Re: [OE-core] [PATCH] hosttools: force a --no-rosegment option if host ld supports it

[Paul Barker <paul@pbarker.dev>]

[-- Attachment #1: Type: text/plain, Size: 1204 bytes --]

On Fri, 2026-02-06 at 14:28 +0100, Alexander Kanavin via
lists.openembedded.org wrote:
> From: Alexander Kanavin <alex@linutronix.de>
> 
> Please see https://bugzilla.yoctoproject.org/show_bug.cgi?id=16087
> for the tricky, unpleasant details. This is addressing an issue with
> recent Fedora distributions where they can contaminate sstate with
> native executables that don't work when transferred onto alma/rocky 8/9.
> 
> [YOCTO #16087]

Hi Alex,

Thanks for working on this, we really appreciate the time is has taken
to investigate!

I think we need a little more detail in the commit message so that it is
understandable if the bugzilla data is ever not accessible. Perhaps just
something along the lines of:

  From Fedora 42 onwards, the --rosegment argument is passed to the
  linker by default when building native binaries. This triggers
  re-arrangement of sections by patchelf, resulting in binaries which
  are loaded incorrectly by older kernel versions. These "contaminated"
  native binaries end up in sstate and cause build failures when they
  are used on older distros such as AlmaLinux 8 & 9 or Rocky Linux 8 &
  9.

Best regards,

-- 
Paul Barker


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 252 bytes --]