Unfortunately the (reverted) first cut cause had problems with kernel builds - setting CFLAGS at the top-level can cause problems. As a result, I've pushed the more aggressive plan B: - introduces USERLAND_CFLAGS make variable (so, hopefully, it is clear that the flags are not intended for the kernel) - adds mk/userland-cflags.mk file to construct USERLAND_CFLAGS value - moved most of the CFLAGS+=-D options to that file The old USERCOMPILE barely survived (greatly stripped down, there are separate DEBUG_CFLAGS and OPTIMIZE_CFLAGS for instance). The old WERROR flag didn't.
It didn't touch -L, -I and -l flags as they get into ordering problems. Leave them for another day. Andrew On 11 March 2015 at 11:13, Andrew Cagney <[email protected]> wrote: > I've pushed this. > > On 6 March 2015 at 12:21, Andrew Cagney <[email protected]> wrote: > >> Hi, >> >> This patch: >> >> commit f12e856de21a3dd1ca3d49946c3902 >> a04cf9ec70 >> Author: Paul Wouters <[email protected]> >> Date: Fri Mar 6 10:06:52 2015 -0500 >> >> FIPS: Do not allow PSK in FIPS mode (submitted patch badly merged in) >> >> This was merged in wrong years ago in upstream openswan, so only >> RHEL6 versions of openswan properly enforced this. >> >> which is about lines like: >> >> ifeq ($(USE_FIPSCHECK),true) >> CFLAGS+=-DFIPS_CHECK >> endif >> >> littered through the Makefiles prompts the question. Should we simply >> move all of these to mk/config.mk? >> >> Baring objections, I'll build/test/push this Monday. As they say, this >> should be safe ... >> >> Andrew >> >> PS: It doesn't include things like this: >> >> ifeq ($(USE_LINUX_AUDIT),true) >> LIBSPLUTO += -laudit >> endif >> >> or this: >> >> ifeq ($(USE_NETKEY),true) >> # Linux always supports udpfromto >> UDPFROMTO_SRCS=udpfromto.c >> endif >> >> >
_______________________________________________ Swan-dev mailing list [email protected] https://lists.libreswan.org/mailman/listinfo/swan-dev
