jacks updates

jacks updates

[Tom Tromey]

Ok, I've checked in the minor Jacks bug fixes I've collected from the
recent past.  Mostly just typos, but also an additional test for
method overriding (it seems that we could use more here) and a real
fix to some test output.

I also have a patch to change jacks to use tclsh8.4 -- I made this
change locally since FC2 doesn't ship tclsh8.3.  If nobody objects
I'll make this change in the mauve cvs repository as well.

Also, I have some tests (some by me, some posted to the jacks list
this year) for 1.5 things: boxing, foreach, and enums.  In my
repository I have these in tests/non-jls/, but I suppose they probably
belong in tests/jls now.  Chris, what do you think?

Also there's the question of what to do about code that was invalid
in 1.4 but is now valid in 1.5.  For instance something like

    Object x = 5;

falls into this category due to auto-boxing.  We could either modify
these tests so that they are always invalid, or we could mark them so
that we can continue testing compliance to 1.4 and 1.5.  I'm inclined
toward the latter offhand.

Tom

Re: jacks updates

[Chris Abbey]

Yesterday, Tom Tromey wrote:
> I also have a patch to change jacks to use tclsh8.4 -- I made this
> change locally since FC2 doesn't ship tclsh8.3.  If nobody objects
> I'll make this change in the mauve cvs repository as well.

I've been keeping a local copy of what I bet is the same patch for ages
now... so I certainly have no objection to moving up to something modern.

> Also, I have some tests (some by me, some posted to the jacks list
> this year) for 1.5 things: boxing, foreach, and enums.  In my
> repository I have these in tests/non-jls/, but I suppose they probably
> belong in tests/jls now.  Chris, what do you think?

I've never really been good at picking the location for where tests should
go in the hierarchy... it's often a subjective issue.

Perhaps tests/jcr/NNN/  ?

> Also there's the question of what to do about code that was invalid
> in 1.4 but is now valid in 1.5.  For instance something like
...
>   we could mark them so
> that we can continue testing compliance to 1.4 and 1.5.

Personally, I'm inclined to like this idea... it's likely something we
should make extensible, so for example a test could be marked as
"expect fail on < 1.5, expect pass on >= 1.5" or similar. I think we ended
up throwing away some tests that were expecting different behaviour in pre
1.2 than was seen on 1.2 and higher... and I'm sure there are tests that
behave differntly on 1.0 than any other version... and we adjusted them
with a bias against 1.0. (But 1.0 was a very different transition than 1.5
is... there was very little 1.0 code, everyone was clamouring for 1.1... I
expect there will be a lot of Java2 code for a while, the move to Java5
won't be as fast.)

-- 
Never make a technical decision based upon the politics of the situation.
Never make a political decision based upon technical issues.
The only place these realms meet is in the mind of the unenlightened.
			-- Geoffrey James, The Zen of Programming

Re: jacks updates

[Tom Tromey]

Chris> I've never really been good at picking the location for where
Chris> tests should go in the hierarchy... it's often a subjective
Chris> issue.

Yeah.  I looked and there is already tests/non-jls/assert.
So how about tests/non-jls/<feature name>?

Chris> Personally, I'm inclined to like this idea... it's likely
Chris> something we should make extensible, so for example a test
Chris> could be marked as "expect fail on < 1.5, expect pass on >=
Chris> 1.5" or similar.

Sounds good, let's do it.

Tom