Re: [OE-core] [PATCH 1/2] elfutils: Fix backtrace ptests on musl

["Mathieu Dubois-Briand" <mathieu.dubois-briand@bootlin.com>]

On Mon Sep 28, 2026 at 7:00 AM CEST, Khem Raj via lists.openembedded.org wrote:
> musl is built with -fno-asynchronous-unwind-tables, so libc.so carries
> almost no .eh_frame and its CFI only lives in .debug_frame. Without it the
> unwinder cannot get out of abort () and run-backtrace-dwarf.sh fails with
> "No DWARF information found". Add musl-dbg to the ptest dependencies, like
> glibc-dbg for glibc.
>
> run-backtrace-native-core.sh then unwinds the thread correctly but loops on
> musl's __clone, which has no end-of-stack CFI. That is the same reason
> run-backtrace-native.sh is already an expected failure on musl, so mark the
> core variant as one too.
>
> Signed-off-by: Khem Raj <khem.raj@oss.qualcomm.com>
> ---

Hi Khem,

I suspect this is breaking elfutils ptest on qemux86-64:

AssertionError:
Failed ptests:
{'elfutils': 'START: ptest-runner\n'
             '2026-09-28T06:47\n'
             'make  \\\n'
	     ...
	  '===============================================================\n'
             '2 of 258 tests did not behave as expected (2 unexpected passes)\n'
             '(44 tests were not run)\n'

https://autobuilder.yoctoproject.org/valkyrie/#/builders/110/builds/1263

Ptest logs can be found here:
https://valkyrie.yocto.io/pub/non-release/20260928-40/testresults/qemux86-64-musl-ptest/core-image-ptest-elfutils/

It does look like it reproduces on every builds, but does not affect
other platforms.

Can you have a look at the issue?

Thanks,
Mathieu

-- 
Mathieu Dubois-Briand, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com