Arch Linux User Repository

@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...

@gyscos Thanks - couple things - the only requirement is for make dependencies to be managed in the root package for all split packages. I guess runtime dependencies would be moved up to base if they were required for configuration/linking/inclusion purposes - but I can't think of a time I've had to do that. Make dependencies being pushed to 'depends' would only cause problems for users and won't solve anything in this situation.

At build time, all make + runtime dependencies must be installed, not just makedepends. It's pretty clear in the guidelines (https://man.archlinux.org/man/PKGBUILD.5#PACKAGE_SPLITTING) that the global depends + make depends need to include anything needed to build all packages. Essentially, if every split package overrides the global depends array, then that global depends array behaves exactly like the makedepends. It will not be present in the dependencies of any build package (since they all override that array), but is only required at build time.

I can see that python-numpy should be added to makedepends

qt5-base is also a dependency of one of the split packages (so it probably should be present at build time) but is not included in either make or runtime global depend.

@gyscos Thanks - couple things - the only requirement is for make dependencies to be managed in the root package for all split packages. I guess runtime dependencies would be moved up to base if they were required for configuration/linking/inclusion purposes - but I can't think of a time I've had to do that. Make dependencies being pushed to 'depends' would only cause problems for users and won't solve anything in this situation.

I can see that python-numpy should be added to makedepends - you can try install qt5-datavis3d, python-numpy, and makepkg -si from a clone of https://aur.archlinux.org/pyside2.git ... when reporting issues please use makepkg and not helpers.

I'll rebuild it in a short while to confirm things are working correctly with that change and make changes elsewhere if not

That message was initially copied from the output of makepkg -s, which cannot work anyway because qt5-datavis3d is in AUR.

The output from paru -B --noconfirm ., which is conceptually similar to makepkg -s, has a similar error message:

(... installs a bunch of dependencies ...)
fetching devel info...
==> Making package: pyside2 5.15.16+3.1-2 (Mon 30 Jun 2025 05:42:58 PM UTC)
==> Retrieving sources...
  -> Downloading pyside-setup-opensource-src-5.15.16.tar.xz...
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100   315  100   315    0     0    662      0 --:--:-- --:--:-- --:--:--   663
100 3502k  100 3502k    0     0  5343k      0 --:--:-- --:--:-- --:--:-- 32.7M
  -> Downloading pyside2_5.15.16-3.1.debian.tar.xz...
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100 28976  100 28976    0     0   716k      0 --:--:-- --:--:-- --:--:--  725k
  -> Found Modify-sendCommand-signatures.patch
  -> Found cmake-findpython.patch
==> Validating source files with sha256sums...
    pyside-setup-opensource-src-5.15.16.tar.xz ... Passed
    pyside2_5.15.16-3.1.debian.tar.xz ... Passed
    Modify-sendCommand-signatures.patch ... Passed
    cmake-findpython.patch ... Passed
==> Making package: pyside2 5.15.16+3.1-2 (Mon 30 Jun 2025 05:43:02 PM UTC)
==> Checking runtime dependencies...
==> Missing dependencies:
  -> qt5-multimedia
  -> qt5-tools
  -> qt5-sensors
  -> qt5-charts
  -> qt5-webengine
  -> qt5-datavis3d
  -> qt5-websockets
  -> qt5-speech
  -> qt5-3d
  -> qt5-svg
  -> qt5-script
  -> qt5-scxml
  -> qt5-x11extras
  -> qt5-quickcontrols2
  -> qt5-serialport
  -> qt5-remoteobjects
==> Checking buildtime dependencies...
==> ERROR: Could not resolve all dependencies.
error: failed to build 'pyside2-5.15.16+3.1-2 (shiboken2 pyside2 python-shiboken2 pyside2-tools)':
error: packages failed to build: pyside2-5.15.16+3.1-2 (shiboken2 pyside2 python-shiboken2 pyside2-tools)

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?

As mentioned in my previous message, if the global depends array is updated to properly reflect the state of each package depends array (or vice versa) as per the linked PKGBUILD guideline, then things build without issue.