Newsgroups: gnu.g++.bug
Path: sparky!uunet!cis.ohio-state.edu!ukcsd.UK.SUn.COM!jeremy
From: jer...@ukcsd.UK.SUn.COM (Jeremy Allison - Sun UK - Tech Support)
Subject: gcc-2.0 bug
Message-ID: <kqkdteINNhfm@cronkite>
Sender: gnuli...@ai.mit.edu
Reply-To: Jeremy.Alli...@uk.sun.com
Organization: Sun Microsystems UK Answer Centre
Distribution: gnu
Date: Tue, 25 Feb 1992 12:33:50 GMT
Approved: bug-...@prep.ai.mit.edu
Lines: 71
Firstly I'd like to say a _BIG_ thanks to GNU & FSF for writing and
publishing this code - it's the only way someone with a sun386i can
get up to date software tools.
I have discovered a problem with the c++ part of gcc-2.0, whilst
porting the libg++ 1.39.0 library to gcc-2.0.
Firstly I removed all the #pragma once's from the library header
code, then patched up the definitions of memcpy and strlen in
the header files to match the internal ones (so I stopped getting
the error messages).
The src make went fine - producing a libg++.a, and indeed, if I link
against this with a "Hello World" program everything is fine, however,
on make'ing the test directory I find that linking stops with some
undefined references - in particular the tRandom test.
Here is the error message when making this explicitly.
g++ tRandom.o -o tRandom -Wall -v -L../src -lg++ -lm
Reading specs from /usr/local/lib/gcc/sun386i/2.0/specs
gcc version 2.0
/usr/local/lib/gcc/sun386i/2.0/ld -e _start -dc -dp -o tRandom /lib/crt0.o \
-L../src -L/usr/local/lib/gcc/sun386i/2.0/ -L/usr/local/lib/gcc/ \
-L/usr/local/lib/ tRandom.o -lg++ -lm -lg++ -lgcc -lc -lgcc
ld: Undefined symbol
_3RNG$singleMantissa
_3RNG$doubleMantissa
collect: /usr/bin/ld returned 1 exit status
*** Error code 1
make: Fatal error: Command failed for target `tRandom'
On investigation I have found this to be caused by gcc2.0 failing to
output definitions for static member data declatations in a class. The
other link failures I have are also due to this - looking at the
assembler files produced show references to these static member data
items in the code but no declarations for them.
The .cc file in question in the libg++ source directory is
RNG.cc.
The #pragma settings are
#pragma implementation
in the RNG.cc file, and
#pragma interface
in the RNG.h file.
The output when make'ing RNG.o is :-
make RNG.o
g++ -I../g++-include -O -fstrength-reduce -felide-constructors -fsave-memoized -Wall -v -c RNG.cc
Reading specs from /usr/local/lib/gcc/sun386i/2.0/specs
gcc version 2.0
/usr/local/lib/gcc/sun386i/2.0/cpp -lang-c++ -v -I../g++-include -undef -D__GNUC__=2 -D__GNUG__=2 -D__cplusplus -Dunix -Di386 -Dsun386 -Dsun -D__unix__ -D__i386__ -D__sun386__ -D__sun__ -D__unix -D__i386 -D__sun386 -D__sun -D__OPTIMIZE__ -Wall RNG.cc /usr/tmp/cca19237.cpp
GNU CPP version 2.0 (80386, Sun syntax)
/usr/local/lib/gcc/sun386i/2.0/cc1plus /usr/tmp/cca19237.cpp -quiet -dumpbase RNG.cc -O -Wall -version -fstrength-reduce -felide-constructors -fsave-memoized -o /usr/tmp/cca19237.s
GNU C++ version 2.0 (80386, Sun syntax) compiled by GNU C version 2.0.
RNG.cc: In method `RNG::RNG ()':
RNG.cc:24: warning: statement with no effect
as -i386 -o RNG.o /usr/tmp/cca19237.s
Hopefully this is already known about (and not just on 386i's :-).
Thanks once again for this software.
Jeremy Allison.
Disclaimer: Sun don't speak for me - I don't speak for Sun.
Newsgroups: gnu.gdb.bug
Path: sparky!uunet!zaphod.mps.ohio-state.edu!cis.ohio-state.edu!uk.SUn.COM!Jeremy.Allison
From: Jeremy.Alli...@uk.SUn.COM (Jeremy Allison - Sun UK - Tech Support)
Subject: gdb4.4 on 386i
Message-ID: <9202271243.AA07835@whippet.uk.sun.com>
Sender: gnuli...@ai.mit.edu
Organization: GNUs Not Usenet
Distribution: gnu
Date: Thu, 27 Feb 1992 12:43:54 GMT
Approved: bug-...@prep.ai.mit.edu
Lines: 25
Firstly - thanks to GNU & FSF for producing this software.
They make many things possible for those of us with older machines.
I have a sun386i at home and am trying to compile up the latest version
of gdb - 4.4 for it.
There is no configure file for this machine - however, on reading the gcc
lists I found the following information given in a patch file to
allow gcc2.0 to be installed on a sun386i.
+ #ifdef NO_STATIC_DEBUG
+ /*
+ * now that GDB-4 with Peter Schauer's patches can handle
+ * shared libs on the sun386, we can skip %{g:-Bstatic}
+ * by defining NO_STATIC_DEBUG
+ */
I have mailed the poster direct, but for completeness, does anyone know where
these patches (Peter Schauer's) may be found ?
If anyone can let me know I'd be very grateful, and will pass the
info back to this list if I get an answer.
Thanks,
Jeremy Allison.