Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Tue, 22 Oct 2019 08:09:59 -0700 (PDT)
Subject:     Re: [PATCH 4/8] riscv: ensure RISC-V C model definitions are passed to static analyzers
In-Reply-To: <alpine.DEB.2.21.9999.1910172138320.3026@viisi.sifive.com>
CC:     luc.vanoostenryck@gmail.com, linux-riscv@lists.infradead.org,
        linux-kernel@vger.kernel.org
From:   Palmer Dabbelt <palmer@sifive.com>
To:     Paul Walmsley <paul.walmsley@sifive.com>
Message-ID: <mhng-8e06672c-a3e1-4dde-bbe9-ba1d8ab79645@palmer-si-x1e>
Mime-Version: 1.0 (MHng)
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 8bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Thu, 17 Oct 2019 21:39:29 PDT (-0700), Paul Walmsley wrote:
> On Fri, 18 Oct 2019, Luc Van Oostenryck wrote:
>
>> On Thu, Oct 17, 2019 at 05:49:25PM -0700, Paul Walmsley wrote:
>> > Static analysis tools such as sparse don't set the RISC-V C model
>> > preprocessor directives such as "__riscv_cmodel_medany", set by the C
>> > compilers.  This causes the static analyzers to evaluate different
>> > preprocessor paths than C compilers would.  Fix this by defining the
>> > appropriate C model macros in the static analyzer command lines.
>> >
>> > Signed-off-by: Paul Walmsley <paul.walmsley@sifive.com>
>> > ---
>> >  arch/riscv/Makefile | 2 ++
>> >  1 file changed, 2 insertions(+)
>> >
>> > diff --git a/arch/riscv/Makefile b/arch/riscv/Makefile
>> > index f5e914210245..0247a90bd4d8 100644
>> > --- a/arch/riscv/Makefile
>> > +++ b/arch/riscv/Makefile
>> > @@ -47,9 +47,11 @@ KBUILD_CFLAGS += -DCONFIG_PAGE_OFFSET=$(CONFIG_PAGE_OFFSET)
>> >
>> >  ifeq ($(CONFIG_CMODEL_MEDLOW),y)
>> >  	KBUILD_CFLAGS += -mcmodel=medlow
>> > +	CHECKFLAGS += -D__riscv_cmodel_medlow
>> >  endif
>> >  ifeq ($(CONFIG_CMODEL_MEDANY),y)
>> >  	KBUILD_CFLAGS += -mcmodel=medany
>> > +	CHECKFLAGS += -D__riscv_cmodel_medany
>>
>> I can teach sparse about this in the following days.
>
> That would be great.  Would you be willing to follow up with me via E-mail
> or mailing list post when it's fixed?  If so, then in the meantime, I'll
> just drop this patch.

It's probably worth going through all our argument-dependent builtin 
definitions at the same time.  They're generated by riscv_cpu_cpp_builtins(): 
https://github.com/gcc-mirror/gcc/blob/master/gcc/config/riscv/riscv-c.c#L35 .

>
>
> - Paul
>
> _______________________________________________
> linux-riscv mailing list
> linux-riscv@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-riscv