On Wed, Dec 23, 2015 at 1:26 PM, Burton, Ross <[email protected]> wrote: > CCing Robert as he wrote the patch that introduced the ARCH assignment. > > On 23 December 2015 at 20:49, Andre McCurdy <[email protected]> wrote: >> >> Error is due to do_compile_ptest() forcing ARCH="${TARGET_ARCH}". The >> tests Makefile uses ARCH to setup include paths but there isn't a >> direct mapping between OE's arch values and the ones expected by >> strace. In this case ARCH should be "i386" in order to find the header >> file correctly, not "i586". >> >> Removing ARCH="${TARGET_ARCH}" from do_compile_ptest() and letting >> strace auto detect fixes the build failure. >> >> Forcing ARCH was introduced here: >> >> http://git.openembedded.org/openembedded-core/commit/?id=af76eb0405352d66dfd00d7f856f401d892855e1 >> >> But I'm not 100% clear what the original problem was though, so I'm >> not sure about just undoing that change without some more >> investigation. Any thoughts? > > I guess if it autodetects correctly now then the workaround isn't required. > Robert, what do you think?
Forcing ARCH (and OS) for strace 4.10 seems to be a purely cosmetic fix to make the include paths on the compiler command lines look prettier. None of the sources in tests seem to require the headers in the $(OS)/$(ARCH) directories, so it doesn't matter if ARCH is undefined or forced to something incorrect. For strace 4.11, forcing OS and ARCH is wrong and should be removed. The WIP change currently in mut is OK as a temporary solution, but maybe I'll send a new patch series for strace 4.11 in the next few days since I think some other things need fixing too (the configure.ac patch and acl / libaio dependencies are no longer correct and the git-version-gen script needs updating or dropping). > Ross -- _______________________________________________ Openembedded-core mailing list [email protected] http://lists.openembedded.org/mailman/listinfo/openembedded-core
