With 2.6.24.3-12.fc8 on x86 a make check almost always fails (x86_64 seems fine though). Downgrading to 2.6.23.15-137.fc8 makes make check (at least in frysk-core) completely clean again. The tests fail (and the TestRunner gets killed!) somewhat randomly. It is easily reproducible through something like: ./TestRunner -r 100 frysk.proc.TestTaskTerminateObserver Which always fails within a couple of rounds with as output: frysk.sys.Errno$Einval: tkill: Invalid argument (task 0)
On x86-64; look the same? Running testGetTasks(frysk.proc.TestProcGet) ...ERROR frysk.sys.Errno$Einval: tkill: Invalid argument (task 0, signal SIGCONT (18)) Running testGetChildren(frysk.proc.TestProcGet) ...Killed
testMultiThreadedRequestRemoveDetached(frysk.proc.TestProcForceDetach)java.lang.RuntimeException: teardown process invalid: 0 at frysk.testbed.TearDownProcess.add(TestRunner) at frysk.testbed.TearDownProcess$1.build(TestRunner) at frysk.sys.proc.ProcBuilder.scan(TestRunner) at frysk.sys.proc.ProcBuilder.construct(TestRunner) at frysk.sys.proc.ProcBuilder.construct(TestRunner) at frysk.testbed.TearDownProcess.tearDown(TestRunner) at frysk.testbed.TestLib.tearDown(TestRunner) at frysk.junit.Runner.runCases(TestRunner) at frysk.junit.Runner.runTestCases(TestRunner) at TestRunner.main(TestRunner)
This likely means that readdir of /proc/PID/task is finding a dirent entry of "0".
s/is likely/is/ This trace is from the code scaning /proc/PID/task/TID and finding Running testTerminateKillINT(frysk.proc.TestTaskTerminateObserver) ...WARNING: 00:00:48.970 11317.11317 frysk.sys.proc.ProcBuilder: /proc/13589/task contains bad pid: 0 ERROR java.lang.RuntimeException: /proc/$$/task contains bad pid (pid 0) pushing upstrea.
Fedora https://bugzilla.redhat.com/show_bug.cgi?id=438414