Jython status.

Jython status.

[Alexandre Petit-Bianco]

I can now run Jython (given certain file configuration hacks which
should go away) enough so that it runs the Jython test suite (which is
based on the Python test suite.)

We're not behaving like the JDK interpreted Jython for this test:

  r(SystemExit)
  import sys
  try: sys.exit(0)
  except SystemExit: pass

We seem to be really calling exit, whereas the JDK ran one doesn't
quite quit and let the next test run. I'll have to debug.

The serialization tests are failing, I'm guessing this could be a
libgcj bug. Some other tests are failing because our libgcj isn't
complete enough, for example, we don't have `awt.Color.RGBtoHSB' which
is used to conduct a test on arrays...

At this point, I'm hitting the failures I would get with the regular
Jython, we're sometimes doing a little better because we are avoiding
Sun's JDK bugs.

I need to fix the general problem for Jython now which is what to find
where. It needs a path to some python files it finds only if I put
them right under its nose. It's also still necessary to put a symlink
to <libgcj_buildir>.../java in the cwd in order for Jython to be able
to load information about the Java objects.

Note that if I'm using gcj, I don't need the symlink, so the binaries
and the VM differ in terms of knowing where things could be found --
does it ring a bell to anyone?

./A

Re: Jython status.

[Anthony Green]

Alex wrote:
> I need to fix the general problem for Jython now which is what to find
> where. It needs a path to some python files it finds only if I put
> them right under its nose. It's also still necessary to put a symlink
> to <libgcj_buildir>.../java in the cwd in order for Jython to be able
> to load information about the Java objects.
>
> Note that if I'm using gcj, I don't need the symlink, so the binaries
> and the VM differ in terms of knowing where things could be found --
> does it ring a bell to anyone?

I'm not sure I understand all this.   Perhaps if I could see it in action...

Do you think you could import this stuff now?  BTW - remember to use "-ko" on
your cvs import in order to maintain the original rcs ID strings.  Not a big
deal - but I noticed that this wasn't done for gnu.readline.

AG

Re: Jython status.

[Alexandre Petit-Bianco]

Anthony Green writes:

> Do you think you could import this stuff now?

Pretty soon. But right now it can't be installed and expected to
work.

>  BTW - remember to use "-ko" on your cvs import in order to maintain
> the original rcs ID strings.  Not a big deal - but I noticed that
> this wasn't done for gnu.readline.

OK, I will. Thanks.

./A

Re: Jython status.

[Anthony Green]

Alex wrote:
> Pretty soon. But right now it can't be installed and expected to
> work.

I don't think that matters, as long as it builds at least. 

AG