Ubuntu 25.10: base-passwd fails to chown. (pseudo issue)

Ubuntu 25.10: base-passwd fails to chown. (pseudo issue)

[Gijs Peskens]

I'm just posting this here as a reference so others who run into this 
can hopefully find it, as it took me a while to figure out what was 
happening ;)

Ubuntu 25.10 obviously isn't supported, so I'm aware I'm on my own....

On Ubuntu 25.10 packages will fail to chown, this is due to Ubuntu 
having moved away from gnu coreutils to the rust based uutils coreutils. 
I'm presuming that Pseudo works by overriding libC provided functions, 
and thus intercepting the syscalls, and that the rust based coreutils do 
their syscalls directly (without a libC), which breaks pseudo and will 
result in failure to chown errors (and possibly other errors as well).

As a work-around, on 25.10, it's still possible to switch back to 
gnu-coreutils:
sudo apt-get remove --allow-remove-essential coreutils-from-uutils

(at your own risk)

Hope it helps others who run into this! :)

(and perhaps some thought must be given to this before 26.04 comes out, 
if the intention is to support that LTS).

Gijs Peskens

Re: [OE-core] Ubuntu 25.10: base-passwd fails to chown. (pseudo issue)

[Alexander Kanavin]

On Wed, 22 Oct 2025 at 11:56, Gijs Peskens via lists.openembedded.org
<gijs.peskens=munisense.com@lists.openembedded.org> wrote:
> I'm just posting this here as a reference so others who run into this
> can hopefully find it, as it took me a while to figure out what was
> happening ;)
>
> Ubuntu 25.10 obviously isn't supported, so I'm aware I'm on my own....
>
> On Ubuntu 25.10 packages will fail to chown, this is due to Ubuntu
> having moved away from gnu coreutils to the rust based uutils coreutils.
> I'm presuming that Pseudo works by overriding libC provided functions,
> and thus intercepting the syscalls, and that the rust based coreutils do
> their syscalls directly (without a libC), which breaks pseudo and will
> result in failure to chown errors (and possibly other errors as well).
>
> As a work-around, on 25.10, it's still possible to switch back to
> gnu-coreutils:
> sudo apt-get remove --allow-remove-essential coreutils-from-uutils
>
> (at your own risk)

Thanks for the report. I was dreading the day this happens, and I'm
sure others as well. Chances are, other distros will follow, and then
Yocto project has a problem.

Should we build C based native coreutils as early as possible, and
then use only that? That might buy some time, but isn't a real
solution, if GNU stops their development.

Alex

Re: [OE-core] Ubuntu 25.10: base-passwd fails to chown. (pseudo issue)

[Gijs Peskens]

On 10/22/25 13:11, Alexander Kanavin wrote:
> Thanks for the report. I was dreading the day this happens, and I'm
> sure others as well. Chances are, other distros will follow, and then
> Yocto project has a problem.
>
> Should we build C based native coreutils as early as possible, and
> then use only that? That might buy some time, but isn't a real
> solution, if GNU stops their development.
I'd guess that's the lowest energy approach, other routes would 
potentially involve writing the utils themselves with pseudo in mind and 
calling back into host utilities where needed. I don't believe GNU would 
stop development on coreutils anywhere in the near(ish) future, nor that 
the approach of building native coreutils would break since it should 
more or less only depend on the libC and kernel, things which aren't 
going away anytime soon.

Perhaps some magic right now inside bitbake that checks whether the 
host-utils are uutils and fails early would be nice (again: I'm aware 
I'm in unsupported land, energy shouldn't be wasted too much).

Gijs

Re: [OE-core] Ubuntu 25.10: base-passwd fails to chown. (pseudo issue)

[Ross Burton]

On 22 Oct 2025, at 17:05, Gijs Peskens via lists.openembedded.org <gijs.peskens=munisense.com@lists.openembedded.org> wrote:
> On 10/22/25 13:11, Alexander Kanavin wrote:
>> Thanks for the report. I was dreading the day this happens, and I'm
>> sure others as well. Chances are, other distros will follow, and then
>> Yocto project has a problem.
>> 
>> Should we build C based native coreutils as early as possible, and
>> then use only that? That might buy some time, but isn't a real
>> solution, if GNU stops their development.
> I'd guess that's the lowest energy approach, other routes would potentially involve writing the utils themselves with pseudo in mind and calling back into host utilities where needed. I don't believe GNU would stop development on coreutils anywhere in the near(ish) future, nor that the approach of building native coreutils would break since it should more or less only depend on the libC and kernel, things which aren't going away anytime soon.
> 
> Perhaps some magic right now inside bitbake that checks whether the host-utils are uutils and fails early would be nice (again: I'm aware I'm in unsupported land, energy shouldn't be wasted too much).

I just filed a bug to ensure that we detect this and tell the user to install the GNU tools:https://bugzilla.yoctoproject.org/show_bug.cgi?id=16028

But I’m curious how Ubuntu is successfully doing builds of packages with the rust coreutils, considering there are (afaik) non-zero packages that need to be built with fakeroot still, which does interception the same was a pseudo does. Maybe their builder just has coreutils installed?

Ross