Return-Path: Date: Mon, 23 Jan 2012 13:54:56 +0200 From: Johan Hedberg To: Siarhei Siamashka Cc: Bastien Nocera , linux-bluetooth@vger.kernel.org Subject: Re: 4.98 fails to build with newer gcc Message-ID: <20120123115456.GA19251@x220> References: <1326460244.2112.48.camel@novo.hadess.net> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii In-Reply-To: Sender: linux-bluetooth-owner@vger.kernel.org List-ID: Hi, On Fri, Jan 13, 2012, Siarhei Siamashka wrote: > On Fri, Jan 13, 2012 at 3:10 PM, Bastien Nocera wrote: > > A few warnings: > > > > sbc/sbc.c:766:34: warning: always_inline function might not be inlinable [-Wattributes] > > sbc/sbc.c:558:34: warning: always_inline function might not be inlinable [-Wattributes] > > sbc/sbc.c:163:31: warning: always_inline function might not be inlinable [-Wattributes] > > Looks like the new version of gcc requires both "inline" and > "__attribute__((always_inline))" to be used in SBC_ALWAYS_INLINE > according to [1]. > > > and an error > > > > sbc/sbc_primitives_mmx.c: In function 'sbc_calc_scalefactors_mmx': > > sbc/sbc_primitives_mmx.c:294:4: warning: asm operand 2 probably doesn't match constraints [enabled by default] > > sbc/sbc_primitives_mmx.c:294:4: error: impossible constraint in 'asm' > > Here gcc 4.7 thinks that > "i" ((char *) &sb_sample_f[1][0][0] - > (char *) &sb_sample_f[0][0][0]), > is an impossible constraint. According to [2], "i" means "An immediate > integer operand (one with constant value) is allowed. This includes > symbolic constants whose values will be known only at assembly time or > later". > Earlier versions of gcc could see that this expression is a compile > time constant, but gcc 4.7 can't. As a workaround, "i" can be changed > to "r" here. The downside is that this needs one more register for > inline assembly and the number of available registers is really > limited for 32-bit x86 systems. As a test, it is possible to try "gcc > -O2 -mmmx -c -m32 -fno-omit-frame-pointer -fPIC sbc_primitives_mmx.c". > Both "-fno-omit-frame-pointer" and "-fPIC" options reduce the number > of available registers. So is someone going to send a patch for this? I think we should try to have it fixed before doing the next BlueZ release. Johan