8218458: [TESTBUG] runtime/NMT/CheckForProperDetailStackTrace.java fails with Expected stack trace missing from output
Reviewed-by: cjplummer, zgu
This commit is contained in:
parent
bb31c7d085
commit
d7097cb0f4
@ -78,7 +78,6 @@ gc/metaspace/CompressedClassSpaceSizeInJmapHeap.java 8193639 solaris-all
|
||||
|
||||
# :hotspot_runtime
|
||||
|
||||
runtime/NMT/CheckForProperDetailStackTrace.java 8218458 generic-all
|
||||
runtime/SharedArchiveFile/SASymbolTableTest.java 8193639 solaris-all
|
||||
runtime/containers/docker/TestCPUSets.java 8220672 generic-all
|
||||
runtime/jni/terminatedThread/TestTerminatedThread.java 8219652 aix-ppc64
|
||||
|
@ -23,6 +23,7 @@
|
||||
|
||||
/*
|
||||
* @test
|
||||
* @bug 8133747 8218458
|
||||
* @key nmt
|
||||
* @summary Running with NMT detail should produce expected stack traces.
|
||||
* @library /test/lib
|
||||
@ -36,26 +37,50 @@ import jdk.test.lib.process.OutputAnalyzer;
|
||||
import java.util.regex.Matcher;
|
||||
import java.util.regex.Pattern;
|
||||
|
||||
/**
|
||||
* We are checking for details that should be seen with NMT detail enabled.
|
||||
* In particular the stack traces from os::malloc call sites should have 4
|
||||
* (based on NMT detail stack depth setting) 'interesting' frames and skip
|
||||
* the higher-level allocation frames and frames specific to the NMT logic.
|
||||
* The actual stack trace is affected by the native compiler's inlining ability
|
||||
* and the type of build, so we need to check for a number of possible stacks.
|
||||
* This information does not change often enough that we are concerned about the
|
||||
* stability of this test - rather we prefer to detect changes in compiler behaviour
|
||||
* through this test and update it accordingly.
|
||||
*/
|
||||
public class CheckForProperDetailStackTrace {
|
||||
|
||||
/* The stack trace we look for by default. Note that :: has been replaced by .*
|
||||
to make sure it maches even if the symbol is not unmangled. */
|
||||
public static String stackTraceDefault =
|
||||
to make sure it matches even if the symbol is not unmangled.
|
||||
*/
|
||||
private static String stackTraceDefault =
|
||||
".*Hashtable.*allocate_new_entry.*\n" +
|
||||
".*ModuleEntryTable.*new_entry.*\n" +
|
||||
".*ModuleEntryTable.*locked_create_entry.*\n" +
|
||||
".*Modules.*define_module.*\n";
|
||||
|
||||
/* The stack trace we look for on Solaris and Windows slowdebug builds. For some
|
||||
reason ALWAYSINLINE for AllocateHeap is ignored, so it appears in the stack strace. */
|
||||
public static String stackTraceAllocateHeap =
|
||||
/* Alternate stacktrace that we check if the default fails, because
|
||||
new_entry may be inlined.
|
||||
*/
|
||||
private static String stackTraceAlternate =
|
||||
".*Hashtable.*allocate_new_entry.*\n" +
|
||||
".*ModuleEntryTable.*locked_create_entry.*\n" +
|
||||
".*Modules.*define_module.*\n" +
|
||||
".*JVM_DefineModule.*\n";
|
||||
|
||||
/* The stack trace we look for on AIX, Solaris and Windows slowdebug builds.
|
||||
ALWAYSINLINE is only a hint and is ignored for AllocateHeap on the
|
||||
aforementioned platforms. When that happens allocate_new_entry is
|
||||
inlined instead.
|
||||
*/
|
||||
private static String stackTraceAllocateHeap =
|
||||
".*AllocateHeap.*\n" +
|
||||
".*ModuleEntryTable.*new_entry.*\n" +
|
||||
".*ModuleEntryTable.*locked_create_entry.*\n" +
|
||||
".*Modules.*define_module.*\n";
|
||||
|
||||
/* A symbol that should always be present in NMT detail output. */
|
||||
private static String expectedSymbol =
|
||||
"locked_create_entry";
|
||||
private static String expectedSymbol = "locked_create_entry";
|
||||
|
||||
public static void main(String args[]) throws Exception {
|
||||
ProcessBuilder pb = ProcessTools.createJavaProcessBuilder(
|
||||
@ -67,23 +92,22 @@ public class CheckForProperDetailStackTrace {
|
||||
|
||||
output.shouldHaveExitValue(0);
|
||||
|
||||
// We should never see either of these frames because they are supposed to be skipped. */
|
||||
// We should never see either of these frames because they are supposed to be skipped.
