Subject: Re: Trouble compiling libg++-2.6.2
To: Douglas Thomas Crosher <dtc@scrooge.ee.swin.oz.au>
From: Rob Healey <rhealey@kas.helios.mn.org>
List: netbsd-help
Date: 03/01/1995 07:23:30
> I have been trying to compile libg++-2.6.2 on a NetBSD-1.0 i486
> machine, but I get an error - see below - when compiling.
> 
> I have installed gcc-2.6.3, this seems to be OK, although it could be
> the cause of the problem; my understanding of c++ is to limited to say.
> 
> libg++ does compiles on a Sun OS 5.3 sparc with gcc-2.6.3; this makes
> me suspect something related to my installation of NetBSD.
> 
> Does anyone have any clues?
> 
[ Deleted output showing failed compile of Fix.cc ]

Someone sent a fix to current-users a while back when I tried to compile
libg++ on NetBSD/m68k. The fixes allowed Fix.cc to compile but it still
failed all the tests. Luckily my application doesn't need to use the Fix
class. Everything else in lib++ passed the tests. Since x86 NetBSD systems
also appear to have at least the compile problem maybe this libg++ bug
is specific to NetBSD? Have SPARC or any other port's been able to get
gcc 2.6.3 + libg++-2.6.2 to compile and run the regression tests successfully?

The tests looked like they failed on NetBSD/m68k due to sign extention
weirdness in 33 bit and 17 bit quantitys in the Fix class tests. The
two other related classes, Fix24 and Fix16(?), passed their tests fine
although their source .cc files didn't need patching like Fix.cc did.

All this was using the 2.6.3 compiler too.

	-Rob