Re: Fix for test gnu/testlet/java/lang/Integer/parseInt.java

[Dr Andrew John Hughes <ahughes at redhat dot com>]

On 10:14 Thu 18 Aug     , Pavel Tisnovsky wrote:
> Dr Andrew John Hughes wrote:
> > On 18:20 Wed 17 Aug     , Pavel Tisnovsky wrote:
> >> Greetings,
> >>
> >> I think that the regression test "gnu/testlet/java/lang/Integer/parseInt.java"
> >> should be fixed to work correctly for pre JDK1.7 case (JDK1.0 .. JDK1.6) and
> >> also for JDK1.7 case. The difference between JDK1.6 and JDK1.7 is in the
> >> behaviour of method Integer.parseInt() when string beginning with '+' sign is
> >> parsed:
> >>
> >> Integer.parseInt("+42")
> >>
> >> JDK1.6 throws exception in this case, but parsing the same string is perfectly
> >> valid in JDK1.7. So I added the condition to the test which checks what JDK is
> >> being tested (I can't figure out how Mauve harness checks the "Tags: JDKxx"
> >> tag). I also changed the message printed in case JDK1.7 does not work correctly.
> >>
> >> Is it possible to push the fixed test to Mauve repository please?
> >>
> >> Thank you in advance,
> >> Pavel
> > 
> > Such tests are dubious because they assume compliance to one particular version,
> > which has never been true of GNU Classpath and can't be verified due to the lack
> > of a Free Software TCK.  There will also be versions of OpenJDK7 that report
> > as "1.7" but don't yet have this change.
> 
> Hi Andrew,
> 
> you are right that we have not FS TCK so the precise JRE behaviour is
> questionable, but I think that any JDK/JRE which identifies itself as "1.6"
> should conforms to "Java Platform, Standard Edition 6 API Specification", at
> least from common developers point of view.
> 

That doesn't reflect reality.  That only works if you make the assumption that
the only available JDK implementations implement a complete specification, which
fails with Free Software.  As I say, it's just as true for OpenJDK.  Will every
build of OpenJDK8 implement the whole 1.8 specification?  I'm pretty sure it will
be reporting 1.8 before the specification is signed off.  1.7 certainly did.

> > Have you tested this with GNU Classpath?  I'm pretty sure it will break the
> > test there as it has the modern behaviour, but most GNU Classpath VMs tend
> > to report earlier versions due to the level of bytecode they support.
> 
> Yeah, that's true because I've tried it using gij:
> $ gij -version
> java version "1.5.0"
> gij (GNU libgcj) version 4.6.0 20110603 (Red Hat 4.6.0-10)
> 
> Its interesting because it is identified itself as "1.5.0" and parse the string
> "+10" without problems.
> 

Yes, because it supports 1.5 *bytecode*.  We've implemented quite a few 1.6 methods
in GNU Classpath and this 1.7 behaviour is in parseInt.  1.7 methods will start
been added as needed.
-- 
Andrew :)

Free Java Software Engineer
Red Hat, Inc. (http://www.redhat.com)

Support Free Java!
Contribute to GNU Classpath and IcedTea
http://www.gnu.org/software/classpath
http://icedtea.classpath.org
PGP Key: F5862A37 (https://keys.indymedia.org/)
Fingerprint = EA30 D855 D50F 90CD F54D  0698 0713 C3ED F586 2A37