On Thu, Feb 25, 2021 at 09:35:13PM +0100, Marek Behun wrote:
On Thu, 25 Feb 2021 15:28:56 -0500 Tom Rini trini@konsulko.com wrote:
On Thu, Feb 25, 2021 at 09:07:40PM +0100, Marek Behun wrote:
On Thu, 25 Feb 2021 14:31:42 -0500 Simon Glass sjg@chromium.org wrote:
We should not need CONFIG_DM here...it should be enabled for all boards. You can always disable MTD for a board if not, or send a removable patch.
If for some reason you do, please use if (IS_ENABLED() so that 'dev' can always be declared.
Simon, it still isn't enabled for all boards. For example tqma6s_wru4_mmc_defconfig does not compile with this. I actually wrote this into commit message:
Although CONFIG_DM is compulsory since v2020.01, there are still some boards (for example tqma6s_wru4_mmc_defconfig) that don't enable it.
But if breaking such boards is not a problem anymore, I will gladly just remove the ifdefs :) Should I?
So, I'm working hard at dropping boards that are well past migration deadlines. That specific one fails DM_MMC more importantly, and I will be dropping it if it's not migrated, after v2021.04 is out. What else fails? If you rebase your series on my WIP/remove-non-AHCI_LIBATA-drivers (as it has the most board removals in it), what fails to build if your series is just DM only?
I haven't tried building all boards, just quickly found a defconfig with disabled CONFIG_DM :) With removing the CONFIG_DM ifdefs this series shouldn't break anything that is already past migration deadline.
Well, past migration deadline doesn't mean it can cause CI to fail. I'm just now throwing things out that are 2 years past the deadline, and we'll probably be doing that for a few releases.