From mboxrd@z Thu Jan 1 00:00:00 1970 From: Thomas Petazzoni Date: Mon, 17 Jul 2017 14:11:11 +0200 Subject: [Buildroot] [PATCH v6 1/4] infra: add support for MIPS NaN In-Reply-To: References: <20170628151713.10746-1-Vincent.Riera@imgtec.com> <20170716164000.415ca6f3@windsurf> <20170717113412.6306d69a@windsurf> Message-ID: <20170717141111.7c200c1d@windsurf> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: buildroot@busybox.net Hello, On Mon, 17 Jul 2017 13:03:05 +0100, Vicente Olivert Riera wrote: > > Is it because the gcc of this external toolchain is too old ? > > Yes, the -mnan= option was introduced in gcc 4.9.0 by this commit: > > 0bd32132d471995f3e333138363d1d928a9dde60 > > The gcc of that external toolchain is 4.8.2. OK, that explains it. > > What can we do about this ? > > I guess we could use an option called BR2_GCC_SUPPORTS_MNAN in order to > avoid adding the -mnan option to the CFLAGS and to the tc-wrapper if the > external toolchain hasn't selected that option. For internal toolchain, > since the oldest gcc version we support is 4.9.4, then it should be > always selected. Sounds good to me. I guess those older gcc versions (4.8 and before) anyway only supported the legacy NaN encoding, and therefore behave as if -mnan=legacy is passed, correct? You might also want to check if the same problem will happen (or not) for the FP32 mode, and the DSP/MSA options :) Thanks! Thomas -- Thomas Petazzoni, CTO, Free Electrons Embedded Linux and Kernel engineering http://free-electrons.com