Created attachment 3874 [details] rpmbuild I found that in some cases rpmbuild fails trying to build a SPEC file, see attached log. I was running oe-selftest, $ oe-selftest -r buildoptions.ArchiverTest.test_arch_work_dir_and_export_source And causes the test to fail, i suspect that is some kind of race because if i modified package_rpm.bbclass adding verbosity -vv to the rpmbuild command and the test passes.
I'm using a modified version of oe-selftest that allows to use more than 1 thread to execute unittests, this consume a lot of cpu/memory and may be the error is shown because that. Using 8 threads for execute unittests other packages starts to fail building rpm, ERROR: nativesdk-patchelf-0.9-r0 do_package_write_rpm: Function failed: BUILDSPEC ERROR: libtool-2.4.6-r0 do_package_write_rpm: Function failed: BUILDSPEC Search for that in the attached log.
Created attachment 3878 [details] oe-selftest run 8 threads
(In reply to comment #1) > I'm using a modified version of oe-selftest that allows to use more than 1 > thread to execute unittests, this consume a lot of cpu/memory and may be the > error is shown because that. > > Using 8 threads for execute unittests other packages starts to fail building > rpm, > > ERROR: nativesdk-patchelf-0.9-r0 do_package_write_rpm: Function failed: > BUILDSPEC > ERROR: libtool-2.4.6-r0 do_package_write_rpm: Function failed: BUILDSPEC > > Search for that in the attached log. My theory seems to be true using 4 pythreads for unittest execution the failures building the RPM's didn't happen.
Created attachment 3879 [details] oe-selftest run 4 threads
This is with a modified oe-selftest using threading and is likely an issue in that threading code. We now have parallelism in ie-selftest with the -j option and don't see this issue.