Request for review 7122219: Passed StringTableSize value not verified

[harold seigel]

Hi David,

Thanks for the comments.  I used 'rollover' instead of 'overflow' 
because I think of 'overflow' as something that causes a SIGFPE.

Also, I thought that "StringTable size" was perhaps a bit more 
understandable because the user is specifying the size of the 
StringTable, but maybe it's not more understandable?

Thanks again,
Harold

On 11/10/2012 7:43 PM, David Holmes wrote:
> On 10/11/2012 5:11 AM, harold seigel wrote:
>> Please review the following change to fix bug 7122219.
>>
>> Summary:  This change prevents the SIGFPE and SIGSEGV problems that can
>> occur when using this option. It checks that the specified value is >=
>> the current default (1009) and <= the maximum requestable heap size, and
>> issues an error if it is not.
>>
>> This webrev also deletes an unused create_table() method and adds a
>> check for integer rollover to method os::malloc().
>
> Minor nit: we generally refer to overflow rather than rollover.
>
>> With this change, if users request an invalid value, they will get an
>> error message like this:
>>
>>     % $JAVA_HOME/bin/java -XX:StringTableSize=1g -cp .. 
>> <java-main-class>
>>     StringTable size of 1073741824 is invalid; must be between 1009 and
>
> Shouldn't that be "StringTableSize of ..."
>
> Was the option of renaming StringTableSize to something reflecting 
> that it is actually the bucket count rejected? I know there is a 
> follow on CR in this area but giving this variable a more accurate 
> name doesn't seem unreasonable.
>
> That said I'm okay if this is pushed as-is.
>
> David
>
>>     536870911
>>     Error: Could not create the Java Virtual Machine.
>>     Error: A fatal exception has occurred. Program will exit.
>
>
>
>> Open webrev at http://cr.openjdk.java.net/~hseigel/bug_7122219
>> <http://cr.openjdk.java.net/%7Ehseigel/bug_7122219>
>> Bug link at http://bugs.sun.com/view_bug.do?bug_id=7122219
>>
>> The changes were tested with JCK, JPRT, JTREG, and UTE tests, and with
>> hand-run tests on 32-bit Linux and 64-bit Windows.