Request for review 7053130: hs_err file does not record specified CLASSPATH
Krystal Mo
krystal.mo at oracle.com
Tue Oct 16 14:25:16 PDT 2012
Hi Harold,
I'm okay with the change as-is, but wonder if this is a good-enough
tradeoff when Java agents are in the picture.
There's a method in Java agent that could change the effective system
class path, namely
java.lang.instrument.Instrumentation.appendToSystemClassLoaderSearch(JarFile
file). It's ultimately implemented by
JvmtiEnv::AddToSystemClassLoaderSearch(const char* segment), in
prims/jvmtiEng.cpp. As you can see from the example of this function,
the effective system class path doesn't have to be the one stored in
Arguments::_java_class_path; in fact, after java.lang.System class is
initialized, the copy in Arguments::_java_class_path doesn't matter anymore.
I'd like to hear other people's opinion on this issue.
Nitpicking: Why did you use PropertyList_get_value() to get the value of
_java_class_path, instead of just calling _java_class_path->value()?
If you're going to use PropertyList_get_value() then you don't need to
do the null check upfront because PropertyList_get_value() does that for
you already.
- Kris
On 2012/10/17 4:43, harold seigel wrote:
> Summary: Added code to write the value of the java.class.path
> property to the hs_err file. The new text in the hs_err file would
> look like this:
>
> ...
> VM Arguments:
> jvm_args: -XX:+CrashGCForDumpingJavaThread
> java_command: Queens
> java_class_path: .:/root/dir:/home
> Launcher Type: SUN_STANDARD
>
> Environment Variables:
> JAVA_HOME=/java/udrt/jdk1.8.0/
> CLASSPATH=.:/home/:/some/where
> ....
>
>
> Open webrev at http://cr.openjdk.java.net/~coleenp/Bug7053130
> <http://cr.openjdk.java.net/%7Ecoleenp/Bug7053130>
>
> Bug link at http://bugs.sun.com/view_bug.do?bug_id=7053130
>
> Tested by generating hs_err files and looking at their contents.
>
> Thanks, Harold
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.openjdk.java.net/pipermail/hotspot-runtime-dev/attachments/20121017/809d7b64/attachment.html
More information about the hotspot-runtime-dev
mailing list