[PATCH] riscv: Terminate the frame-pointer chain at entry points [BZ #34950]

Peter Bergner bergner@oss.tenstorrent.com
Thu Nov 19 22:41:27 GMT 2026


On 11/12/26 1:00 PM, Palmer Dabbelt wrote:>
> At least Fedora defaults to enabling frame pointers already.

The reason for that is a sorry tale and I wouldn't take Fedora enabling
-fno-omit-frame-pointer on their builds as we (the riscv community) thinking
that is a good idea and we probably should audit that change and whether we
(the riscv toolchain community) recommend its use or not in Fedora or any
distro.

The reason is was enabled, was that someone convinced people it was good for
x86 debugging.  I know there were some from the toolchain community who were
strongly opposed to the change, and that should have been that.  However, the
people asking for the change brought it to a vote when the toolchain people
were away on vacation and it passed.  Fedora likes to use common options
across all architectures (if possible), so it was globally enabled for
all architectures without the input from the other architectures.  When I
was still at IBM, we had to go back to Fedora and force them to disable it
on Power and S390, since it was a performance loss for those arches.

Peter





More information about the Libc-alpha mailing list