Re: [Buildroot] [PATCH v5 1/3] package/python-numpy: bump to version 1.26.4

From: Thomas Petazzoni via buildroot <buildroot@buildroot.org>
To: James Hilliard <james.hilliard1@gmail.com>
Cc: buildroot@buildroot.org, Joseph Kogut <joseph.kogut@gmail.com>,
	Guillaume William Brs <guillaume.bressaix@gmail.com>,
	Asaf Kahlon <asafka7@gmail.com>,
	Jagan Teki <jagan@amarulasolutions.com>
Subject: Re: [Buildroot] [PATCH v5 1/3] package/python-numpy: bump to version 1.26.4
Date: Sat, 26 Oct 2024 23:31:01 +0200	[thread overview]
Message-ID: <20241026233101.323e096b@windsurf> (raw)
In-Reply-To: <CADvTj4qxs4=Er411V1-baHEVmr9FrTpOWzD3s-3oqM7ppqAb-Q@mail.gmail.com>

Hello James,

On Sat, 26 Oct 2024 15:15:03 -0600
James Hilliard <james.hilliard1@gmail.com> wrote:

> > Please document in the .mk file how you could those values. You
> > explained it in a reply to your v4, but I believe it should be kept in
> > the package for posterity. More comment on this below.  
> 
> Not sure the best way to describe how I got these, like should I put my
> test output somewhere?

A comment that explains how you got it, maybe even the test program, etc.

> > # Some good comment here about how those values were found
> > config BR2_PACKAGE_PYTHON_NUMPY_LONGDOUBLE_FORMAT
> >         string
> >         default "IEEE_DOUBLE_LE" if BR2_... || BR2_... || BR2_...
> >         default "IEEE_DOUBLE_BE" if BR2_... || BR2_...
> >
> > config BR2_PACKAGE_PYTHON_NUMPY_ARCH_SUPPORTS
> >         bool
> >         default y if BR2_PACKAGE_PYTHON_NUMPY_LONGDOUBLE_FORMAT != ""
> >
> > This way we cannot miss any case :-)  
> 
> Yeah, not sure what style is best, I kind of just went with this one
> because we essentially do the same sort of thing for meson:
> https://github.com/buildroot/buildroot/blob/2024.08.1/package/pkg-meson.mk#L32-L69

Yeah, I agree both options can work. What prompted me to suggest the
Config.in solution is precisely that you had forgotten some
architectures in the .mk file, showing that it made sense to tie
together the list of architectures with the long double format list.

> > Could you have a look into this? I wanted to rework myself, but since
> > the value is missing for some architectures (i386), I couldn't really
> > do it easily.  
> 
> Oh, I did actually generate that value here, must have forgotten to copy it:
> https://gist.github.com/jameshilliard/b057791b55bd32884301734df5b781db#file-numpy-qemu-test-txt-L16-L18

That's my point: your list of long double format in the .mk file was
not matching with the list of supported CPU architectures described in
the Config.in file.

Thomas
-- 
Thomas Petazzoni, co-owner and CEO, Bootlin
Embedded Linux and Kernel engineering and training
https://bootlin.com
_______________________________________________
buildroot mailing list
buildroot@buildroot.org
https://lists.buildroot.org/mailman/listinfo/buildroot