|
||||
output.shouldNotContain("NativeCallStack::NativeCallStack");
|
||||
output.shouldNotContain("os::get_native_stack");
|
||||
|
||||
// AllocateHeap shouldn't be in the output because it is supposed to always be inlined.
|
||||
// We check for that here, but allow it for Aix, Solaris and Windows slowdebug builds
|
||||
// because the compiler ends up not inlining AllocateHeap.
|
||||
Boolean okToHaveAllocateHeap =
|
||||
Platform.isSlowDebugBuild() &&
|
||||
(Platform.isAix() || Platform.isSolaris() || Platform.isWindows());
|
||||
Boolean okToHaveAllocateHeap = Platform.isSlowDebugBuild() &&
|
||||
(Platform.isAix() || Platform.isSolaris() || Platform.isWindows());
|
||||
if (!okToHaveAllocateHeap) {
|
||||
output.shouldNotContain("AllocateHeap");
|
||||
}
|
||||
|
||||
// See if we have any stack trace symbols in the output
|
||||
boolean hasSymbols =
|
||||
output.getStdout().contains(expectedSymbol) || output.getStderr().contains(expectedSymbol);
|
||||
boolean hasSymbols = output.getStdout().contains(expectedSymbol) ||
|
||||
output.getStderr().contains(expectedSymbol);
|
||||
if (!hasSymbols) {
|
||||
// It's ok for ARM not to have symbols, because it does not support NMT detail
|
||||
// when targeting thumb2. It's also ok for Windows not to have symbols, because
|
||||
@ -92,21 +116,37 @@ public class CheckForProperDetailStackTrace {
|
||||
return; // we are done
|
||||
}
|
||||
output.reportDiagnosticSummary();
|
||||
throw new RuntimeException("Expected symbol missing missing from output: " + expectedSymbol);
|
||||
throw new RuntimeException("Expected symbol missing from output: " + expectedSymbol);
|
||||
}
|
||||
|
||||
/* Make sure the expected NMT detail stack trace is found. */
|
||||
String expectedStackTrace =
|
||||
(okToHaveAllocateHeap ? stackTraceAllocateHeap : stackTraceDefault);
|
||||
if (!stackTraceMatches(expectedStackTrace, output)) {
|
||||
output.reportDiagnosticSummary();
|
||||
throw new RuntimeException("Expected stack trace missing missing from output: " + expectedStackTrace);
|
||||
// Make sure the expected NMT detail stack trace is found
|
||||
System.out.println("Looking for a stack matching:");
|
||||
if (okToHaveAllocateHeap) {
|
||||
System.out.print(stackTraceAllocateHeap);
|
||||
if (stackTraceMatches(stackTraceAllocateHeap, output)) {
|
||||
return;
|
||||
}
|
||||
} else {
|
||||
System.out.print(stackTraceDefault);
|
||||
if (!stackTraceMatches(stackTraceDefault, output)) {
|
||||
System.out.println("Looking for alternate stack matching:");
|
||||
System.out.print(stackTraceAlternate);
|
||||
if (stackTraceMatches(stackTraceAlternate, output)) {
|
||||
return;
|
||||
}
|
||||
} else {
|
||||
return;
|
||||
}
|
||||
}
|
||||
// Failed to match so dump all the output
|
||||
output.reportDiagnosticSummary();
|
||||
throw new RuntimeException("Expected stack trace missing from output");
|
||||
}
|
||||
|
||||
public static boolean stackTraceMatches(String stackTrace, OutputAnalyzer output) {
|
||||
Matcher stdoutMatcher = Pattern.compile(stackTrace, Pattern.MULTILINE).matcher(output.getStdout());
|
||||
Matcher stderrMatcher = Pattern.compile(stackTrace, Pattern.MULTILINE).matcher(output.getStderr());
|
||||
Pattern p = Pattern.compile(stackTrace, Pattern.MULTILINE);
|
||||
Matcher stdoutMatcher = p.matcher(output.getStdout());
|
||||
Matcher stderrMatcher = p.matcher(output.getStderr());
|
||||
return (stdoutMatcher.find() || stderrMatcher.find());
|
||||
}
|
||||
}
|
||||
|
Loading…
x
Reference in New Issue
Block a user