🚀 go-pugleaf

A number of people (myself including) have gotten gcc and g++
running on DECstations. There are couple of bugs, at least
with g++. One is that a non-inlined methods (virtual or not)
cannot return structures. (This is because gcc wants to use
the first parameter register for the address of the result,
while g++ wants to use it for the 'this' pointer. What should
happen is that the first shold be used for the result address,
and the second for the 'this' pointer, but (the machine-
independent part of) gcc isn't really set up to handle that for
architectures that pass parameters in registers. This may be
fixed in gcc-2.0.) Another problem I experienced
was that "i += f(x)" should sometimes fail to update 'i'.
I've managed to kludge around around these problems, and use
gcc and g++ productively on a near-daily basis.

Another problem is that debugging isn't supported very well.
Michael Meissner (who did most of the mips port) has a kludge for
emitting symbol table support, but it won't work for g++ (because
of the "inventive" mips symbol table format). Peter Ham
has a version of g++ that emits debuggable code, but I
haven't tried it, and it seems to be based on old versions
of the software. I believe there are people actively working
on the solving the debuggability problem(s).

I have g++ version 1.37.2 alpha 2 and gcc version 1.37.91.
I suspect most of the problems will be solved when gdb-4.0
and gcc-2.0 come out, or soon thereafter - in a month or four...
-- 
	--Per Bothner
bothner@cs.wisc.edu Computer Sciences Dept, U. of Wisconsin-Madison