Hi Wolfgang,
On Tue, 4 Aug 2020 at 01:59, Wolfgang Wallner wolfgang.wallner@br-automation.com wrote:
Hi Simon,
-----"Simon Glass" sjg@chromium.org schrieb: -----
Betreff: Re: chromebook_coral: build failure
Hi Wolfgang,
On Fri, 31 Jul 2020 at 05:44, Wolfgang Wallner wolfgang.wallner@br-automation.com wrote:
Hi Simon,
while trying out your ACPI patches I tried to compile
chromebook_coral_defconfig
as a reference. Building this defconfig fails for multiple
definitions
of "_X86EMU_env":
$ git checkout v2020.07 $ make distclean $ make chromebook_coral_defconfig $ make
[...] LDS u-boot.lds LD u-boot ld.bfd: drivers/built-in.o:(.bss._X86EMU_env+0x0): multiple
definition of `_X86EMU_env'; arch/x86/lib/built-in.o:(.bss._X86EMU_env+0x0): first defined here
make: *** [Makefile:1755: u-boot] Error 1
A quick and dirty workaround was to drop "#define CONFIG_BIOSEMU"
from
include/configs/x86-chromebook.h. With this change
chromebook_coral_defconfig
compiles fine.
Do you know what is going on here, and what might be a proper fix?
I don't see that problem myself, [...]
It seems to be triggered by a change of default flags in GCC 10.
I can reliably reproduce the build error on my main machine (Arch Linux with GCC 10), but it builds fine on a Debian 10 with GCC 8. A colleague told me about "-fno-common", which was changed to be on by default in GCC 10:
https://gcc.gnu.org/gcc-10/porting_to.html
I can build chromebook_coral_defconfig on my Arch with GCC 10 with:
make KCFLAGS="-fcommon"
However, I'm not sure what the initial problem is, as I don't know the relevant code parts. Both of the following files declare a global variable "_X86EMU_env":
drivers/bios_emulator/x86emu/sys.c arch/x86/lib/bios.c
Should they both be part of the build for Chromebook Coral?
I think I figured this out and sent a patch. Thanks for explanation.
[...] but I do see an ACPI compile problem now. that I rebase. I have added a patch for that and pushed the tree to u-boot-dm/coral-working
Yes, this is the reason why I wanted to compile chromebook_coral_defconfig as a reference :) I saw the ACPI compile problem when I tried your ACPI patches on my APL board and wanted to see whether I had broken something.
Could you please move the fix to the beginning of the part D series when you send V2? It might make testing individual patches of that series easier if the fix is at the beginning.
I can't remember what the ACPI compile problem is, sorry.
Regards, Simon