Re: patch: gcc 2.96 compile fixes, file creation mode, ServerSocket bind cleanup

[Tom Tromey <tromey@cygnus.com>]

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

Bryce> 2. The mode of new files created by java.io.File is wrong
Bryce> (should be 644 not 755 - verified against the JDK). The mode of
Bryce> directories is correct so removed the fixme comment.

Thanks.  Obviously I still can't do octal -- what was I thinking with
755?

Bryce> I'll commit this tommorrow if there are no objections.

It would be easier if the patch were actually a few smaller chunks:
cast cleanups, the mode fix, and the net fixes.  I try to only commit
related things in a single commit (sometimes I fail, but still it's
worth a shot).  This makes it easier to revert individual fixes.
Plus, on a different level, I'm not really competent to comment on the
java.net changes.

Bryce> +  GC_REGISTER_FINALIZER_NO_ORDER (object, call_finalizer, (GC_PTR)meth,

A nasty little nit: there should be a space after the close paren:

	(GC_PTR) meth

Bryce> +  ehinfo->eh_info.match_function = (__eh_matcher)_Jv_type_matcher;

Ditto.

Bryce> -   return new java::lang::Boolean (false);
Bryce> +   return new java::lang::Boolean ( (__java_boolean)false );

Use "jboolean", and fix the spacing thusly:

	((jboolean) false)

(Another silly nit.  And though we aren't totally consistent in
libgcj, we do try...)

Actually, I find this change a bit distressing.  Is this due to a C++
compiler change?  I think for seamless C++/Java integration to really
be seamless, C++ boolean constants and Java boolean constants should
be identical.  That is, I think we shouldn't need a cast here.

So for this little piece (the boolean cast) I'd rather get more
information before checking it in.

Tom