Re: ia64 libjava java-signal.h build failure

[Mark Mitchell <mark at codesourcery dot com>]

>>>>> "Alexandre" == Alexandre Oliva <aoliva@redhat.com> writes:

Thanks for your reply.

Your points about the --enable-languages=... stuff are valid -- but
orthogonal.  If that configury option is difficult/risky to use, then
we should improve it.

    Alexandre> How about doing it the other way round, taking into
    Alexandre> account that libgcj is already disabled for almost all
    Alexandre> platforms: enable GCJ by default, and, whenever we get
    Alexandre> a report that it fails to build on some platform, we
    Alexandre> disable it?

We are trying to avoid risk.  Your suggestion will create a broken
tree for those people where it causes a problem.  Then, they will have
to wait for us to fix the problem.

I think the burden should be on the people who want the feature
enabled by default.  

It sounds like all we need is a one-line patch to the top-level
configure, followed by the simple (but admittedly tedious) exercise of
testing the patch on a variety of platforms.  I think that someone
with access to lots of machines could spend an hour kicking off all
those tests and then go to sleep; the next day everything would be OK
(since by hypothesis this "just works"), and we could check in the
change, having greatly reduced the risk of inconveniencing people.

Thanks,

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com