Fwd: RFR 8073423: Remove LazyClassPathEntry support if no longer needed
Lois Foltan
lois.foltan at oracle.com
Thu Jun 25 19:06:27 UTC 2015
On 6/25/2015 9:25 AM, harold seigel wrote:
> Hi Lois,
>
> Thanks for the review.
>
> I tried the bad.jar / good.jar example that you mention below both
> with my proposed change and without my proposed change. In both
> cases, the JVM ignored bad.jar and successfully found the class in
> good.jar.
>
> Why do you think the JVM would stop traversing the BootClassPath if
> one of the jar files on the path couldn't be opened?
Well I wasn't advocating that the VM should stop traversing, I more just
wanted to understand if any changes in behavior would occur. Thanks for
the clarification. I am fine with this change.
Lois
>
> Thanks, Harold
>
> On 6/24/2015 7:30 PM, Lois Foltan wrote:
>>
>> On 6/24/2015 1:23 PM, harold seigel wrote:
>>> Hi,
>>>
>>> Please review this updated webrev for bug JDK-8073423:
>>> http://cr.openjdk.java.net/~hseigel/bug_8073423.2/
>>>
>>> The updated webrev differs from the previous one in two ways:
>>>
>>> 1. The LazyBootClassLoader option is deprecated instead of just
>>> deleted.
>>>
>>> 2. To preserve compatibility, bad jar files detected while setting up
>>> the boot class loader path are ignored. With the previous fix, they
>>> caused the JVM to terminate. (This caused test JTReg test
>>> hotspot/test/runtime/LoadClass/LoadClassNegative.java to fail.)
>>
>> Hi Harold,
>>
>> Looks good. Just for my clarification on #2, previously if there was
>> a bad jar file on the boot class loader's -Xbootclasspath/a, a
>> LazyClassPathEntry was created and an attempt was not made to open
>> the bad jar file until there was a need. Now, that bad jar file is
>> not put on the boot class path list at all, isn't this a change in
>> behavior? Someone could have have two .jar files in the
>> -Xbootclasspath/a directory specified, say bad.jar and good.jar.
>> Today if an attempt to open bad.jar, a ClassNotFoundException is
>> thrown. With this change, bad.jar is no longer on the boot class
>> path and the class being requested has the potential of being found
>> in good.jar, and the test passes.
>>
>> Thanks,
>> Lois
>>
>>>
>>> This updated webrev was tested with hotspot JTreg tests, JCK lang
>>> and VM tests, and UTE Quick tests. Additionally, the deprecation of
>>> the LazyBootClassLoader option was tested by hand.
>>>
>>> Thanks, Harold
>>>
>>> -------- Forwarded Message --------
>>> Subject: RFR 8073423: Remove LazyClassPathEntry support if no
>>> longer needed
>>> Date: Wed, 01 Apr 2015 15:06:25 -0400
>>> From: harold seigel <harold.seigel at oracle.com>
>>> Organization: Oracle Corporation
>>> To: hotspot dev runtime <hotspot-runtime-dev at openjdk.java.net>
>>>
>>>
>>>
>>> Hi,
>>>
>>> Please review this change to remove LazyClassPathEntry support from the
>>> JVM. With the demise of rt.jar, this is no longer used.
>>>
>>> Open webrev: http://cr.openjdk.java.net/~hseigel/bug_8073423/
>>>
>>> JBS Bug: https://bugs.openjdk.java.net/browse/JDK-8073423
>>>
>>> The change was tested with JCK lang, vm, and api tests, hotspot jtreg
>>> tests, and testbase split_verifier and quick tests. Additionally, I
>>> tested '$JAVA_HOME/bin/java -XX:+LazyBootClassLoader -version" to test
>>> for the deprecation message for -XX:+LazyBootClassLoader.
>>>
>>> Thanks, Harold
>>>
>>>
>>>
>>
>
More information about the hotspot-runtime-dev
mailing list