Wouldn't it make more sense for the logic of converting a string into
a proper argument to live over in Flag instead of being buried in
attachListener? It seems more likely that no one will forget that way
and I think the result would be quite a bit shorter. It looks ok as
it too.
tom
On Nov 4, 2009, at 11:49 AM, Paul Hohensee wrote:
6898160: Need serviceability support for new vm argument type
'uint64_t'.
Webrev here
http://cr.openjdk.java.net/~phh/6898160/webrev.01/
A new vm argument type, 'uint64_t', was added as part of 6887571.
The jdk library
code in jdk/share/native/sun/management/Flag.c that fetches all the
vm product flags
throws an exception if it doesn't understand the type of an
argument. The change
for 6887571 didn't include full serviceability support for uint64_t.
The fix is to add serviceability support for uint64_t. In addition,
if a flag has a type
unknown to the jdk library, we now assert in debug builds and ignore
it in product
builds. Thus, no jdk library change is necessary.
The failing test,
jdk/test/com/sun/management/HotSpotDiagnosticMXBean/
GetDiagnosticOptions.java,
now passes.
Thanks,
Paul
