Re: [Buildroot] [PATCH] package/e2fsprogs: fix build with headers >= 5.10 and < 5.12

From: Thomas Petazzoni via buildroot <buildroot@buildroot.org>
To: Markus Mayer <mmayer@broadcom.com>
Cc: Julien Olivain <ju.o@free.fr>,
	Buildroot List <buildroot@buildroot.org>,
	Bernd Kuhls <bernd@kuhls.net>
Subject: Re: [Buildroot] [PATCH] package/e2fsprogs: fix build with headers >= 5.10 and < 5.12
Date: Fri, 28 Nov 2025 10:21:46 +0100	[thread overview]
Message-ID: <20251128102146.2afd777b@windsurf> (raw)
In-Reply-To: <CAGt4E5vMkqsHt2sUUqLWERZyRvQXyp1Z90E4wx8yPRP6CEiWgg@mail.gmail.com>

Hello Markus,

On Thu, 27 Nov 2025 11:53:36 -0800
Markus Mayer <mmayer@broadcom.com> wrote:

> I submitted a different patch for this to Ted T'so a week ago, on
> November 21, to be exact. Unfortunately, there doesn't seem to be a
> mailing list for e2fsprogs, so there is no way of quoting said email.
> Unfortunately, I haven't heard back from Ted as of yet.
> 
> I think the easiest way of showing my patch set here at this point in
> time is by providing a Github link.
> 
> https://github.com/tytso/e2fsprogs/compare/master...mmayer:e2fsprogs:kern-compat-master
> 
> This page should show the entirety of the diff, combining the two
> commits in that tree (manual changes and derived autoconf changes).
> The important changes are in the first commit (configure.ac and
> kern_compat.h)[1].
> 
> The big difference between your patch and my patch is that my patch
> doesn't disable functionality when the build environment is too old.
> Instead, it provides the missing definitions of the data structures,
> so the binaries can still be built with full functionality compiled
> in. So, while your patch strips functionality out of e2fsprogs based
> on the build environment, mine does not.

I don't really have a strong preference between the two solutions.

In your solution, what happens if the kernel is actually too old to
support the fsverity ioctl() ? Does e2fsprogs gracefully handles the
situation? That's a bit the problem with building code with "newer"
kernel headers than the running kernel: how is user-space going to be
have when the kernel features exposed in the kernel headers are
actually not implemented by the kernel.

Best regards,

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