On 28 Jun 2000, Brian Jones wrote:
> Tom Tromey <[EMAIL PROTECTED]> writes:
>
> > Brian> What's very confusing here is libgcj's implementation has a
> > Brian> classpath header but does not appear to be exactly the
> > Brian> same... so I guess these aren't merged yet. Is anyone working
> > Brian> on that?
> >
> > Warren has been working on serialization lately. He told me he is
> > going to check in the changes to Classpath soon. I don't know when.
> > We're definitely not going to allow divergence in code that we share.
Yes, Tom's absolutely right here. I have all my patches sitting here in
my classpath tree waiting to be checked in. Thanks to Jeff Bailey
<[EMAIL PROTECTED]>, I should now have write access into the tree (FWIW,
the Classpath folks were extremely prompt about getting me access, I was
just extremely slow in sending them my identity.pub for the setup).
I'm trying to get some other serialization stuff straightened out at the
moment and then I'll see if I can actually write to the tree (I gather
from the Gnu html pages that I need to run a script to massage my CVS
files from 'anoncvs' to 'warrenl' on subversions; I probably also have to
switch from 'pserver' to something else like 'ext', I'm guessing I'll have
to play around with this one to get it right. (BTW, I don't really want
to check out a whole other tree since my mods are already in place in the
current one)).
> Thanks for the info, I wanted to send a small patch concerning this
> class against libgcj's implementation but for some reason anoncvs is
> broken this morning (can't update or do a diff), and I already removed
> the original... Anyway, in ObjectStreamClass.lookup (Class) it should
> look something like this with the check for serialization moved to
> somewhere more appropriate to match the comments and Sun's own docs.
Thanks, I'll try it out. BTW, I have one other change in the libgcj
java.io.ObjectStreamClass that I wasn't yet ready to put in Classpath
(I wasn't sure if it was necessary there). It pulls in the SHA algorithm
available at gnu.java.security.provider in Classpath by doing the
following in the setUID method:
try
{
md = MessageDigest.getInstance ("SHA");
}
catch (NoSuchAlgorithmException e)
{
// If a provider already provides SHA, use it; otherwise, use this.
Gnu gnuProvider = new Gnu();
Security.addProvider(gnuProvider);
md = MessageDigest.getInstance ("SHA");
}
This allows me to get the serialVersionUID right (well, mostly right from
what I'm seeing) in libgcj. If folks aren't getting the right values in
Classpath, then I probably need to do it there too. Let me know if you
agree.
> I'm eagerly awaiting the merge. :)
Me too. ;-)
--warrenl
