Re: [PATCH] Fix pr16789 failures on the SPARC

[Bryce McKinlay <mckinlay at redhat dot com>]

Tom Tromey wrote:

> Bryce> In any case, libgcj should not be considered frozen -
> Bryce> especially for bug fixes - just because the compiler is in
> Bryce> regression-only mode.
>
> Mark specifically asked language maintainers not to violate the
> regression-only rule.  It is unclear if this really applied to Java
> (given earlier statements of his, like at the Summit), but we never
> asked... but now we don't have to, today he change the status to
> "normal stage 3", meaning ordinary bug fixes are ok.
>  

Fair enough. Next time, we should ask if there is any doubt, but I have 
no reason not assume that Mark's multiple previous statements that the 
stage rules don't apply strictly to Java are still in effect. Changes to 
the front end are, perhaps, a gray-area, but for libgcj it doesn't make 
sense to me to hold back development on account of the compiler 
freezing. Its a different matter once we actually branch, but going for 
weeks without any tree where new development can be done hurts us a lot, 
I think. Of course, we'll need to get stricter once we reach a point 
when we're really claiming ABI stability and have a lot of applications 
depending on that.

Regards

Bryce