Re: bug in gcj, maybe with multiple .class inputs

[Per Bothner <per at bothner dot com>]

Bryce McKinlay <bryce@albatross.co.nz> writes:

> We really need to disable multi-file on the branch - there are too many
> problems with it, affecting both source and class compilation.

This is news to me.  The regression testsuite passes, and Kawa (which uses
the multi-file thing heavility) passes its large testsuite.  I really
do not want to disable multi-file compilation for compiling to native
because I've got my Kawa Makefiles set up to depend on it.  It doesn't
seem like that should break anything that does actually use multi-file
compilation.  Multi-file compilation to class files is less critical,
since that doesn't affect Makefiles.

> Per posted a patch to do this, but its slightly buggy.
> I've been intending to check it in once I get around to fixing it.

I don't know which patch you're talking about.  All my (posted
non-obsoleted) patches are checked in (to both the trunk and the
branch).  If you have a testcase that you think breaks due to
multi-fle compilation, let me know and I'll take a look at it.
-- 
	--Per Bothner
per@bothner.com   http://www.bothner.com/~per/