Re: Bug#775611: [sh]: FTBFS due to unknown compiler option '-m32'

[Ben Hutchings <ben@decadent.org.uk>]

[-- Attachment #1: Type: text/plain, Size: 1668 bytes --]

On Sun, 2015-01-18 at 12:15 +0100, Geert Uytterhoeven wrote:
> On Sun, Jan 18, 2015 at 10:58 AM, Ben Hutchings <ben@decadent.org.uk> wrote:
> > On Sun, 2015-01-18 at 00:35 +0100, John Paul Adrian Glaubitz wrote:
> >> Package: linux
> >> Version: 3.16.7-ckt4-1
> >> Severity: normal
> >> Tags: patch
> >>
> >> Hi!
> >>
> >> The kernel package currently fails to build from source on sh4 since
> >> the build scripts try to pass the '-m32' compiler option on gcc which
> >> is not available with gcc on sh4 (also according to the manpage).
> >>
> >> Selecting HAVE_C_RECORDMCOUNT in arch/sh/Kconfig (see attached patch
> >> by Ben Hutchings) fixes the issues. However, I also suggest removing
> >> the line "$cc .= " -m32";" for sh compiler options in scripts/
> >> recordmcount.pl.
> >
> > Steven and SH experts, what do you think of this?
> 
> It seems most compilers don't support -m32, unless they're biarch.
> On an Ubuntu 14.04 LTS box with lots of cross-compilers from korg/crosstool:
[...]

Right.  It's also not mentioned at all in the documentation of SH
options for gcc 4.9.  There appear to be multiple 32-bit modes and
instruction encodings to choose from, so it wouldn't be very useful.

The next question is whether recordmcount.c works properly for SH.

Adrian, can you check whether ftrace seems to work on the kernel built
with HAVE_C_RECORDMCOUNT selected?  (See Documentation/trace/ftrace.txt
and particularly the 'function' example.)

Ben.

-- 
Ben Hutchings
The generation of random numbers is too important to be left to chance.
                                                            - Robert Coveyou

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 811 bytes --]