, I'll be happy to help you get through it.

Indeed, I was wrong about qt5-base, but what about the other questions?

@gyscos: Not to worry, I understand how split packages work. qt5-base is an implicit and explicitly requiring it is redundant and unnecessary. In order to cut down on the noise I'll confirm the package is okay later today...