Re: Build regressions/improvements in v4.12-rc2

[Arnd Bergmann]

On Mon, May 22, 2017 at 11:49 AM, Geert Uytterhoeven
<geert@xxxxxxxxxxxxxx> wrote:
> Hi Arnd, Helge,
>
> On Mon, May 22, 2017 at 11:35 AM, Geert Uytterhoeven
> <geert@xxxxxxxxxxxxxx> wrote:
>> Below is the list of build error/warning regressions/improvements in
>> v4.12-rc2[1] compared to v4.11[2].
>>
>> Summarized:
>>   - build errors: +2/-3
>>   - build warnings: +27717/-896
>
>> [1] http://kisskb.ellerman.id.au/kisskb/head/08332893e37af6ae779367e78e444f8f9571511d/ (267 out of 270 configs)
>
>> 27717 warning regressions:
>>
>> [Deleted 26393 lines about "warning: ... [-Wpointer-sign]" on parisc-allmodconfig]
>
> Do you know where these are coming from? They appeared in v4.12-rc1.
> The toolchain on kisskb hasn't changed (still gcc-4.6.3-nolibc).
>
>> [Deleted 1166 lines about "warning: -ffunction-sections disabled; it makes profiling impossible [enabled by default]" on parisc-allmodconfig]
>
> This has been introduced a while ago.  Would be good to get it fixed, too.
> I guess the "cflags-y += -ffunction-sections" still needs some CONFIG check,
> despite commit 366dd4ea9d5f0eb7 ("parisc: Fix ftrace function tracer")?

I found this mailing list discussion:

https://gcc.gnu.org/ml/gcc-help/2008-11/msg00141.html
https://gcc.gnu.org/ml/gcc-help/2008-11/msg00139.html

This patch was merged into gcc-4.8:

https://patchwork.ozlabs.org/patch/197134/

As the -ffunction-sections is a workaround for another problem, we could
use make it conditional on having a more recent compiler here.

        Arnd