Re: [cp-patches] [patch] let gjavah accept -source 1.[567]

[Andrew Hughes <gnu.andrew@redhat.com>]

----- Original Message -----
> On Wed, Dec 19, 2012 at 02:01:10PM +0100, Matthias Klose wrote:
> > Currently gjavah only accepts -source 1.4 and lower, and errors out
> > for any
> > other value. Would it be reasonable to accept higher versions too?
> 
> I think that should be fine for gjavah, I cannot think of something
> in the bytecode that would impact jni/cni header generation.
> 
> But your patch is for gjdoc. There I think there are source
> constructs
> that might be a problem in newer versions. It should support some of
> the new 1.5 source level features, but I am not sure if it handles
> everything nor whether it handles any 1.6 and 1.7 extensions.
> 
> > Index: classpath/tools/gnu/classpath/tools/gjdoc/Main.java
> > ===================================================================
> > --- classpath/tools/gnu/classpath/tools/gjdoc/Main.java	(Revision
> > 194604)
> > +++ classpath/tools/gnu/classpath/tools/gjdoc/Main.java
> > 	(Arbeitskopie)
> > @@ -1339,10 +1310,13 @@
> >              option_source = args[0];
> >              if (!"1.2".equals(option_source)
> >                  && !"1.3".equals(option_source)
> > -                && !"1.4".equals(option_source)) {
> > +                && !"1.4".equals(option_source)
> > +                && !"1.5".equals(option_source)
> > +                && !"1.6".equals(option_source)
> > +                && !"1.7".equals(option_source)) {
> 
> If you really meant gjdoc I think it would be OK to try to accept it,
> but maybe with a warning message that it is untested?
> 
> Cheers,
> 
> Mark
> 
> 
> 

The question would make a lot more sense for gjdoc than gjavah :-)

As is, it does support 1.5 to a degree (enough to compile the Classpath
docs anyway; we should try OpenJDK at some point).  There are no source
changes in 1.6 and minimal ones in 1.7 so these should be easy to support,
if they aren't already.

I'm generally in favour of removing unnecessary version checks as they
only mask the real bugs (e.g. when the parser hits a construct it can't handle)
and would have to be allowed to actually implement the newer versions anyway.
I think this is more a case of neglected code (given there were 1.5 updates)
rather than a design choice.

Note that I have a JAPI run of gjdoc vs. javadoc's API which points out some
of the missing API (mainly, if not all, 1.5 stuff AFAICS) and I intend to look
at implementing this soon.

Thanks,
-- 
Andrew :)

Free Java Software Engineer
Red Hat, Inc. (http://www.redhat.com)

PGP Key: 248BDC07 (https://keys.indymedia.org/)
Fingerprint = EC5A 1F5E C0AD 1D15 8F1F  8F91 3B96 A578 248B DC07