On 23 September 2010 18:43, Pekka Enberg <penb...@kernel.org> wrote: > From: Roland Brand <roland.br...@ergon.ch> > > As explained by AdamB: > > On Linux, Classpath appears to implement the System.nanoTime() function using > the gettimeofday function (see java_lang_VMSystem.c). Since gettimeofday > returns a wall-clock time this makes it unreliable for measuring elapsed > time, > (because the system clock could be changed by another process at any moment. > > According to the Sun javadocs: > > ... This method can only be used to measure elapsed time and is not > related to any other notion of system or wall-clock time. The > value returned represents nanoseconds since some fixed but > arbitrary time ... > > In short, gettimeofday is not appropriate for measuring elapsed time. > > This patch fixes System.nanoTime() to use CLOCK_MONOTONIC which is what > OpenJDK > does as well. > > Cc: Andrew Hughes <gnu_and...@members.fsf.org> > Cc: AdamB <ad...@videx.com> > Cc: Mark Wielaard <m...@klomp.org> > Cc: Roland Brand <roland.br...@ergon.ch> > Bugzilla-URL: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44411 > Signed-off-by: Pekka Enberg <penb...@kernel.org> > --- > v1 -> v2: Use strerror() to pass human readable error message to FatalError() > as suggested by Andrew Hughes > > native/jni/java-lang/java_lang_VMSystem.c | 30 > ++++++++++++++++++++++++++++- > vm/reference/java/lang/VMSystem.java | 5 +--- > 2 files changed, 30 insertions(+), 5 deletions(-) > > diff --git a/native/jni/java-lang/java_lang_VMSystem.c > b/native/jni/java-lang/java_lang_VMSystem.c > index d203227..4ff7355 100644 > --- a/native/jni/java-lang/java_lang_VMSystem.c > +++ b/native/jni/java-lang/java_lang_VMSystem.c > @@ -39,8 +39,10 @@ exception statement from your version. */ > > #include <jcl.h> > > +#include <time.h> > #include <sys/time.h> > #include <stdlib.h> > +#include <errno.h> > > /* > * Class: java_lang_VMSystem > @@ -121,6 +123,32 @@ Java_java_lang_VMSystem_nanoTime > (JNIEnv * env __attribute__ ((__unused__)), > jclass thisClass __attribute__ ((__unused__))) > { > + jlong result; > + struct timespec tp; > + > + if (clock_gettime (CLOCK_MONOTONIC, &tp) == -1) { > + char error[64]; > + snprintf(error, 64, "clock_gettime call failed: %s.", strerror(errno)); > + (*env)->FatalError (env, error); > + } > + > + result = (jlong) tp.tv_sec; > + result *= (jlong)1000000000L; > + result += (jlong)tp.tv_nsec; > + > + return result; > +} > + > +/* > + * Class: java_lang_VMSystem > + * Method: currentTimeMillis > + * Signature: ()J > + */ > +JNIEXPORT jlong JNICALL > +Java_java_lang_VMSystem_currentTimeMillis > + (JNIEnv * env __attribute__ ((__unused__)), > + jclass thisClass __attribute__ ((__unused__))) > +{ > /* Note: this implementation copied directly from Japhar's, by Chris > Toshok. */ > jlong result; > struct timeval tp; > @@ -131,7 +159,7 @@ Java_java_lang_VMSystem_nanoTime > result = (jlong) tp.tv_sec; > result *= (jlong)1000000L; > result += (jlong)tp.tv_usec; > - result *= (jlong)1000; > + result /= (jlong)1000L; > > return result; > } > diff --git a/vm/reference/java/lang/VMSystem.java > b/vm/reference/java/lang/VMSystem.java > index a194663..0b3d692 100644 > --- a/vm/reference/java/lang/VMSystem.java > +++ b/vm/reference/java/lang/VMSystem.java > @@ -129,10 +129,7 @@ final class VMSystem > * @return the current time > * @see java.util.Date > */ > - static long currentTimeMillis() > - { > - return nanoTime() / 1000000L; > - } > + static native long currentTimeMillis(); > > /** > * <p> > -- > 1.7.0.4 > > >
Thanks! strerror was the function I wanted but I couldn't think of it at the time, and the perror manpage listed sys_errlist. I'll check out the copyright situation on the patch and commit it as soon as I can. -- Andrew :-) Free Java Software Engineer Red Hat, Inc. (http://www.redhat.com) Support Free Java! Contribute to GNU Classpath and the OpenJDK http://www.gnu.org/software/classpath http://openjdk.java.net PGP Key: 94EFD9D8 (http://subkeys.pgp.net) Fingerprint: F8EF F1EA 401E 2E60 15FA 7927 142C 2591 94EF D9D8
