Groups > linux.debian.maint.python > #15066 > unrolled thread

Python module installation with cmake

Started by	Gregor Riepl <onitake@gmail.com>
First post	2023-07-28 09:30 +0200
Last post	2023-07-28 09:50 +0200
Articles	2 — 2 participants

Back to article view | Back to linux.debian.maint.python

  Python module installation with cmake Gregor Riepl <onitake@gmail.com> - 2023-07-28 09:30 +0200
    Re: Python module installation with cmake Timo Röhling <roehling@debian.org> - 2023-07-28 09:50 +0200

#15066 — Python module installation with cmake

From	Gregor Riepl <onitake@gmail.com>
Date	2023-07-28 09:30 +0200
Subject	Python module installation with cmake
Message-ID	<GWq5c-3jWm-7@gated-at.bofh.it>

Hi,

I'm looking for help with a regression in cmake 3.27, that is currently 
producing incorrect module installation paths for the Python package 
Uranium: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1042157

Uranium's cmake script uses the cmake variable Python_SITELIB to 
determine the installation path for Python modules, and that used to 
work correctly up until cmake 3.26.
In 3.27, the value of this variable has changed, and it now points to 
/usr/local, which seems wrong to me. Furthermore, it includes the full 
Python version, which also doesn't look right:

In 3.26.4-4, Python_SITELIB is /usr/lib/python3/dist-packages, and the 
documentation[1] says:
 > Information returned by 
distutils.sysconfig.get_python_lib(plat_specific=False,standard_lib=False) 
or else sysconfig.get_path('purelib').

In 3.27.1-1, it's /usr/local/lib/python3.11/dist-packages, coming from[2]:
 > Information returned by sysconfig.get_path('purelib').

There don't seem to be any other useful cmake variables that could be 
used. The relevant cmake script parts look like this:

####
if(NOT DEFINED Python_SITELIB_LOCAL)
     set(Python_SITELIB_LOCAL
             "${Python_SITELIB}"
             CACHE PATH "Local alternative site-package location to 
install Uranium" FORCE)
endif()
install(DIRECTORY UM DESTINATION "${Python_SITELIB_LOCAL}")
####

I do think this is a regression in cmake, but perhaps something was 
wrong with the script in the first place and only showed up now?

Can someone help with sorting this issue out?

Thanks,
Gregor

[1] https://cmake.org/cmake/help/v3.26/module/FindPython.html
[2] https://cmake.org/cmake/help/v3.27/module/FindPython.html

[toc] | [next] | [standalone]

#15067

From	Timo Röhling <roehling@debian.org>
Date	2023-07-28 09:50 +0200
Message-ID	<GWqoy-3k3b-9@gated-at.bofh.it>
In reply to	#15066

[Multipart message — attachments visible in raw view] — view raw

Hi Gregor,

* Gregor Riepl <onitake@gmail.com> [2023-07-28 09:21]:
>I'm looking for help with a regression in cmake 3.27, that is 
>currently producing incorrect module installation paths for the Python 
>package Uranium: 
>https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1042157

I posted an explanation to debian-devel yesterday:
https://lists.debian.org/debian-devel/2023/07/msg00307.html


Cheers
Timo

-- 
⢀⣴⠾⠻⢶⣦⠀   ╭────────────────────────────────────────────────────╮
⣾⠁⢠⠒⠀⣿⡁   │ Timo Röhling                                       │
⢿⡄⠘⠷⠚⠋⠀   │ 9B03 EBB9 8300 DF97 C2B1  23BF CC8C 6BDD 1403 F4CA │
⠈⠳⣄⠀⠀⠀⠀   ╰────────────────────────────────────────────────────╯

[toc] | [prev] | [standalone]

csiph-web

Python module installation with cmake

Contents

#15066 — Python module installation with cmake

#15067