Dear Nico,
thank you for your reply.
Am Mittwoch, den 27.03.2013, 20:49 +0100 schrieb Nico Huber:
Am 27.03.2013 14:04, schrieb Paul Menzel:
Dear coreboot folks,
using latest master
commit 3cc0d1eb3f611cb7bf0e45d8ccdb0c84f54f54dc Author: David Hendricks <dhendrix@chromium.org> Date: Tue Mar 26 16:28:21 2013 -0700 exynos5250: assign RAM resources in cpu_init() […] Reviewed-on: http://review.coreboot.org/2923
it looks like libpayload’s PDCurses backend is not including `libpayload-config.h` in some files so the configuration is ignored, right? I noticed this due to the following warning.
/src/coreboot/payloads/libpayload(master) $ make clean /src/coreboot/payloads/libpayload(master) $ make […] CC curses/pdcurses-backend/pdcscrn.libcurses.o curses/pdcurses-backend/pdcscrn.c: In function 'PDC_scr_open': curses/pdcurses-backend/pdcscrn.c:75:5: warning: "CONFIG_SPEAKER" is not defined [-Wundef] […]
I don't see the include missing: `pdcscrn.c` includes `lppdc.h` which, in turn, includes `libpayload-config.h`.
No warnings are printed for the other files, because `#ifdef` is used there. But these macros will never be defined, when `libpayload-config.h` is not included, if I am not mistaken.
So is the solution to just include `config.h` or `libpayload-config.h` everywhere?
/src/coreboot/payloads/libpayload(master) $ grep -R CONFIG_SPEA . ./configs/defconfig:CONFIG_SPEAKER=y ./curses/pdcurses-backend/pdcscrn.c:#ifdef CONFIG_SPEAKER ./curses/pdcurses-backend/pdcutil.c:#ifdef CONFIG_SPEAKER ./curses/tinycurses.c:#ifdef CONFIG_SPEAKER ./drivers/Makefile.inc:libc-$(CONFIG_SPEAKER) += speaker.c ./build/config.h:#define CONFIG_SPEAKER 1 ./build/auto.conf:CONFIG_SPEAKER=y ./build/libpayload-config.h:#define CONFIG_SPEAKER 1 ./.config:CONFIG_SPEAKER=y ./tests/libpayload-config.h:#define CONFIG_SPEAKER 1
If you tell me the preferred solution, I am going to submit a patch.
Libpayload's config mechanism doesn't define unset options (unlike the one coreboot uses).
My problem, as seen in my paste above, is that `CONFIG_SPEAKER` is defined and I still get the error.
Therefore, the use of `#ifdef` is correct. You can change `#if CONFIG_...` to `#ifdef` where appropriate.
I (accidentally) pushed such a patch to Gerrit [1].
Thanks,
Paul