RFR 7014526: "java -version" crash on sparc with some values of HeapBaseMinAddress
Coleen Phillimore
coleen.phillimore at oracle.com
Tue Feb 18 17:51:55 UTC 2014
On 2/18/14 12:48 PM, Yumin Qi wrote:
> Hi,
>
> I joined late for this discussion. We parse 'G', 'M', 'K' and if
> there is no appendix to numbers, it is Byte, is this right? If so,
> using a number as default (G) will cause more confuse I think.
Yes, this is right, we do parse G, M and K. I only use G for the
default value but I've also put the whole number in because we don't
save that the user specified the number with G, M or K.
Coleen
>
> Thanks
> Yumin
>
> On 2/18/2014 9:41 AM, Coleen Phillimore wrote:
>>
>> On 2/18/14 11:13 AM, Thomas Schatzl wrote:
>>> Hi Coleen,
>>>
>>> On Tue, 2014-02-18 at 09:54 -0500, Coleen Phillimore wrote:
>>>> I fixed the indentation and I changed it to %d because it's %dM. The
>>>> %x was temporary for my debugging.
>>> iirc there is a UINTX_FORMAT constant specifically to be used as the
>>> format string for uintx values.
>>>
>>> Is it possible to also write the value selected by the user to make
>>> the problem more clear? I.e. "HeapBaseMinAddress must be at least
>>> ..., but
>>> is ...".
>>>
>>> Also I am not sure if dividing by M does not cut significant digits too
>>> much, but I guess people are mostly going to use MB here anyway.
>>
>> I divided by M to avoid this sort of output, which looks like a
>> meaningless bunch of digits to me.
>>
>> % java -XX:+PrintCommandLineFlags -version
>> -XX:InitialHeapSize=1035993088 -XX:MaxHeapSize=16575889408
>> -XX:+PrintCommandLineFlags -XX:+UseCompressedClassPointers
>> -XX:+UseCompressedOops -XX:+UseParallelGC
>>
>> How about:
>>
>> java -d64 -XX:HeapBaseMinAddress=128 -XX:+PrintMiscellaneous
>> -XX:+Verbose -version
>> ...
>> HeapBaseMinAddress must be at least 6442450944 (6G) which is greater
>> than value given 128
>> ...
>>
>> All the defaults are in G. If that changes we could make the message
>> divide by M.
>>
>> Thanks,
>> Coleen
>>
>>>
>>> Thanks,
>>> Thomas
>>>
>>>
>>>
>>
>
More information about the hotspot-runtime-dev
mailing list