NetBSD-Bugs archive
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]
port-arm/52528: CPUFLAGS=-mfpu=neon-vfpv4 breaks some programs
>Number: 52528
>Category: port-arm
>Synopsis: CPUFLAGS=-mfpu=neon-vfpv4 breaks some programs
>Confidential: no
>Severity: non-critical
>Priority: low
>Responsible: port-arm-maintainer
>State: open
>Class: sw-bug
>Submitter-Id: net
>Arrival-Date: Tue Sep 05 08:55:00 +0000 2017
>Originator: Manuel Bouyer
>Release: NetBSD 8.0_BETA
>Organization:
>Environment:
System: NetBSD chartplotter 8.0_BETA NetBSD 8.0_BETA (CUBIEBOARD) #0: Fri Sep 1 13:15:30 CEST 2017 bouyer%bop.soc.lip6.fr@localhost:/dsk/l1/misc/bouyer/tmp/earmv7hf/obj/dsk/l1/misc/bouyer/netbsd-8/src/sys/arch/evbarm/compile/CUBIEBOARD evbarm
Architecture: earmv7hf
Machine: evbarm
>Description:
building a release with
./build.sh -m evbarm -a earmv7hf -V CPUFLAGS="-mfpu=neon-vfpv4"
(call this a earmv7hf-neonvfpv4 release) cause some programs to
fail on the target system.
Here is an example:
cat > /tmp/conftest.cpp << EOF
// (C) Copyright John Maddock 2001.
// Use, modification and distribution are subject to the
// Boost Software License, Version 1.0. (See accompanying file
// LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
// See http://www.boost.org/libs/config for most recent version.
// MACRO: BOOST_NO_EXCEPTIONS
// TITLE: exception handling support
// DESCRIPTION: The compiler in its current translation mode supports
// exception handling.
namespace boost_no_exceptions{
void throw_it(int i)
{
throw i;
}
int test()
{
try
{
throw_it(2);
}
catch(int i)
{
return (i == 2) ? 0 : -1;
}
catch(...)
{
return -1;
}
return -1;
}
}
int main(){ return boost_no_exceptions::test(); }
EOF
c++ -o conftest conftest.c
It doens't matter if the program is built on earmv7hf-neonvfpv4 or
plain earmv7hf. When run on earmv7hf the program runs as expected,
on earmv7hf-neonvfpv4 it fails with:
terminate called after throwing an instance of 'int'
terminate called recursively
Abort (core dumped)
I tracked it down to libc (running on a plain earmv7hf with the
earmv7hf-neonvfpv4 libc is enough to make it fail), and more
specifically to libunwind.o in libc: on a earmv7hf-neonvfpv4 system,
c++ -static -o conftest conftest.c
gives a non-working program, but
c++ -static -o conftest conftest.c libunwind.o
gives a working one, with libunwind.o extracted from a plain
earmv7hf libc.a
>How-To-Repeat:
see above.
>Fix:
unkown.
Home |
Main Index |
Thread Index |
Old Index