Messages in this thread | | | Subject | Re: nds32 build failures | From | Guenter Roeck <> | Date | Mon, 16 Apr 2018 06:59:38 -0700 |
| |
On 04/16/2018 06:22 AM, Guenter Roeck wrote: > On 04/16/2018 02:06 AM, Greentime Hu wrote: >> 2018-04-16 11:58 GMT+08:00 Guenter Roeck <private@roeck-us.net>: >>> I thought I should give the brand new architecture a try. Unfortunately, >>> that was not very successful. >>> >>> Build reference: v4.17-rc1 >>> gcc version: nds32le-elf-gcc (GCC) 7.3.0 >>> >>> Building nds32:defconfig ... failed >>> >>> arch/nds32/include/asm/nds32.h: In function 'GIE_ENABLE': >>> arch/nds32/include/asm/nds32.h:25:2: error: implicit declaration of function >>> '__nds32__gie_en'; did you mean '__nds32__'? >>> >>> arch/nds32/include/asm/nds32.h: In function 'CACHE_SET': >>> arch/nds32/include/asm/nds32.h:38:18: error: implicit declaration of >>> function '__nds32__mfsr'; did you mean '__nds32__'? >>> >>> arch/nds32/include/asm/nds32.h:38:32: error: 'NDS32_SR_ICM_CFG' undeclared >>> >>> arch/nds32/include/asm/nds32.h:41:32: error: 'NDS32_SR_DCM_CFG' >>> >>> Am I missing something ? >>> >> Hi, Guenter: >> >> This built failure is because the toolchain version you used is not >> supported the latest intrinsic function/macro. >> We are sending the latest patchset now and we expect the whole new >> features will be supported in gcc8.0.0 and binutil2.31+. >> >> If you'd like to get these new features of toolchain, you may use the >> github version. >> This is the built-script repo. https://github.com/andestech/build_script.git >> > Hmm. How comes that the defines are not there ? Also, is nds32_intrinsic.h supposed to be auto-generated ? > Answering my own question. This file is part of gcc. Presumably, the missing defines are declared in that file ? Interesting, I always thought that the Linux kernel is supposed not to include compiler include files, to avoid issues such as license poisoning. One lives and learns.
Guenter
> The toolchain as provided isn't usable for me since it is not position independent. > > Arnd, it might make sense to drop the nds32 toolchain from kernel.org; its presence is at best misleading. > > Guenter > >
| |