>While that may work, it is technically wrong.

Yeah, I konw, so was initializing it to zero, which is what the old
code did.

 > I want to fold the chang back to FSF/Cygnus. A comment says the coded
 > fragment is(was) also used on OSF/1 on pmaxes, and i have no clue what
 > OSF/1 does with sigemtpyset().

>So you need HAVE_SIGEMPTYSET...

Oh, foo.... I dont even know if the autoconfig in our dist/gcc tests
for that.  

But if you know how to fix this in a way that's OK for our tree, and
that works gcc 2.x on ancient OSF/1 mips boxes, that the FSF will buy
back, and you know where to turn on the HAVE_SIGSEMPTYSET in our tree
(we dont seem to have a nice config.h, as for newer GNU imports) and
you know what to do if HAVE_SIGEMPTYSET isn't on, and you still care,
please go ahead and fix it, or tell me where to ...

All I want is to get gcc 2.7.2 built one last time so I can get a
snapshot out, and get on with moving to egcs.