On Tuesday 12 July 2022 19:15:45 Tom Rini wrote:
On Wed, Jul 13, 2022 at 01:11:35AM +0200, Pali Rohár wrote:
On Tuesday 12 July 2022 18:58:31 Tom Rini wrote:
[snip]
It really isn't, that's why the plain text environment stuff is a good step in the right direction. What I'm not sure about how to best handle is making it easy for a user, rather than developer, to override environment values too. What I mean he
Maybe the important question is: Why is needed CONFIG_PREBOOT define at all? Why it is not enough just to set preboot= env variable? And same questions for all other CONFIG_<name_of_env> options.
I'm sorry if this sounds snarky, I don't intend it to. But, size. Maybe that doesn't matter so much these days for the size savings on run_preboot_environment_command().
Size is really a good argument.
For all this stuff is needed some stronger tool/language than Kconfig, e.g. C preprocessor (like it was before in config.h) or any similar stronger macro language (e.g. m4) or script languages (shell / python).
Yes, so, take a look at 5c3f6a320678d64c9fa0c42755488822a033b567 as an example of moving a board away from board config.h and to something else. What I'm not sure on is how to best handle preboot in this case as I'm not quite liking "Enable CONFIG_PREBOOT and then edit .../boardname.env to set preboot to the right value".
-- Tom
With that commit, all env variables are moved into separate file. This file can be parsed by other tools and at compile it is possible to extract current value of preboot= env variable or check if preboot= is defined. Cannot be by this logic automatically defined/expanded CONFIG_PREBOOT symbol? Of course, it would not be Kconfig symbol anymore.
That sounds interesting.
-- Tom
And this solve also issue with size (if we can determinate at compile time if preboot= env is defined or not).