[Buildroot] Analysis of build failures

Peter Korsgaard jacmet at uclibc.org
Sat Feb 8 21:43:50 UTC 2014


>>>>> "Gustavo" == Gustavo Zacarias <gustavo at zacarias.com.ar> writes:

Hi,

 >> Missing setns() on CodeSourcery toolchain. Since util-linux is really
 >> needed for many packages, it would be good to find a fix that doesn't
 >> involve disabling util-linux completely.
 >> 
 >> checking for syscall setns... no
 >> configure: WARNING: Unable to detect syscall setns.
 >> configure: error: nsenter selected but setns syscall not found
 >> make: *** [/home/test/test/3/output/build/util-linux-2.23.2/.stamp_configured] Error 1
 >> 
 >> Gustavo, an idea?

 > Yes, blame commit 9f91d79601da42516752318beecfd080dc05aac9
 > Before it was an option util-linux decided if it was appropiate to build
 > it according to setns() being available, now we may be forcing it when
 > the syscall may simply be unavailable (like this toolchain).
 > We can do a ton of toolchain kludges or just revert, i'd prefer the
 > second option since it's not like a super-common command (actually most
 > distros don't have it yet because it's pretty new for util-linux).
 > Regards.

Hmm, looking closer at the issue I think I agree. Without the explicit
option nsenter automatically gets built if possible. Nsenter is very
small compared to the rest of util-linux (~20KB), and people can always
remove it in a post-build script if they really care.

-- 
Bye, Peter Korsgaard



More information about the buildroot mailing list