Re: [platform-releng-dev] improved checking for internal references

[Olivier Thomann <Olivier_Thomann@xxxxxxxxxx>]

As soon as you use a 1.5 constructs
(annotations, new for loop, generics, static imports, enums, ....) you
need -source 1.5 which requires -target 1.5. So the generated .class file
can only run on a 1.5 VM. If you don't use 1.5 constructs, then you can
change the source level to what you want 1.4 ou less and have a compatible
target level.

Hope I am clear that time :-),

Olivier

Jeff McAffer/Ottawa/IBM@IBMCA
Sent by: platform-releng-dev-admin@xxxxxxxxxxx

2005-02-17 17:00

Please respond to
platform-releng-dev

To	platform-releng-dev@xxxxxxxxxxx
cc
Subject	Re: [platform-releng-dev] improved checking for internal references

Just to be clear, with the 1.5 support you can no longer generate class
files for any other Java (1.4, 1.3, ...)? Or is it that you have to flip
the big switch and under the covers a 1.4, 1.3, ... compiler is actually
being used? 

Jeff

Olivier Thomann/Ottawa/IBM@IBMCA
Sent by: platform-releng-dev-admin@xxxxxxxxxxx

02/17/2005 04:18 PM

Please respond to
platform-releng-dev

To	platform-releng-dev@xxxxxxxxxxx
cc
Subject	Re: [platform-releng-dev] improved checking for internal references

Maybe I missed something, but when you use annotations, you need a 1.5
compiler. 
Even if you use the SOURCE retention policy, the generated .class file
can run only on a 1.5 VM because of the major version being 49 instead
of 48 for 1.4 VM. So a 1.4 VM could not run the generated .class file even
if it doesn't contain any reference to annotations.

So I don't think this is compatible with the limitation to a subset of
the libraries. 

Just my 2 cents, 

Olivier

Steven Wasleski <wasleski@xxxxxxxxxx>
Sent by: platform-releng-dev-admin@xxxxxxxxxxx

2005-02-17 15:56

Please respond to
platform-releng-dev

To	platform-releng-dev@xxxxxxxxxxx
cc
Subject	Re: [platform-releng-dev] improved checking for internal references

Good tooling will be desirable for what ever annotation method(s) we choose
or we run the risk of annotation and intention getting out of sync.  I
think we agree that a single source for the information is best now or
not so long from now.

As for Java 5 Annotations vs Javadoc, there are good arguments both ways
and the RCP point from your comment and the bug are well taken.  However,
it might be possible to use these annotations even in this case.  The
Java 5 Annotations provide for three different retention policies for different
annotation types (http://java.sun.com/j2se/1.5.0/docs/api/java/lang/annotation/RetentionPolicy.html).
 One of them directs the compiler to throw them away.  This essentially
makes them fancy Javadoc comments as far as I can tell, but perhaps we
could have non-RCP types use a longer retention policy.

I think I could be convinced either way on the Java 5 Annotations vs Javadoc
debate.  I think the decisive point would be which one could be more
easily and completely tooled, both for authoring annotations for an api
and enforcing them as others write code that uses an api.

Thanks,
Steve