On Wed, Sep 16, 2015 at 01:46:34PM +0200, Ludger Dreier wrote:
Hi Tom,
Am 2015-09-15 18:54, schrieb Tom Rini:
Was there perhaps some sort of init ordering issue in a board which did have EEPROM used as the backing store for env? I could see a case where perhaps (and Ludger, can you test this as well please?) the problem is that for env_eeprom env_init() needs to just default to the built-in (like it's basically doing today) and env_relocate_spec() needs to do the read, check which is valid, etc, etc, dance which it is _not_ doing today.
as I understand your idea, the code from both versions of env_init (redundant env and non-redunant) which is framed by "#ifdef ENV_IS_EMBEDDED" may potentially be moved to, or called by env_relocate_spec. I think this could work. If this ok for you, I can come up with a tested proposal. But this would need some days.
Roughly speaking, yes. env_init() should just be the "use the default built-in env" code and env_relocate_spec() should do all of the hard work that env_init() used to do.