Re: [PATCH v6 1/4] pylibfdt: allow build out of tree

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]



On Mon, Nov 08, 2021 at 07:07:56PM -0600, Rob Herring wrote:
> On Mon, Nov 8, 2021 at 6:09 PM Simon Glass <sjg@xxxxxxxxxxxx> wrote:
> >
> > Hi Rob,
> >
> > On Mon, 8 Nov 2021 at 09:59, Rob Herring <robh@xxxxxxxxxx> wrote:
> > >
> > > On Mon, Nov 8, 2021 at 9:58 AM Simon Glass <sjg@xxxxxxxxxxxx> wrote:
> > > >
> > > > Hi,
> > > >
> > > > On Sun, 7 Nov 2021 at 22:14, David Gibson <david@xxxxxxxxxxxxxxxxxxxxx> wrote:
> > > > >
> > > > > On Fri, Nov 05, 2021 at 04:01:38PM -0500, Rob Herring wrote:
> > > > > > On Wed, Nov 4, 2020 at 1:45 PM <marcandre.lureau@xxxxxxxxxx> wrote:
> > > > > > >
> > > > > > > From: Marc-André Lureau <marcandre.lureau@xxxxxxxxxx>
> > > > > > >
> > > > > > > With meson, we have to support out-of-tree build.
> > > > > >
> > > > > > I noticed that meson still builds pylibfdt/libfdt.py and
> > > > > > pylibfdt/libfdt_wrap.c in tree. And it makes the result unusable
> > > > > > without installing. For example the README instructions don't work:
> > > > > >
> > > > > >     $ cd pylibfdt
> > > > > >     $ python3 -c "import libfdt; help(libfdt)"
> > > > > >
> > > > > > I tried to fix the in tree part with swig option '-outdir', but that
> > > > > > didn't work.
> > > > >
> > > > > Urgh.  This has always fit poorly with the rest of the build system.
> > > > > In addition, I've discovered that with Fedora 35, I now get this scary warning:
> > > > >
> > > > >          PYMOD pylibfdt/_libfdt.so
> > > > > /usr/lib/python3.10/site-packages/setuptools/dist.py:498: UserWarning: The version specified ('1.6.1-g4eda2590-dirty') is an invalid version, this may not work as expected with newer versions of setuptools, pip, and PyPI. Please see PEP 440 for more details.
> > > > >   warnings.warn(
> > > > > make[1]: Leaving directory '/home/dwg/src/dtc'
> > > > >
> > > > > Merely replacing 'distutils' with 'setuptools' in the code does not
> > > > > fix this.
> > > >
> > > > I don't know much about this. Python seems to want to use .' 'instead
> > > > of '-' but I'm not sure why they decided this so recently.
> > >
> > > I'm working on patches for both of these issues.
> > >
> > > > > > Really what I'm after is what uses need to be supported? I'm looking
> > > > > > at getting pip and pypi packaging to work. I'm actually wondering if
> > > > > > pylibfdt needs to be so tightly coupled to the build system(s!) and
> > > > > > source tree. The dependencies are really just libfdt headers and
> > > > > > library which are easily installed either from source or a distro pkg.
> > > > > > The '..libfdt/' references are problematic with the packaging tools.
> > > > > > The easier route may be a standalone repo with a copy of libfdt.i and
> > > > > > a different setup.py.
> > > > >
> > > > > Unfortunately, I don't know enough about either Meson or Python
> > > > > packaging to fix this.  Simon?
> > > >
> > > > I'm not quite sure what the problem is with pip3, etc. I am not an
> > > > expert on any of these but can probably find some help if we have a
> > > > clear problem. Rob, do you have a patch that causes the problem?
> > >
> > > Python has certain thoughts on what the directory structure looks like
> > > and building an sdist can't include something outside of pylibfdt/
> > > (the headers specifically). Basically, this is what needs to work
> > > locally for an install from PyPI to work:
> > >
> > > ./setup.py sdist
> > > pip install -f ./dist/ libfdt
> > >
> > > We also need to statically link if we want to create a wheel (python
> > > binary) for PyPI. (setuptools has to be used for a wheel)
> > >
> > > > I am not keen on splitting the repos as pylibfdt has a source
> > > > dependency on libfdt.h
> > >
> > > Yes, but libfdt.h is part of the ABI and should only change in an ABI
> > > compatible way, right? If you really want the version of libfdt you
> > > built, you do have to install it AFAICT. I don't think the wrapper
> > > will load ../libfdt/libfdt.so.
> > >
> > > I think the easiest way to solve these problems is moving setup.py up
> > > to the project root and perhaps dropping the pylibfdt parts from
> > > make/meson. Then working with the python bits all works 'the Python
> > > way'.
> >
> > That seems OK to me, but can we have the Makefile invoke the Python
> > stuff? How does having a meson option (who uses it?) affect all this?
> 
> Everyone. The goal is meson will replace at least the bulk of the
> makefiles and they just become wrappers. The issue is getting setup.py
> to do out of tree builds that I mentioned at the start of this thread.

I'm ok with that, but could you please update the README to include
"meson native" build instructions.

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

Attachment: signature.asc
Description: PGP signature


[Index of Archives]     [Device Tree]     [Device Tree Spec]     [Linux Driver Backports]     [Video for Linux]     [Linux USB Devel]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [Yosemite Backpacking]

  Powered by Linux