RFR: 8216303: JFR: Simplify generated files

Erik Joelsson erik.joelsson at oracle.com
Thu May 28 20:55:00 UTC 2020


On 2020-05-28 13:51, Erik Gahlin wrote:
> Hello Erik,
>
>> On 28 May 2020, at 19:47, Erik Joelsson <erik.joelsson at oracle.com> wrote:
>>
>> Hello Erik,
>>
>> I noticed that you added an import for java.util.HashSet, but it doesn't seem to be used.
> Fixed.
>
>> You also replaced the only use of HashMap with LinkedHashMap, which we like as it gives a predictable iteration order, but forgot to remove the import.
>>
>> Do you know if the generated files are stable over several builds?
> They should be stable. Events are numbered after how they are specified in /hotspot/share/jfr/metadata/metadata.xml.

Good thanks! Then I'm happy with this.

/Erik

> Erik
>
>> /Erik
>>
>> On 2020-05-28 10:37, Erik Gahlin wrote:
>>> Hi,
>>>
>>> Could I have a review of a fix that removes legacy in the generated JFR event files, in particular hard coded numbers for type / event IDs. Event IDs are now numbered from zero, which has the benefit that most of them can be represented as a single byte instead of two. This will reduce the size of recordings with a few percent.
>>>
>>> I’m looping in build-dev, since I am modifying make/src/classes/build/tools/jfr/GenerateJfrFiles.java, but there are no changes to how things are built. The generated JFR files are just slightly different.
>>>
>>> Old and new files:
>>> http://cr.openjdk.java.net/~egahlin/8216303_generated/
>>>
>>> Bug:
>>> https://bugs.openjdk.java.net/browse/JDK-8216303
>>>
>>> Webrev:
>>> http://cr.openjdk.java.net/~egahlin/8216303/
>>>
>>> Testing:
>>> - Building with and without jfr the feature enabled
>>> - Tier 1/2 (Mac, Linux and Windows)
>>> - JFR tests in test/jdk/jdk/jfr
>>>
>>> Thanks
>>> Erik
>>>
>>>



More information about the build-dev mailing list