[Buildroot] Analysis of build failures

Thomas Petazzoni thomas.petazzoni at free-electrons.com
Sat Jan 23 08:01:48 UTC 2016


Frank,

On Fri, 22 Jan 2016 20:05:01 -0500, Frank Hunleth wrote:

> I received further advice from the Erlang mailing list. The summary is:
> 
>    1. Upgrade to Erlang 18 (BR is currently using 17.5)
>    2. Prefer the use of Erlang 18's native atomics implementation over
> libatomic_ops.
> 
> I verified that upgrading to Erlang 18 and not using libatomic_ops
> works for all qemu_*_defconfigs (with Erlang enabled), the aarch64
> autobuilder failure, and a couple other configs. It does not work with
> sparc_v8 as recently found with Erlang 17.5. I also tested whether
> Erlang 18's configure scripts would choose libatomic_ops if it were
> present, and they do not. This was tedious, so I am quite open to
> others repeating the tests (or telling me to do so) if this does not
> sound right. As a result of this testing, I disabled the use of
> libatomic_ops for Erlang as you will see in a subsequent patch set.
> 
> Unfortunately, the upgrade to Erlang 18 was nontrivial due to a couple
> major changes to the language and APIs. I had to bump the versions of
> ejabberd and nearly all Erlang packages to fix this. While Erlang is
> very important to me, I have never used the rebar infrastructure or
> ejabberd. I verified that the new version of ejabberd compiles across
> all qemu_*_defconfigs except the sparc_v8 one. I do not know if it
> runs correctly, though. Johan - I cc'd you since I saw that you are
> using ejabberd. Hopefully you can review too.
> 
> The patch set will follow shortly.

Thanks a lot for working on this. Upgrading to Erlang 18 definitely
looks like a good thing.

Thanks,

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com


More information about the buildroot mailing list