Hi Tom,
On Tue, Jan 26, 2016 at 10:47 PM, Tom Rini trini@konsulko.com wrote:
On Tue, Jan 26, 2016 at 12:26:07PM +0800, Bin Meng wrote:
Hi Tom,
On Mon, Jan 25, 2016 at 10:12 PM, Tom Rini trini@konsulko.com wrote:
On Mon, Jan 25, 2016 at 11:18:26AM +0800, Bin Meng wrote:
Hi Tom,
On Mon, Jan 25, 2016 at 11:05 AM, Tom Rini trini@konsulko.com wrote:
On Mon, Jan 25, 2016 at 10:34:16AM +0800, Bin Meng wrote:
Hi Tom,
On Mon, Jan 25, 2016 at 12:19 AM, Tom Rini trini@konsulko.com wrote: > On Sun, Jan 24, 2016 at 12:00:42PM +0800, Bin Meng wrote: > >> Hi, >> >> Summary of 71 commits for 1100 boards (24 threads, 1 job per thread) >> 01: iocon / bamboo: Drop CONFIG_SYS_LONGHELP >> blackfin: + bf561-acvilon cm-bf561 blackstamp bf537-minotaur >> bct-brettl2 cm-bf527 bf506f-ezkit ip04 bf527-sdp pr1 bf527-ad7160-eval >> bf609-ezkit bf537-stamp bf527-ezkit-v2 cm-bf537e tcm-bf518 cm-bf537u >> bf527-ezkit cm-bf533 bf533-ezkit ibf-dsp561 bf537-pnav bf537-srv1 >> cm-bf548 bf538f-ezkit bf548-ezkit bf525-ucr2 blackvme tcm-bf537 >> bf533-stamp dnp5370 bf518f-ezbrd bf526-ezbrd bf561-ezkit br4 >> aarch64: + test >> avr32: + atngw100mkii grasshopper atstk1002 atngw100 >> sh: + sh7753evb sh7785lcr_32bit sh7785lcr >> arc: + arcangel4-be axs101 axs103 tb100 arcangel4 >> openrisc: + openrisc-generic >> powerpc: + TQM834x katmai >> arm: + mx28evk mx28evk_nand xfi3 bg0900 sansa_fuze_plus >> mx23evk m28evk sc_sps_1 mx28evk_spi apx4devkit mx23_olinuxino >> mx28evk_auart_console >> nds32: + adp-ag101p > > I need to finally fetch a few toolchains as I don't do > avr32/sh/openrisc/nds32 iirc. As a tangent, x86 is very broken with gcc > 5.x, can you look into it? :) >
Sure, I can look into x86. Which gcc 5.x toolchain are you using?
Pretty much any, I've seen it for a long time but not had time to poke at the libgcc "fun" that's involved here.
I mean if there is any prebuilt gcc 5.x for me to grab and test? Or do I need build one from gcc source?
Fedora has shipped with gcc 5.x for a release or two and Debian/unstable is how I get all of my gcc 5.x toolchains for build testing.
I installed a fresh Fedora 23 and got gcc 5.1.1 up and running. However I see _zero_ compiler warnings/errors using gcc 5.1.1 when building all of the 9 x86 boards.
The only thing I noticed, which is probably best matched to what you said "x86 is very broken with gcc 5.x", is:
HOSTCC tools/aisimage.o In file included from tools/aisimage.c:10:0: include/image.h:923:27: fatal error: openssl/evp.h: No such file or directory compilation terminated. scripts/Makefile.host:111: recipe for target 'tools/aisimage.o' failed make[1]: *** [tools/aisimage.o] Error 1 Makefile:1196: recipe for target 'tools' failed make: *** [tools] Error 2
Are you using the stock configs?
Yes.
+(galileo,crownbay,coreboot-x86,bayleybay,qemu-x86,chromebook_link,minnowmax,chromebox_panther) arch/x86/lib/built-in.o: In function `__wrap___udivdi3': +(galileo,crownbay,coreboot-x86,bayleybay,qemu-x86,chromebook_link,minnowmax,chromebox_panther) build/../arch/x86/lib/gcc.c:25: undefined reference to `__normal___udivdi3'
Is your Fedora a 64-bit installation? And a 64-bit gcc compiler, right? I am using 32-bit Fedora with a 32-bit gcc. This is nothing related to gcc 5.x too. The same thing happens in gcc 4.x. If we want to use 64-bit gcc to build 32-bit U-Boot, gcc must ship with multilib support.
Regards, Bin