[Buildroot] Analysis of build results for 2017-08-14

[Thomas Petazzoni <thomas.petazzoni@free-electrons.com>]

Hello,

On Tue, 15 Aug 2017 07:58:43 -0500, Matthew Weber wrote:

> > All of these are qemu-user failures, due to the host kernel headers
> > being too old. This error is normally ignored by autobuild-run, but
> > Julien's autobuilder is running a too old version of autobuild-run. So
> > I've temporarily added some filtering on the server side to reject such
> > bogus failures.
> 
> You're going to see some more unknown builds today (end log snippets
> below).  I've updated my autobuilder and trying to figure out the
> following new error.  At this point I disabled my result submissions
> and doing further testing offline.
> 
> http://autobuild.buildroot.net/results/e19/e196ceb333f554748081e78858eafb2739e0e317//
>  [7m>>>   Sanitizing RPATH in target tree [27m  
> /accts/mlweber1/instance-3/buildroot/support/scripts/fix-rpath target
> make: Leaving directory `/accts/mlweber1/instance-3/buildroot'
> make: Entering directory `/accts/mlweber1/instance-3/buildroot'
> package/coxpcall/coxpcall.mk:11: *** COXPCALL_SITE cannot be empty
> when COXPCALL_SOURCE is not.  Stop.
> make: Leaving directory `/accts/mlweber1/instance-3/buildroot'
> 
> 
> http://autobuild.buildroot.net/results/d3f/d3fa24b5589cc75f040a8f7aedb442ffda9b6814//
>  [7m>>>   Sanitizing RPATH in target tree [27m  
> /accts/mlweber1/instance-2/buildroot/support/scripts/fix-rpath target
> make: Leaving directory `/accts/mlweber1/instance-2/buildroot'
> make: Entering directory `/accts/mlweber1/instance-2/buildroot'
> package/lua-bit32/lua-bit32.mk:13: *** LUA_BIT32_SITE cannot be empty
> when LUA_BIT32_SOURCE is not.  Stop.
> make: Leaving directory `/accts/mlweber1/instance-2/buildroot'
> 
> Has anyone else seen errors like these?
> 
> The only modifications I have to the autobuilder script are to use a
> local cache of archives for my BR2_PRIMARY_SITE.  ( we refresh it
> daily out-of-band for internal firewall/proxy download failure
> reasons)   As part of setting BR2_PRIMARY_SITE  we set
> BR2_PRIMARY_SITE_ONLY=y.

Having BR2_PRIMARY_SITE_ONLY=y is going to trigger some download
errors, when your mirror will not contain a tarball that is needed.
Please don't enable BR2_PRIMARY_SITE_ONLY=y if you contribute to the
public autobuild.b.o.

However, I indeed spotted this "unknown" error on coxpcall from your
machine, and wasn't able to reproduce it. Can you investigate what's
going on ?

In the mean time, I would ask you to stop using this BR2_PRIMARY_SITE +
BR2_PRIMARY_SITE_ONLY setup. We don't want to be flooded with "unknown"
errors at this time, especially now that we understand those will
happens for pretty much all luarocks packages in your scenario.

Best regards,

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