[Buildroot] Analysis of build results for 2017-11-25

Johan Oudinet johan.oudinet at gmail.com
Thu Nov 30 15:00:55 UTC 2017


Hello,

On Wed, Nov 29, 2017 at 2:34 PM, Thomas Petazzoni
<thomas.petazzoni at free-electrons.com> wrote:
>
> On Wed, 29 Nov 2017 14:28:55 +0100, Johan Oudinet wrote:
>
>> > I don't have much time at the moment but I'll take a look at this when I
>> > can (in a previous life I was an Erlang programmer, it's great to see it
>> > as part of buildroot!). It would be sad to disable HiPE so I'll see if
>> > we can just bump the version up to 20.
>> >
>> > Or Johan if you get to it first, I'll certainly test it :-)
>>
>> I haven't had the time to do it yet, however, it seems OTP 20 has
>> simply applied the patch I mentioned:
>> https://github.com/erlang/otp/pull/1394/commits/024e438101c010f17e3a374117c3b376d48483f3
>>
>> If I understand correctly this configure, there is a test in the
>> configure to automatically enable HiPE for a list of architectures.
>> ppc64 is present in this list but not ppc64le. So, I think HiPE would
>> be disabled on ppc64le architectures even if we bump to OTP 20.
>
> Well, if this commit made it to OTP 20, then there is no need to
> disable HiPE explicitly: my understanding is that this commit makes
> sure ppc64le is not mistaken as ppc64, in particular to avoid having
> HiPE being enabled. See https://github.com/erlang/otp/blob/master/erts/configure.in:
>
> if test "$cross_compiling" != "yes" && test X${enable_hipe} != Xno; then
>   if test -z "$M4"; then
>         enable_hipe=no
>         AC_MSG_NOTICE([HiPE disabled as no valid m4 is found in PATH])
>   else
>         case "$ARCH-$OPSYS" in
>             x86-linux|amd64-linux|x86-darwin*|amd64-darwin*|ppc-linux|ppc64-linux|ppc-darwin|arm-linux|amd64-freebsd|x86-freebsd|x86-sol2|amd64-sol2|ultrasparc-linux)
>                 enable_hipe=yes
>                 ;;
>         esac
>   fi
> fi
>
> ppc64le used to be handled as ppc64, so it was matching the case
> "$ARCH-$OPSYS", and enable_hipe was set to yes.
>
> With ARCH=ppc64le, this will no longer match, and enable_hipe will no
> longer be set to yes.
>

I had issues with some Erlang packages when testing Erlang OTP 20.1:
https://github.com/processone/xmpp/issues/18
However, everything looks good with OTP 20.0.
Since Christophe warned me about issues with OTP 20.1, I'm going to
send a patch to bump to version 20.0 only, which should be enough to
fix the ppc64le error reported here.

Best,
-- 
Johan



More information about the buildroot mailing list