RFR: 8242188: [TESTBUG] error in jtreg test jdk/jfr/api/consumer/TestRecordedFrame.java on linux-aarch64

Nick Gasson nick.gasson at arm.com
Wed May 13 06:01:55 UTC 2020


OK, thanks for the review! 

--
Nick

On 05/12/20 19:38 pm, Erik Gahlin wrote:
> I think what you have is fine.
>
> If we remove -Xcomp we depend on OSR happening after 100 000 iteration which seems brittle.
>
> At some point, I will rewrite some of the test so they use Event Streaming. In this case, it could retry if an event is not compiled.
>
> Erik
>
>> On 12 May 2020, at 12:40, Nick Gasson <nick.gasson at arm.com> wrote:
>> 
>> Hi Erik,
>> 
>>> Looks reasonable (but perhaps still fragile??)
>>> 
>> 
>> Yes it is a bit fragile. Another way of doing it is to remove the -Xcomp
>> and let main() loop enough to trigger OSR compilation. Something like
>> this:
>> 
>> 
>>    public static int counter = 0;
>> 
>>    public static void main(String[] args) throws IOException {
>>        if ("false".equals(System.getProperty("interpreted"))) {
>>            for (int i = 0; i < 100000; i++) {
>>                counter++;
>>            }
>>        }
>> 
>>        test(); // Records the line number and BCI for the main method/frame
>>    }
>> 
>> 
>> What do you think?
>> 
>> --
>> Thanks,
>> Nick



More information about the hotspot-jfr-dev mailing list