| Summary: | AckDaemonProcess doesn't appear to fully wait | ||
|---|---|---|---|
| Product: | frysk | Reporter: | Mike Cvet <mcvet> |
| Component: | general | Assignee: | Mike Cvet <mcvet> |
| Status: | RESOLVED FIXED | ||
| Severity: | normal | CC: | cagney, mark, mcvet |
| Priority: | P2 | ||
| Version: | unspecified | ||
| Target Milestone: | --- | ||
| Host: | Target: | ||
| Build: | Last reconfirmed: | ||
| Bug Depends on: | |||
| Bug Blocks: | 2935, 3346, 4289, 4394 | ||
Without the sleep call, may give errors such as the following: 1) testBacktrace(frysk.rt.TestStackBacktrace)java.lang.RuntimeException: Failed to attach to created proc at frysk.rt.SteppingEngine$ThreadLifeObservable.addFailed(TestRunner) at frysk.proc.LinuxPtraceTaskState$16.handleAddObservation(TestRunner) at frysk.proc.Task.handleAddObservation(TestRunner) at frysk.proc.TaskObservation.handleAdd(TestRunner) at frysk.proc.LinuxPtraceProcState$Attaching.allAttached(TestRunner) at frysk.proc.LinuxPtraceProcState$Attaching$ToMainTask.handleTaskAttachCompleted(TestRunner) at frysk.proc.Proc$4.execute(TestRunner) at frysk.event.EventLoop.runEventLoop(TestRunner) at frysk.event.EventLoop.runPolling(TestRunner) at frysk.proc.TestLib.assertRunUntilStop(TestRunner) at frysk.proc.TestLib.assertRunUntilStop(TestRunner) at frysk.rt.TestStackBacktrace.testBacktrace(TestRunner) at frysk.junit.Runner.runCases(TestRunner) at frysk.junit.Runner.runArchCases(TestRunner) at frysk.junit.Runner.runTestCases(TestRunner) at TestRunner.main(TestRunner) Caused by: java.lang.RuntimeException: destroyed at frysk.proc.LinuxPtraceTaskState$16.handleAddObservation(TestRunner) ...27 more 2) testThreadedBacktrace(frysk.rt.TestStackBacktrace)java.lang.RuntimeException: getting task's executable at frysk.proc.IsaFactory.getIsa(TestRunner) at frysk.proc.IsaFactory.getIsa(TestRunner) at frysk.proc.LinuxPtraceTask.sendrecIsa(TestRunner) at frysk.proc.Task.getIsa(TestRunner) at frysk.rt.StackCallbacks.<init>(TestRunner) at frysk.rt.StackFactory.createStackFrame(TestRunner) at frysk.rt.StackFactory.createStackFrame(TestRunner) at frysk.rt.TestStackBacktrace.firstTestBacktraceAssertions(TestRunner) at frysk.rt.TestStackBacktrace$LockObserver.update(TestRunner) at java.util.Observable.notifyObservers(libgcj.so.7rh) at frysk.rt.SteppingEngine$SteppingObserver.updateExecuted(TestRunner) at frysk.proc.Proc$14.add(TestRunner) at frysk.proc.LinuxPtraceTaskState$Attached.handleAddObservation(TestRunner) at frysk.proc.Task.handleAddObservation(TestRunner) at frysk.proc.TaskObservation.handleAdd(TestRunner) at frysk.proc.LinuxPtraceProcState$Attaching.allAttached(TestRunner) at frysk.proc.LinuxPtraceProcState$Attaching$ToOtherTasks.handleTaskAttachCompleted(TestRunner) at frysk.proc.Proc$4.execute(TestRunner) at frysk.event.EventLoop.runEventLoop(TestRunner) at frysk.event.EventLoop.runPolling(TestRunner) at frysk.proc.TestLib.assertRunUntilStop(TestRunner) at frysk.proc.TestLib.assertRunUntilStop(TestRunner) at frysk.rt.TestStackBacktrace.testThreadedBacktrace(TestRunner) at frysk.junit.Runner.runCases(TestRunner) at frysk.junit.Runner.runArchCases(TestRunner) at frysk.junit.Runner.runTestCases(TestRunner) at TestRunner.main(TestRunner) Caused by: lib.elf.ElfFileException: Could not open /proc/12927/exe at lib.elf.Elf.<init>(TestRunner) at frysk.proc.IsaFactory.getIsa(TestRunner) ...39 more Frysk uses SIGIO (SIGPOLL is an alias) internally. Try a signal like SIGUSR1. Sig.USR1 fixes the problem * TestStepping.java (testLineStepFunctionCall): Use Sig.USR1 instead of Sig.POLL. Fixes #4431. (testLineStepIfStatementPass): Ditto. (testLineStepIfStatementFail): Ditto. (testLineStepFunctionReturn): Ditto. (testStepSigLongJmp): Ditto. (testStepGoto): Ditto. (testStepSigRaise): Ditto. * TestSteppingEngine.java (testInstructionStepping): Ditto. (testLineStepping): Ditto. (testStepOver): Ditto. (testInstructionNext): Ditto. (testStepOut): Ditto. |
In most of the RT testcases, AckDaemonProcesses are used to start up a test process. However, often, (usually second iteration of the test), the test will fail becase the grabbed stack trace is not a proper stack trace of the process, and as noticed in #4394: "I grabbed some stack traces of the traced process, and ended up with varying results, including the following: #0 0x00002aaaaaab1a2b in _dl_map_object_from_fd () #1 0x00002aaaaaab2d9c in _dl_map_object () #2 0x00002aaaaaab610d in openaux () #3 0x00002aaaaaab7d26 in _dl_catch_error () #4 0x00002aaaaaab67d6 in _dl_map_object_deps () #5 0x00002aaaaaaae306 in dl_main () #6 0x00002aaaaaabde0b in _dl_sysdep_start () #7 0x00002aaaaaaac468 in _dl_start () #8 0x00002aaaaaaabb58 in [unknown] or #0 0x00002aaaaaabeff6 in strcmp () #1 0x00002aaaaaab3c87 in check_match.8134 () #2 0x00002aaaaaab4084 in do_lookup_x () #3 0x00002aaaaaab4282 in _dl_lookup_symbol_x () #4 0x00002aaaaaab57a5 in _dl_relocate_object () #5 0x00002aaaaaaae77a in dl_main () #6 0x00002aaaaaabde0b in _dl_sysdep_start () #7 0x00002aaaaaaac468 in _dl_start () #8 0x00002aaaaaaabb58 in [unknown] or #0 0x00002aaaaaab7bc3 in _dl_fixup () #1 0x00002aaaaaabd3a2 in _dl_runtime_resolve () #2 0x00002aaaaaccb628 in __pthread_initialize_minimal () #3 0x00002aaaaaccaf29 in _init () or just plain: #0 0x00002aaaaaabeff4 in [unknown]" The relevant code is usually something like: AckDaemonProcess process = new AckDaemonProcess (Sig.POLL, new String[] { getExecPath ("funit-rt-threader"), "" + Pid.get (), "" + Sig.POLL_ }); myTask = process.findTaskUsingRefresh(true); SteppingEngine.setProc(myTask.getProc()); However, the problem is mitigated by placing try { Thread.sleep(500); } catch (Exception e) {} between the initialization of the AckDaemonProcess, and the call to SteppingEngine. It appears as though the process needs time to fully load and start running before the test begins.