Arch Linux User Repository

This comment from @AvacadoCookie should be pinned, IMO.

If anyone is getting errors about Cython or setuptools, and they are using Conda, that comment has the answer.

If anyone is getting errors about Cython or setuptools, and they are using pyenv, there are 2 possible ways to fix it:

1. pyenv local system to set Python back to the system installed Python for this session.
2. pip install Cython setuptools to install the necessary packages to your preferred python installation.

For all users with ModuleNotFoundError: No module named 'Cython' or 'setuptools' reported, please try the following methods to address the problem:

1. Quit conda environment. Now after which python typed in console, the output should be /usr/bin/python.
2. Try again.
3. If the installation still failed, try again after pacman -S cython python-setuptools.

@Xavier Finally! Thanks!

@freso, that wrong path is dynamic, it depends on where you are building the package, so it is more complicated to just do a sed besides in future updates there could be more files (or not, we don´t know). Of course I prefer not to manipulate the file system outside the package manager, but we don't know the consequences of having wrong paths in the speed-up debugging.

With 4479465010ef2f6a58bc764b613e929daef79a6e (pkgrel=5) I get a hit on /usr/src/debug/pycharm-professional/pycharm-2023.3.3/plugins/python/helpers/pydev which doesn’t exist anywhere on my system.

This feels very similar to “WARNING: Package contains reference to $srcdir” (which I thought was what this was about at first, hence my grep’ing) and while I get that it may not be the neatest to have a "wrong" path in there, I’d argue that it’s way less neat to manipulate the file system outside of the package manager, especially if there is no confirmed issues arising from not doing so. If it turns out there is an issue, it should probably be filed as a bug with upstream. (I guess there’s always the option of sed -i 's|/usr/src/debug/pycharm-professional/pycharm-[0-9.]\+/|/opt/pycharm-professional/| or similar too…)

I’ll try and look into if there’s something that can be passed to build_ext to make it not include these paths, but even if not, I have a working PKGBUILD that doesn’t offend my sensibilities, so whether you choose to adopt any of my changes for that or not, it’ll only be a slight minor annoyance for me come update time if you don’t. :)

@Xavier Thanks for your thinking of that. Because that is really an inconvenience. For example, I currently switch to this non-eap version, because in eap, the UML diagrams are broken. But still I need the eap version to test some other stuff.

In the listing of package (yay -Ql pycharm-professional) the most files are already in /opt/pycharm-professional.

I only see the following files that may potentially conflict with other packages (not yet checked):

/usr/bin/ltedit - can be renamed as ltedit-pycharm-professional for example
/usr/bin/pycharm - can be renamed as pycharm-professional for example
/usr/share/applications/ltedit.desktop - may be renamed as ltedit-pycharm-professional.desktop for example (does not influence on the actual name in the app menu)
/usr/share/icons/hicolor/128x128/apps/pycharm.png - as pycharm-professional.png and edit in the desktop file (there is no extension there)
/usr/share/icons/hicolor/scalable/apps/pycharm.svg - as pycharm-professional.svg and edit in the desktop file (there is no extension there)

@Freso run grep -a "/pycharm-professional" /opt/pycharm-professional/plugins/python/helpers/pydev/_pydevd_bundle/pydevd_cython.cpython-311-x86_64-linux-gnu.so however I am not entirely sure that is really affecting the plugin if it shows a wrong path but I prefer avoid that.

1) I’m not reproducing. This is from my pkgrel=5 PKGBUILD:

freso@din ~ (main)> lddtree /opt/pycharm-professional/plugins/python/helpers/pydev/_pydevd_bundle/pydevd_cython.cpython-311-x86_64-linux-gnu.so
/opt/pycharm-professional/plugins/python/helpers/pydev/_pydevd_bundle/pydevd_cython.cpython-311-x86_64-linux-gnu.so (interpreter => None)
    libc.so.6 => /usr/lib/libc.so.6
        ld-linux-x86-64.so.2 => /usr/lib/ld-linux-x86-64.so.2
freso@din ~ (main)> grep -F 'makepkg' /opt/pycharm-professional/plugins/python/helpers/pydev/_pydevd_bundle/pydevd_cython.cpython-311-x86_64-linux-gnu.so
freso@din ~ (main) [1]> grep -F '/usr/lib' /opt/pycharm-professional/plugins/python/helpers/pydev/_pydevd_bundle/pydevd_cython.cpython-311-x86_64-linux-gnu.so
grep: /opt/pycharm-professional/plugins/python/helpers/pydev/_pydevd_bundle/pydevd_cython.cpython-311-x86_64-linux-gnu.so: binary file matches

either way, the current .install files both install files in the file system not managed by the package manager – and I’m not sure if the chroot’ing of post_remove() prevents this, but if not, this might also remove files that might intentionally exist in the file system outside of this package (e.g., if someone made a package to install a PyCharm plugin to /opt/pycharm-professional/plugins/).

2) It’s still an approved RFC, work is happening. It’ll be ready when it’s ready.

3) Part of the point with that commit is that there currently is no actual conflict between those packages, other than them being listed in this package’s $conflicts array. They can happily exist side-by-side as it is, right now, without any changes to any packages (other than cleaning up the $conflicts).

@Freso thanks so much for your contributions, 1) Of course, before I had the compilation of the speed-up debugging inside the PKGBUILD as you propose, the problem with that it generates wrong paths linked inside the .so files check this comment. 2) interesting about the x86_64_v3 but the discussion is very old, is it going to happen?. 3) I'm going to fix this kind of conflict when I resolve the problem of conflict with other pycharm package that we have been discussing with @ashark

(Relatedly, as far as I can discern, https://gitlab.archlinux.org/freso/pycharm-professional/-/commit/4479465010ef2f6a58bc764b613e929daef79a6e should be enough to allow Ashark to run this with their pycharm-eap build without you renaming /usr/bin/pycharm (and Ashark could also choose to just do this edit to their local PKGBUILD before makepkg’ing it).)

@ashark ok sounds reasonable, I'm going to do that but first I need to check if exists another conflict, I guess it is only the executable.