Re: javax.crypto

[Marco Trudel <mtrudel at gmx dot ch>]

Casey Marshall wrote:
> Marco Trudel wrote:
> > Casey Marshall wrote:
>
> > > However, even if the dynamic loading fails, we initialize our providers
> > > with our default set. You can confirm that these providers are loaded by
> > > calling Security.getProviders, and listing them.
> > Mohan already pointed out how to solve this for a static gcj (see his
> > mail in the thread). In case you're still interested, here the providers:
> >
> > Withhout Mohans static initializer:
> >
> > gnu.java.security.provider.Gnu: name=GNU version=1.0
> >
> > With:
> > gnu.java.security.provider.Gnu: name=GNU version=1.0
> > gnu.javax.crypto.jce.GnuCrypto: name=GNU-CRYPTO version=2.1
> > gnu.javax.crypto.jce.GnuSasl: name=GNU-SASL version=2.1
> > gnu.javax.net.ssl.provider.Jessie: name=Jessie version=1.0
> > gnu.javax.security.auth.callback.GnuCallbacks: name=GNU-CALLBACKS
> > version=2.1

I tested further. Using Mohans initializer 
(http://gcc.gnu.org/ml/java/2006-09/msg00052.html) on linux, I get the 
providers mentioned above and it works.
But using the same java code on mingw fails. I only get the default 
provider, it doesn't matter if I use the initializer or not.
I use a just compiled GCJ and the one for linux is slightly older. So I 
guess it should work...

Any ideas? Workarounds?


Marco


> OK. It seems like this should still work without that initializer,
> though. I do see that the Gnu provider does get loaded, which means that
> the fallback mechanism isn't used. But if the GnuCrypto, et al.,
> providers are referenced in the Security class, shouldn't they be linked
> in? And, all those providers reference all their algorithms statically,
> so they should all get pulled in.
>
> That is, unless I'm misunderstanding how statically linking libgcj works.
>
> > Actually I was only prototyping to check if things work at all. I had an old DES method somewhere and used it... I intend to use AES and updated my prototyp  :-)
>
> OK, but it does look like Classpath/libgcj is missing some algorithms
> (or possibly just aliases) that are supported by Sun.