On Mon, Aug 31, 2026 at 08:03:21AM -0700, Jakub Kicinski wrote: > On Mon, 31 Aug 2026 17:21:28 +0800 Hangbin Liu wrote: > > Many test scripts under tools/testing/selftests/net/ each define their own > > log_test() function with near-identical logic for comparing a return > > code against an expected value and printing OK/FAIL. Add a shared > > log_test_expected() to lib.sh and replace each local definition with a > > thin wrapper that delegates to it. > > > > The function is named log_test_expected() rather than log_test() because > > lib.sh already exports log_test() with a different signature used by > > the forwarding tests. > > > > Two tests required special handling: > > > > - icmp_redirect.sh passed a fourth "xfail" argument to log_test(). The > > underlying issue has been fixed and the argument is always absent now, > > so it is simply dropped. > > > > - fib_nexthops.sh checked for ksft_skip (exit code 4) inside log_test(), > > but no test in the file produces that code. The check is removed, > > callers that need to report a skip should use log_test_skip() instead. > > Two callers compared against exit code 2 are left as-is since code 2 > > correctly counts as a failure in the previous log_test(). > > still breaks fib_tests.sh
Oh... This patch touches too much tests. I only fixed the [ "$VERBOSE" = "1" ] checking and return 1 issue in test_bridge_neigh_suppress.sh. I didn't notice one of fib_test.sh "set -e" and also breaks here. I think there is no test would expect return 1 for log_test. So I will fix it by return 0 in the lib function directly. Thanks Hangbin
