Christopher Faylor wrote: > On Thu, Feb 13, 2003 at 12:43:49PM -0800, neal somos wrote: >> If I make the following change to wv2's configure, I can then make >> the package >> >> 25826c25826 >> < ac_link='$CC -o conftest$ac_exeext $CFLAGS $CPPFLAGS $LDFLAGS >> conftest.$ac_ext $LIBS >&5' >> --- >>> ac_link='$CC -o conftest$ac_exeext $CFLAGS $CPPFLAGS >>> conftest.$ac_ext $LDFLAGS $LIBS >&5' >> >> Basically ... >> gcc -o conftest.exe -O2 -liconv conftest.c >> produces quite different results from >> gcc -o conftest.exe -O2 conftest.c -liconv >> which are the compile lines from the test being run by configure. >> >> How the original test works elsewhere but fails on cygwin >> is beyond me at this time. > > It's probably expecting that it can do something put > > extern int libiconv_close(); > > in a source file and then just link it. That doesn't necessarily work > on Windows. You may have to mark the declaration with > declspec(dllimport);. > > The function should be auto-imported automatically but maybe there is > a warning showing up or something which is confusing autoconf.
No, I looked at wv2's configure.in. They are wrongly putting -liconv into $LDFLAGS, not $LIBS. It's that gcc link order FAQ all over again. Max. -- Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple Bug reporting: http://cygwin.com/bugs.html Documentation: http://cygwin.com/docs.html FAQ: http://cygwin.com/faq/
