Release stuff

Tue Mar 12 13:31:00 GMT 2002

Hi,
On Tue, 2002年03月12日 at 19:30, Tom Tromey wrote:
> I'm trying to track our state so we can try to have meaningful input
> into the 3.1 release process.

> 5912 gcj -R doesn't work for mingw32 cross compiler
> I think we have a patch for this.
> We want mingw32 in this release
Patch: <http://gcc.gnu.org/ml/gcc-patches/2002-03/msg00545.html>
Approved: <http://gcc.gnu.org/ml/gcc-patches/2002-03/msg00577.html>
> 5923 gcj -C generates incorrect bytecode
> This is probably a regression, though I'm not certain
It is not a regression, it also happens with 3.0.4. But it only happens
when the String constant is read from a class file and then written to a
different class file. Compiling everything from source or compiling to
native code does not show it. It always seems to only write the first
character of the String constant to the class file.
> If you think there is a missing PR, please tell me. If there is a
> critical bug that isn't in a PR, point me to the bug report and I will
> file a PR.

4715 Tree check failure in check_inner_circular_reference
 Makes it impossible to compile the latest GNU Classpath 0.03 release.
 gcj 3.0.4 can compile it. Your patch in the PR seems to work for me.
Do we have a set of real world programs/libraries that can be used as
test cases? Making sure that Jacks, Mauve, Classpath and Kawa
compile/work seems like a good idea.
I can setup an autobuilder/tester on a Debian GNU/Linux PowerPC box.
Cheers,
Mark