RFC 6910473: BigInteger negative bit length, value range, and future prospects

[Dmitry Nadezhin]

The updated patch is here:
http://cr.openjdk.java.net/~bpb/6910473/webrev.4/

It contains 4 new small tests - one for each bug:

6910473 BitLengthOverflow.java
8021203 DoubleValueOverflow.java
8021204 StringConstructorOverflow.java
8022780 DivisionOverflow.java

Each test catches possible OutOfMemoryError.
It passes with a message in System.out when this happens.

When I add -javaoption:-Xmx8g to jtreg command-line
these four tests (and also ExtremeShiftingTests) fail before the patch
and pass after the patch - all without OutOfMemoryError.

One of the tests (StringConstructorOverflow) passes because of
OutOfMemoryError
both before and after the patch with default jtreg command-line on my
computer .

SymmetricRangeTests.java is intentionally disabled by invalid jtreg tag "@
test".
It is for manual run only.





On Fri, Oct 18, 2013 at 9:58 PM, Alan Bateman <Alan.Bateman at oracle.com>wrote:

> On 18/10/2013 17:24, Dmitry Nadezhin wrote:
>
>> The WebRev with @ignore in SymmetricRangeTests is here:
>> http://cr.openjdk.java.net/~**bpb/6910473/webrev.3/<http://cr.openjdk.java.net/~bpb/6910473/webrev.3/>
>>
>> jtreg ignores the test with a message
>> "test result: Error. Test ignored: hugeMemory"
>> with default jtreg command-line;
>> jtreg runs it with additional switch "-ignore:run".
>>
> I don't think we want @ignore here as it will cause jtreg to report it as
> an error (unless run with -ignore:quiet). If we are adding the test but
> just not having it run automatically then it might be best to just drop the
> jtreg meta-data completely. Alternatively, if feasible, is to dial-down the
> test so that it runs without requiring a lot of resources, but perhaps a
> second mode for interactive runs that does the full thing.
>
> -Alan.
>