linux-kernel - Re: [PATCH v3 0/2] Don't create Python bytecode when building the kernel

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20250424152648.0979054f@sal.lan>
Date: Thu, 24 Apr 2025 15:26:48 +0800
From: Mauro Carvalho Chehab <mchehab+huawei@...nel.org>
To: Akira Yokosawa <akiyks@...il.com>
Cc: Andy Shevchenko <andriy.shevchenko@...el.com>, airlied@...il.com,
 corbet@....net, dmitry.baryshkov@....qualcomm.com,
 dri-devel@...ts.freedesktop.org, intel-gfx@...ts.freedesktop.org,
 jani.nikula@...ux.intel.com, joonas.lahtinen@...ux.intel.com,
 linux-doc@...r.kernel.org, linux-kbuild@...r.kernel.org,
 linux-kernel@...r.kernel.org, maarten.lankhorst@...ux.intel.com,
 masahiroy@...nel.org, mripard@...nel.org, nathan@...nel.org,
 nicolas.schier@...ux.dev, rodrigo.vivi@...el.com, simona@...ll.ch,
 tursulin@...ulin.net, tzimmermann@...e.de
Subject: Re: [PATCH v3 0/2] Don't create Python bytecode when building the
 kernel

Em Thu, 24 Apr 2025 11:07:05 +0900
Akira Yokosawa <akiyks@...il.com> escreveu:

> On Wed, 23 Apr 2025 19:31:36 +0300, Andy Shevchenko wrote:
> > On Wed, Apr 23, 2025 at 06:30:48PM +0900, Akira Yokosawa wrote:  
> >> On Tue, 22 Apr 2025 10:57:33 +0300, Andy Shevchenko wrote:  
> >>> On Mon, Apr 21, 2025 at 10:35:29AM -0600, Jonathan Corbet wrote:  
> >>>> Dmitry Baryshkov <dmitry.baryshkov@....qualcomm.com> writes:  
> > 
> > [...]
> >   
> >>>>> Would it be possible to properly support O= and create pyc / pycache
> >>>>> inside the object/output dir?  
> >>>>
> >>>> I have to confess, I've been wondering if we should be treating the .pyc
> >>>> files like we treat .o files or other intermediate products.  Rather
> >>>> than trying to avoid their creation entirely, perhaps we should just be
> >>>> sure they end up in the right place and are properly cleaned up...?
> >>>>
> >>>> To answer Dmitry's question, it seems that setting PYTHONPYCACHEPREFIX
> >>>> should do the trick?  
> >>>
> >>> It's not so easy. The Python is written in a way that it thinks it will never
> >>> runs object files separately from the source. Hence that variable sets only
> >>> the folder per script as _home_ for the cache. It's completely unusable. They
> >>> took it wrong. It still can be _painfully_ used, but it will make Makefiles
> >>> uglier.  
> >>
> >> But, PYTHONPYCACHEPREFIX can be set as an environment variable.
> >>
> >> For example, try:
> >>
> >>     export PYTHONPYCACHEPREFIX="$HOME/.cache/__pycache__"
> >>
> >> Wouldn't it be good enough for you?  
> > 
> > Of course not. We have _many_ scripts in python in kernel and having a cache
> > there for _all_ of them is simply WRONG. You never know what clashes can be
> > there with two complicated enough scripts which may have same module names,
> > etc.
> >   
> 
> Interesting...
> 
> I'm suspecting you replied without having tried the setting...
> 
> FYI, this is an excerpt from list of .pyc files under __pycache__ after
> building defconfig kernel and "make htmldocs"; and running
> 
> $ find . -name *.pyc" -print" under ~/.cache/__pycache__
> ---------------------------------------------------------------------
> ./home/.../git/linux/scripts/lib/kdoc/kdoc_files.cpython-312.pyc
> ./home/.../git/linux/scripts/lib/kdoc/kdoc_parser.cpython-312.pyc
> ./home/.../git/linux/scripts/lib/kdoc/kdoc_re.cpython-312.pyc
> ./home/.../git/linux/scripts/lib/kdoc/kdoc_output.cpython-312.pyc
> [...]
> ./usr/lib/python3.12/xml/__init__.cpython-312.pyc
> ./usr/lib/python3.12/xml/parsers/expat.cpython-312.pyc
> ./usr/lib/python3.12/xml/parsers/__init__.cpython-312.pyc
> ./usr/lib/python3.12/xml/etree/ElementPath.cpython-312.pyc
> ./usr/lib/python3.12/xml/etree/__init__.cpython-312.pyc
> ./usr/lib/python3.12/xml/etree/cElementTree.cpython-312.pyc
> ./usr/lib/python3.12/xml/etree/ElementTree.cpython-312.pyc
> ./usr/lib/python3.12/mimetypes.cpython-312.pyc
> [...]
> ./usr/lib/python3/dist-packages/sphinx/deprecation.cpython-312.pyc
> ./usr/lib/python3/dist-packages/sphinx/highlighting.cpython-312.pyc
> ./usr/lib/python3/dist-packages/sphinx/pycode/ast.cpython-312.pyc
> ./usr/lib/python3/dist-packages/sphinx/pycode/__init__.cpython-312.pyc
> ./usr/lib/python3/dist-packages/sphinx/pycode/parser.cpython-312.pyc
> ./usr/lib/python3/dist-packages/sphinx/config.cpython-312.pyc
> [...]
> ./home/.../sphinx-WIP/lib/python3.12/site-packages/sphinx/deprecation.cpython-312.pyc
> ./home/.../sphinx-WIP/lib/python3.12/site-packages/sphinx/highlighting.cpython-312.pyc
> ./home/.../sphinx-WIP/lib/python3.12/site-packages/sphinx/pycode/ast.cpython-312.pyc
> ./home/.../sphinx-WIP/lib/python3.12/site-packages/sphinx/pycode/__init__.cpython-312.pyc
> ./home/.../sphinx-WIP/lib/python3.12/site-packages/sphinx/pycode/parser.cpython-312.pyc
> ./home/.../sphinx-WIP/lib/python3.12/site-packages/sphinx/config.cpython-312.pyc
> [...]
> -------------------------------------------------------------------------
> 
> As you see, each of them are stored at a path corresponding to its original
> .py file.  The final part of the excerpt came from me running in-development
> Sphinx in a python venv with the same PYTHONPYCACHEPREFIX setting.
> 
> I don't see any possibility of clashes as you mentioned above,

I didn't test it, but it sounds good enough for me. Could you please send
us a patch on the top of:

	https://lore.kernel.org/linux-doc/cover.1745453655.git.mchehab+huawei@kernel.org/T/#t

for us to test it? The idea is to set PYTHONPYCACHEPREFIX only when O=
is used.

Regards,
Mauro