On Sun, Oct 1, 2017 at 8:11 PM, Piotr Król piotr.krol@3mdeb.com wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA512
Hi all, PC Engines apu2 (and probably apu1) not booting on recent master. I bisected problem to: d4955f0ade18cafde4a3ea20885eb9fbdc5b4514
AGESA: Move API interface under drivers/
My fault, see: https://review.coreboot.org/21840
That was insufficient testing on my side, I blame too much pending work in the review queues spanning couple branches... I mostly test only without BINARYPI_LEGACY_WRAPPER, and some of that work has still not landed in gerrit.
I'm not sure if some config options are missing. I tried to select DRIVERS_AMD_PI and CPU_AMD_PI in Kconfig, but those not help.
Once you have bisected to an offending commit, you can do builds that strips away all the revision hashes:
abuild --timeless -at pcengines/apu2 diff -qr build-ok/ build-fail/
The list of just the .o files is still a long one, but two further things you can do here: Strip debug symbols from .o files, and define ASSERT() without __FILE__ or __LINE__. If you know the commit was supposed to create a binary identical image, this is one way to pinpoint where the difference is.
This approach has saved the day before when AGESA includes have gone crazy.
Kyösti