checking whether to use MMX intrinsics... disabled checking whether to use SSE2 intrinsics... disabled checking whether to use VMX/Altivec intrinsics... disabled checking whether to use ARM SIMD assembler... yes checking whether to use ARM NEON assembler... yes checking whether to use GNU-style inline assembler... yes .... mv -f .deps/pixman-matrix.Tpo .deps/pixman-matrix.Plo armv4tl-softfloat-linux-gnueabi-gcc -DHAVE_CONFIG_H -I. -I.. -Os -march=armv4t -mtune=arm920t -pipe -MT pixman-arm-neon-asm.lo -MD -MP -MF .deps/pixman-arm-neon-asm.Tpo -c pixman-arm-neon-asm.S -fPIC -DPIC -o .libs/pixman-arm-neon-asm.o armv4tl-softfloat-linux-gnueabi-ar cru .libs/libpixman-arm-simd.a .libs/libpixman_arm_simd_la-pixman-arm-simd.o pixman-arm-neon-asm.h: Assembler messages: pixman-arm-neon-asm.h:565: Error: selected processor does not support `pld [PF_SRC,PF_X,lsl#src_bpp_shift]' pixman-arm-neon-asm.h:565: Error: selected processor does not support `pld [PF_DST,PF_X,lsl#dst_bpp_shift]' Reproducible: Always
configure:21747: result: yes configure:21767: checking whether to use ARM NEON assembler configure:21788: armv4tl-softfloat-linux-gnueabi-gcc -c -x assembler-with-cpp conftest.c >&5 configure:21795: $? = 0
Thanks. It really a valid bug and pixman fails to build when compiling code for armv4 arch. It can be fixed using the following patch from upstream git: http://cgit.freedesktop.org/pixman/commit/?id=a47b5167c4c1b55b2f51e29ab0782c2659bec312
(In reply to comment #1) > configure:21747: result: yes > configure:21767: checking whether to use ARM NEON assembler > configure:21788: armv4tl-softfloat-linux-gnueabi-gcc -c -x assembler-with-cpp > conftest.c >&5 > configure:21795: $? = 0 This is a check which verifies whether your *toolchain* is able to compile NEON code. The real decision whether to run it on real CPU is made at runtime.
Created attachment 214181 [details, diff] patch fixed neon and simd checks This is how it should looks like, but I haven't tested on arm's with neon support, but it should be fine. config.log: configure:11403: checking whether to use ARM SIMD assembler configure:11413: armv4tl-softfloat-linux-gnueabi-gcc -c -Os -march=armv4t -mtune=arm920t -pipe -Wall -fno-strict-aliasing -fvisibility=hidden conftest.c >&5 {standard input}: Assembler messages: {standard input}:24: Error: selected processor does not support `uqadd8 r1,r1,r2' configure:11438: result: no configure:11456: checking whether to use ARM NEON assembler configure:11470: armv4tl-softfloat-linux-gnueabi-gcc -c -Os -march=armv4t -mtune=arm920t -pipe -Wall -fno-strict-aliasing -fvisibility=hidden conftest.c >&5 {standard input}: Assembler messages: {standard input}:24: Error: selected processor does not support `pld r0' {standard input}:27: Error: bad instruction `vmovn.u16 d0,q0' configure.log: checking whether to use ARM SIMD assembler... no checking whether to use ARM NEON assembler... no Some more info: 1. plz respect user's cflags, users know what they want 2. assembler-with-cpp looks like useless for testing, all should be done in C asm
Created attachment 214182 [details, diff] ebuild patch
(In reply to comment #2) > Thanks. It really a valid bug and pixman fails to build when compiling code for > armv4 arch. It can be fixed using the following patch from upstream git: > http://cgit.freedesktop.org/pixman/commit/?id=a47b5167c4c1b55b2f51e29ab0782c2659bec312 > with my configure.ac patch, that commit is unneed!, (also I think that this commit was wrong way to trying to fix armv4 compilation issue)
"Fixed" with the stabilization of binutils-2.19.1.
(In reply to comment #7) > "Fixed" with the stabilization of binutils-2.19.1. > not that bug fixed with binutils-2.19.1, this one is about another issue
I just did tests with bintuils-2.19.1-r1 - armv5tel works for me, but when I compile nativly on armv5tel for armv4tl on armv4tl chroot with armv4tl-softfloat-linux-gnueabi, I still have this issue.
Yes, binutils update only fixes bug 297801 See http://bugs.gentoo.org/show_bug.cgi?id=297794#c12
I've applied the patch from upstream, and i've tested it on armv4tl. Closing
