[Buildroot] [autobuild.buildroot.net] Build results for 2016-03-01

Jeroen Roovers jer at airfi.aero
Wed Mar 2 15:29:25 UTC 2016


On 2 March 2016 at 16:21, Thomas Petazzoni
<thomas.petazzoni at free-electrons.com> wrote:

>> So this is horrible broken :/
>
> Ah, I understand better. I was confused by Jeroen's report, since his
> report was essentially "nginx concludes that SPARC is big endian",
> which was not wrong at all.



> Thanks for the explanation. Then indeed, passing NGX_HAVE_LITTLE_ENDIAN
> seems like the right thing to do.
>
> Does this means that nginx has never worked on little endian platforms ?

It worked fine for me so far. I am looking at some code in
src/http/ngx_http_parse.c where it sets up things according to

#if (NGX_HAVE_LITTLE_ENDIAN && NGX_HAVE_NONALIGNED)

which is not the case on my target system (nginx configuration has a
similarly interesting way to determine NGX_HAVE_NONALIGNED but that's
OK for ARM as they don't account for that at all).

so then I looked if that might be the cause of the problem (being
miscompilation with GCC 4.9.3 which I am now working around by setting
CFLAGS = -O1).

And then I found that on an amd64 host the endian test returned the
wrong result. So I asked this list if the test similarly returns the
wrong result on other systems and if the autobuild system should
perhaps flag that kind of error.

Regards,
     jer


More information about the buildroot mailing list