[9] RFR of 8078672: Print and allow setting by Java property seeds used to initialize Random instances in java.lang numerics tests

Roger Riggs Roger.Riggs at Oracle.com
Tue Apr 28 16:46:01 UTC 2015


Hi,

Note that there is some existing support for Random (and predetermined)
seeds in the Hotspot test library.
See hotspot/test/testlibrary/.../Utils.java

Soon(ish) to be coming to a testlibrary near you.

Roger


On 4/28/2015 3:57 AM, Chris Hegarty wrote:
> On 28 Apr 2015, at 01:44, Brian Burkhalter <brian.burkhalter at oracle.com> wrote:
>
>> Hi Joe,
>>
>> On Apr 27, 2015, at 5:32 PM, joe darcy <joe.darcy at oracle.com> wrote:
>>
>>> The patch looks pretty good. A few questions / comments:
>>>
>>> The test ParseHexFloatingPoint.java didn't get the "(use -Dseed=X to set PRNG seed)" addition; was that intentional?
>> No, it was an oversight. Thanks for catching it. I thought it had been there. Possibly something happened in moving the path to a different issue.
>>
>>> To save some space on the @summary line
>>>
>>>    Numerical tests for hexadecimal inputs to parseDouble, parseFloat
>>>
>>> could be replaced with
>>>
>>>    Numerical tests for hexadecimal inputs to parse{Double, Float}
>> Will change.
>>
>>> Could the java/math/RandomSeed class you added for java.math be used for this changeset instead of another another class with a large subset of the functionality?
>>>
>>> Perhaps the random number handling should be moved up to somewhere under jdk/test/lib? That would more easily let other type who want to handling random numbers in a uniform way use the library (at the cost of slightly more complicated jtreg tagging).
> I think having support for this in the jdk/test/lib/testlibrary/jdk/testlibrary would be good. I was going to suggest that too when the first changeset was pushed.
>
> -Chris.
>
>> I’ll investigate tomorrow whether perhaps this functionality can be moved to a higher level with removal of duplication of code. The java/math/RandomSeed class changeset was the first in which I used @library so I should look over the jtreg tag list again.
>>
>> Thanks,
>>
>> Brian




More information about the core-libs-dev mailing list