Test of primitive types

Test of primitive types

[Daniel Bonniot]

I searched in Mauve CVS for tests of primitive types and operations (integer
arithmetic, promotion, ...), and other core java features. It seems there is none,
but only tests of core library classes. 
Did I overlook some part? Would it make sense for Mauve to include such tests? 
Maybe not, as the FAQ says Mauve is a test for the core libraries. But these tests
would be useful for java (and alikes) compiler writers.
Is anyone aware of such tests avaliable out there?

Daniel

PS: please reply to sender too, I do not belong to the list

Re: Test of primitive types

[Artur Biesiadowski]

Daniel Bonniot wrote:
> 
> I searched in Mauve CVS for tests of primitive types and operations (integer
> arithmetic, promotion, ...), and other core java features. It seems there is none,
> but only tests of core library classes.
> Did I overlook some part? Would it make sense for Mauve to include such tests?
> Maybe not, as the FAQ says Mauve is a test for the core libraries. But these tests
> would be useful for java (and alikes) compiler writers.

There is a problem with such tests in current framework. If you want to
test such things, you cannot rely on testing framework written in java.
You cannot check if 2+2==4 in same code which later compares this answer
with correct one and adds passed/failures together :)

Same goes for many other core features - if jvm needs testing for such
things, it is probable that it might break at any moment in unexpected
way. For such kind of stuff I think you should use testing framework
from other environment - bash/expect and fork separate jvm for every
test.

Mauve is core lib test, not jvm test and I believe it should focus and
stay on that subject. Nailing few jvm test on top of it won't help
anything. Full jvm testing is a subject for different project. And java
compiler testing is a subject for yet another one (in this case you
would need to have fully trusted jvm and fully trusted core lib). 

Just my $0.02.

Artur

Re: Test of primitive types

[Peter Naulls]

In message < 3ADF6533.89684BE6@pg.gda.pl >
          Artur Biesiadowski <abies@pg.gda.pl> wrote:

> Same goes for many other core features - if jvm needs testing for such
> things, it is probable that it might break at any moment in unexpected
> way. For such kind of stuff I think you should use testing framework
> from other environment - bash/expect and fork separate jvm for every
> test.
> 
> Mauve is core lib test, not jvm test and I believe it should focus and
> stay on that subject. Nailing few jvm test on top of it won't help
> anything. Full jvm testing is a subject for different project. And java
> compiler testing is a subject for yet another one (in this case you
> would need to have fully trusted jvm and fully trusted core lib). 

FWIW, I have been testing a JVM using Mauve - and to quite a bit of
success.  Although it certainly doesn't test anywhere near all the
things I'd like, it went a long way towards fixing some major bugs.

(reordered)

> There is a problem with such tests in current framework. If you want to
> test such things, you cannot rely on testing framework written in java.
> You cannot check if 2+2==4 in same code which later compares this answer
> with correct one and adds passed/failures together :)

The comparison would probably be the textual result - Mauve already does
things like this in places.  If you haven't got basic string stuff
working, then you're not ready for a test suite just yet :-)  This is
why "Hello World" is such a crucial test - if you got basic string I/O
working, you can check the results of many things.

Peter



-- 
------------------------------------------------------------------------
 Peter Naulls - peter@erble.freeserve.co.uk -  http://chocky.mine.nu/
 Java for RISC OS and ARM - http://www.java.riscos.org.uk/
 Debian Linux on RiscPCs - http://www.erble.freeserve.co.uk/
------------------------------------------------------------------------

Re: Test of primitive types

[Mark Wielaard]

On Thu, Apr 19, 2001 at 11:29:44PM +0200, Daniel Bonniot wrote:
> 
> I searched in Mauve CVS for tests of primitive types and operations (integer
> arithmetic, promotion, ...), and other core java features. It seems there is none,
> but only tests of core library classes. 
> Did I overlook some part? Would it make sense for Mauve to include such tests? 
> Maybe not, as the FAQ says Mauve is a test for the core libraries. But these tests
> would be useful for java (and alikes) compiler writers.
> Is anyone aware of such tests avaliable out there?
> 
You might want to look at Jacks (Jacks is an Automated Compiler Killing Suite)
< http://oss.software.ibm.com/developerworks/opensource/jacks/ >

Cheers,

Mark
-- 
Stuff to read:
    < http://www.toad.com/gnu/whatswrong.html >
  What's Wrong with Copy Protection, by John Gilmore

Re: Test of primitive types

[Tom Tromey]

>>>>> "Artur" == Artur Biesiadowski <abies@pg.gda.pl> writes:

Artur> Same goes for many other core features - if jvm needs testing
Artur> for such things, it is probable that it might break at any
Artur> moment in unexpected way. For such kind of stuff I think you
Artur> should use testing framework from other environment -
Artur> bash/expect and fork separate jvm for every test.

I agree.

Artur> Mauve is core lib test, not jvm test and I believe it should
Artur> focus and stay on that subject. Nailing few jvm test on top of
Artur> it won't help anything. Full jvm testing is a subject for
Artur> different project. And java compiler testing is a subject for
Artur> yet another one (in this case you would need to have fully
Artur> trusted jvm and fully trusted core lib).

Jacks exists for compiler testing.  I definitely advocate that Java
compiler developers work with the Jacks project when writing tests.  I
intend to push gcj in that direction as time permits.  I've already
occasionally run gcj against Jacks and tried to fix bugs.

I'm unaware of any JVM testing project.  There's nothing preventing us
from making a new module in the Mauve CVS if someone wants to pursue
this.

Tom