On 7/24/2010 6:14 AM, Alan Bateman wrote:
Daniel D. Daugherty wrote:
Greetings,
Porting my most recent batch of test fixes to OpenJDK6 revealed
two new bugs:
6971847 4/4 jmap '-histo:live' option is necessary for proper leak
detection
6971851 4/4 jmap prints an incorrect usage message when SA is not
present
6971847 was partially introduced in the original bug that introduced
these tests (6942989) and the remainder was introduced by the fix
for 6964018 which fixed the tests on Linux. Yes, fixing the tests on
Linux broke them for OpenJDK6 on Windows. 6971851 was introduced when
then 'jmap -histo:live' option was added back in JDK6.
I have a fix in hand for 6971847 and it gets the new
AnonLoggerWeakRefLeak
and LoggerWeakRefLeak tests working properly on OpenJDK6 on Windows.
This
should be (knock on wood) the last fix to these tests as the tests have
now been checked out via JPRT on all the configs of interest.
Here is the webrev URL:
http://cr.openjdk.java.net/~dcubed/6971847-webrev/0/
Thanks, in advance, for any reviews.
Dan
I assume you are running into this because the SA bits weren't shipped
with jdk6.
Yes and the fact that the jmap usage message is wrong (the other bug).
For the issue at hand, then the simplest solution may be just remove
the check for -histo:live from the tests. It shouldn't be needed
anyway because the live sub-option is there since jdk6.
The 'histo:live' check gives me a helpful usage message rather
than a vague one. I would prefer to keep the check. Would you
be okay with the fix as is?
Dan
