[kaffe] Mauve test results

[Dalibor Topic]

--- Audeon Guillaume <AUDEONG@thmulti.com> wrote:
> I've ran the Mauve test suite (downloaded today
> 11-June-2002) on Kaffe and
> get the following results:
gnu.testlet.java.net.DatagramSocket.DatagramSocketTest:
> Error :
> test_Basics failed - 5 Not able to create a socket 
> (number 1)
> 
> Then Kaffe blocks indefinitely. It has already been
> reported that Kaffe has
> some problems with Unicode characters.

I think that you are not running the latest mauve from
CVS. This issue should have been fixed, check
http://sources.redhat.com/ml/mauve-discuss/2002-q2/msg00025.html
for reference.

> The Mauve tests have been configured with the
> following selection keys:
> 
> JDK1.1
> !java.awt

there are no awt tests in mauve.

> !java.security.acl
> !java.security.cert
> !java.security.interfaces
> !java.security.spec

there are no security tests in mauve.

> !java.sql

sql only breaks (i.e. the test programs don't compile)
if you are running mauve's make check  with
KEYS="JDK1.2" The problem is that kaffe does not
implement java.sql.Blob, Clob and Array.

> !javax.accessibility

there are no javax.* tests in mauve.

> Did I miss something?

when you get the latest version from cvs, the mauve
makefile might break down demanding aclocal-1.6a. That
won't work if you are using automake 1.6.1 for
example. You may need to recreate mauve's auto*
generated files. Take a look at kaffe's
developers/autogen.sh on how to do it.

> Can anyone help me in going on further with this
> testing?

There is a nice perl script in the japhar CVS that
will convert the results of a mauve check run with
TESTFLAGS="-verbose" into more pleasant HTML files ;)
I think it's time we put some mauve results on the web
site.

If you use the script, just pipe the output of mauve
check into it. Otherwise it's going to take ages to
fill your hard disk with megabytes of very verbose
checking information.

I'd better write a FAQ.mauve, I guess ;)

One last word regarding Character failures: about 2/3
of the 1019 failures I get on JDK 1.1 tests (using
latest kaffe from trunk) are due to my changes from
Fri May 24, which fixed the behaviour of digit and
getNumericalValue mthods. The JDK 1.4 spec has been
updated since 1.4 to clarify the behaviour of one of
them, while the other one is still underspecified. Its
semantics are explained mich better in the
corresponding java class libraries book by Chan.

I filed a bug with Sun's bug parade, saying that they
should clarify the spec. They haven't managed to
contact me in the last two weeks, and I wanted to see
if Sun would update their spec before I go off and
send the mauve developers a patch updating their
tests. (in fact, most of it is already there, just
commented out). 

cheers,

dalibor topic

__________________________________________________
Do You Yahoo!?
Yahoo! - Official partner of 2002 FIFA World Cup
http://fifaworldcup.yahoo.com