[Buildroot] [PATCH v2] package/uclibc: Enable compile in thumb mode configuration option

[Paul Enman]

Hi All,
Thanks for your feedback and insight. Buildroot does build a working
Cortex-M (thumb instructions only) uClibc toolchain by touching up the
uClibc(-ng) config file with

Enable COMPILE_IN_THUMB_MODE
and
Disable UCLIBC_HAS_CONTEXT_FUNCS

ARMv5 and thumb1 targets shouldn't be effected by this patch; I'll look
into guarding packages that depend on (obsolescent?) context control
functions and prepare an updated version for review.

The toolchain compiles a bootable kernel and both Linuxthreads and NPTL
build without error, however I am looking into why flthdr shows busybox
bFLT flags are set to 0x2 (Has-PIC-GOT).

On Wed, Dec 16, 2015 at 4:08 PM, Arnout Vandecappelle <arnout at mind.be>
wrote:

> On 15-12-15 23:03, Thomas Petazzoni wrote:
> > Paul,
> >
> > On Sun, 13 Dec 2015 19:28:40 -0600, Paul Enman wrote:
> [snip]
> >> +ifeq ($(BR2_ARM_INSTRUCTIONS_THUMB2),y)
> >> +define UCLIBC_ARM_THUMB_CONFIG
> >> +    $(call KCONFIG_ENABLE_OPT,COMPILE_IN_THUMB_MODE,$(@D)/.config)
> >
> > According to the help text of this uClibc option:
> >
> >           Say 'y' here to force building uClibc in thumb mode.
> >           Say 'n' to use your compiler's default mode.
> >
> > When you have BR2_ARM_INSTRUCTIONS_THUMB2, the compiler already
> > produces Thumb2 code by default, because the compiler is built with
> > --with-mode=thumb.
>
>  And in fact, the only thing that this option does is add -mthumb to
> CFLAGS,
> which is indeed quite useless for our internal compiler.
>
>  Or actually, it's not, because it also selects USE_BX and USE_LDREXSTREX.
> USE_BX is currently handled by BR2_UCLIBC_ARM_BX but this is a cleaner way
> I
> think. USE_LDREXSTREX is a better alternative to the SWP instruction so
> that's
> interesting as well.
>
>  Hm, looking at the uClibc code, there's something I don't quite
> understand.
> USE_LDREXSTREX is selected by COMPILE_IN_THUMB_MODE, but in fact
> ldrex/strex are
> only available on ARMv6+, so won't this break down on ARMv5?
>
>
>  Regards,
>  Arnout
>
>
> >
> > So could you be more specific on which cases this patch is fixing by
> > expanding the commit log?
>
> [snip]
>
>
>
> --
> Arnout Vandecappelle                          arnout at mind be
> Senior Embedded Software Architect            +32-16-286500
> Essensium/Mind                                http://www.mind.be
> G.Geenslaan 9, 3001 Leuven, Belgium           BE 872 984 063 RPR Leuven
> LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle
> GPG fingerprint:  7493 020B C7E3 8618 8DEC 222C 82EB F404 F9AC 0DDF
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.busybox.net/pipermail/buildroot/attachments/20151217/5ba375da/attachment.html>