[Buildroot] Analysis of build failures
romain.naour at openwide.fr
Sun May 3 10:39:14 UTC 2015
Le 03/05/2015 10:50, Thomas Petazzoni a écrit :
> On Sat, 2 May 2015 08:30:16 +0200 (CEST), Thomas Petazzoni wrote:
>> powerpc | boost-1.57.0 | NOK | http://autobuild.buildroot.net/results/df8f87a34f59f81aa288c72673a80757f00ea6a2/
> Weird C++ issue.
This issue appears since boost-log has been added (BR2_PACKAGE_BOOST_LOG=y).
May be an issue specific to uClibc 0.9.33 ?
Needs to test with uClibc-snapshot and uClibc-ng.
>> arc | host-gdb-arc-2014.12-gdb | NOK | http://autobuild.buildroot.net/results/d44753dfe7e2aaa9fd803b4a69db65728f8301a8/
>> arc | host-gdb-arc-2014.12-gdb | NOK | http://autobuild.buildroot.net/results/ac3391023b13f405ef633688f714a10c22dc2db2/
> The dynamic linking issue.
There is two proposal to fixes this issue:
Also, it's not a issue specific to the arc toolchain:
>> i686 | numactl-2.0.10 | NOK | http://autobuild.buildroot.net/results/619339810617212a667fe72278ec727ee992ffbf/
> Don't know:
> checking for thread local storage (TLS) class... __thread
> ./configure: line 12681: syntax error near unexpected token `fi'
> ./configure: line 12681: `fi'
Weird because this package is autoreconfed... but wait.
If you look to the build-time.log, you'll see that host-autoconf-archive has
been installed before numactl.
host-autoconf-archive can suppressed several m4 macro bundled with a package.
Needs to test with and without host-autoconf-archive.
More information about the buildroot