On 6/7/2017 1:14 PM, Gavin Smith wrote:
On Tue, Jun 06, 2017 at 09:56:37PM -0400, Ken Brown wrote:
I've committed a change to do this. I copied the test from the
top-level configure.ac to see if we are on MS-Windows:
host_is_windows=no
case "$host" in
*-mingw32 | *-mingw64 | *-msdosdjgpp ) host_is_windows=yes ;;
esac
AM_CONDITIONAL([HOST_IS_WINDOWS], [test "x$host_is_windows" = "xyes"])
Please someone say if this is wrong for a system where the -no-undefined
flag is needed (e.g. cygwin).
It's wrong on Cygwin. Without -no-undefined -lperl I get errors like this:
/bin/sh ./libtool --tag=CC --mode=link gcc -DPERL_USE_SAFE_PUTENV
-D_GNU_SOURCE -U__STRICT_ANSI__ -ggdb -O2 -pipe
-Wimplicit-function-declaration -fwrapv -fno-strict-aliasing
-fstack-protector-strong -D_FORTIFY_SOURCE=2 -DVERSION=\"0\"
-DXS_VERSION=\"1\" "-I/usr/lib/perl5/5.24/x86_64-cygwin-threads/CORE"
-avoid-version -module -o XSParagraph.la -rpath /usr/lib/texinfo
XSParagraph_la-XSParagraph.lo XSParagraph_la-xspara.lo
XSParagraph_la-text.lo ./lib/libgnu.la
libtool: error: can't build x86_64-unknown-cygwin shared library unless
-no-undefined is specified
So I guess it should be
host_is_windows=no
case "$host" in
*-mingw32 | *-mingw64 | *-msdosdjgpp | *-cygwin ) host_is_windows=yes ;;
esac
AM_CONDITIONAL([HOST_IS_WINDOWS], [test "x$host_is_windows" = "xyes"])
Yes, that would do it. My only suggestion is that it might be better to
use something like HOST_NEEDS_NO_UNDEFINED rather than HOST_IS_WINDOWS.
Having HOST_IS_WINDOWS defined on Cygwin could cause confusion in the
future if someone wants to introduce Windows-specific code, since such
code usually isn't appropriate for Cygwin. The need for -no-undefined
is an exception.
Ken
