When attempting to use sstate from a mirror, the siginfo file will be checked twice: first when deciding whether to run the setscene task and again while fetching in the setscene task. If the file is removed from the mirror between those two checks, the setscene task will run and fail. This is acceptable since since the real task will be run instead, but currently the fetcher will emit an error message which causes bitbake to return a nonzero exit code. The build completes successfully but appears to have failed.
This race has occurred several times in practice when using an http sstate server that periodically prunes old files. The simple fix provided here is to downgrade the internal fetcher error message to a warning and let the do_fetch task continue to generate its error level message. This is one of two paches. First openembedded-core must be patched to ignore the "Error:" prefix when scraping the log for this message. Then bitbake's log level can change. Tyler Hall (1): lib/oeqa/selftest: Don't match log level in output meta/lib/oeqa/selftest/bbtests.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) -- 2.0.3 -- _______________________________________________ Openembedded-core mailing list Openembedded-core@lists.openembedded.org http://lists.openembedded.org/mailman/listinfo/openembedded-core