Subject: Proper way to do C++ development under -current
To: None <>
From: Rick Byers <>
List: current-users
Date: 10/30/2000 23:18:40
I'm trying to do some C++ development on my NetBSD-current box.  I keep
getting bitten by what appear to be bugs in egcs-1.1.2.  I tried using
the gcc in pkgsrc, but it apparently requires gas 2.10 (otherwise
exceptions don't work due to a known bug in gas 2.9.1).  What is the
proper way to do real C++ development in NetBSD?  

I know gcc-2.95.2 and buntils-2.10 are going to be in -current soon, but
what are people doing up until then?  I've been able to work around most
of the problems I've had with egcs-1.1.2, but its been a real hassle.  
Right now I'm stumped by the compiler's inability to determine the type of
a thrown exception (the first "catch" line allways catches it, regardless
of type), and I can't figure out a way around it (short of restructuring
all my exceptions).

So I'm trying to install gcc-2.95.2/binutils-2.10 using the new
dist/toolchain code.  I had some problems getting gdb to build (it doesn't
seem to understand that my i386 box is ELF as it's doing stuff to
compensate for link_aout.h).  I think I have it working now (knock on
wood - still needs to be tested), but I'm curious as to what others are 
doing.  Does gnu toolchain stuff normally build properly out of the
box, or are modifications generally required?

Maybe we should consider making a binutils package?  If thats the right
thing to do, I'm willing to try to spend some time on doing that.  Having
an up-to-date (maybe even -current?) toolchain pacakge might be a good
idea anyway.  Once I get my program working properly on -current, I need
to run it on a NetBSD-1.4.2 machine (which I'll probably update to
1.5_BETA).  I have a feeling thats going to be another big challenge as
I'd prefer not to have to install a whole new toolchain set there also.


P.S. The "minor toolchain updates to make mozilla build properly" wouldn't
have any impact on my problems would they?  Are we making an attempt to
patch bugs in egcs-1.1.2 (seems sort of futile).

Rick Byers
University of Waterloo, Computer Science