recent toplev.c changes break eh

[Per Bothner]

> BTW, it would have been helpful if a) you had provided a test-case and
> b) that test-case had been in some test-suite.  I ran the tests before
> check-in, and did not notice any failures.  I would have expected
> making the Java front-end unusable would have resulted in some failure
> somewhere.  (Did I miss the test failure?  That's possible, too.)

Well, the first "testsuite" is to build libgcj.  I was unable to compile
java.io.PrintStream before your patch.

It is difficult to make stand-alone Java tests, since anything will
depend on java.lang.Object - and it turns out Object depends on quite
a few other classes.

Note I'm not criticising you - it is difficult to test all the
front-ends and run all the testsuites.  Often it may just make more
sense to do whatever testing is relatively easy, and hope for the best
- and that someone else will quickly notice other problems - as in
this case.  I am just surprised this did not come up when testing
C++ ...

> Please verify this,

It seems to fix the problem - at least now libgcj compiles without
errors.

> and check in the patch,

OK will, do.

	--Per Bothner
Cygnus Solutions     bothner@cygnus.com     http://www.cygnus.com/~bothner