On Sat, 2006-07-29 at 17:52 +0200, Mark Wielaard wrote: > > And the mauve test that should now pass with the patch is > > java.security.Engine.getInstance testAlgorithmCase > > It did pass before your patch (because the canonicalization was done in > the Provider first) but no regressions, and this way is indeed nicer.
However, it does cause a regression with some other mauve test: FAIL: gnu.java.security.jce.TestOfProvider line 150: getInstance(SHA-256PRNG): java.security.NoSuchAlgorithmException: too many aliases [1] -- forced fail This seems to be caused because your patch doesn't seem to handle aliases pointing to themselves. And the Gnu provider does that. I am not sure why the provider puts in aliases to itself, but the old code handled this case. Could you take a look? Thanks, Mark
