"init_dummy" bug apparently fixed in 1.0.1 - 77% faster now!

[Benjamin Redelings I]

I had a program that used to call init_dummy whenever it felt like
it... or at least that's what gprof would lead me to believe.  It gave
init_dummy 70% of the CPU time!
	I checked egcs 1.0.1 (compiled with --enable-shared), and init_dummy
appears nowhere :)  Also, (probably because of this) version 1.0.1 is
77% faster than snapshot 12-25-97 (-fno-exceptions -O6).
	I'd give the speedup over 2.7.2.3, but that old compiler can't compile
my code any more :)  Hmmmm.......

Thanks for all the work,
-BenRI