nrfx_..._uninit not idempotent: can't be safely called on an already uninitialized driver instance

RE: nrfx_..._uninit not idempotent: can't be safely called on an already uninitialized driver instance

Dan Halbert — Wed, 17 Jan 2024 20:33:55 GMT

The nrfx 3.2.0 release has added functions like nrfx_twim_init_check(), which return true or false based on whether the driver was initialized or not. This satisfies the alternative solution proposed in #43 (comment). Thank you.

RE: nrfx_..._uninit not idempotent: can't be safely called on an already uninitialized driver instance

tesc — Mon, 15 Oct 2018 13:06:13 GMT

Hi,

Thanks for linking to the github issue from this thread! What to do is a decision best to be taken by the SDK team, so I think leaving your PR as is is fine.

Regards,
Terje

RE: nrfx_..._uninit not idempotent: can't be safely called on an already uninitialized driver instance

Dan Halbert — Mon, 15 Oct 2018 12:38:36 GMT

Thank you. I also submitted a GitHub ticket at https://github.com/NordicSemiconductor/nrfx/issues/43 that essentially says the same thing -- I wasn't sure whether that was more logical, since I have a prepared Pull Request. I can modify my PR to do what you suggest (right now it just returns nothing, since the uninit functions return void).

RE: nrfx_..._uninit not idempotent: can't be safely called on an already uninitialized driver instance

tesc — Mon, 15 Oct 2018 11:06:17 GMT

Hi,

Thank you for reporting this. Bug reports and feature requests are always highly appreciated!

While I do not currently have any suggestions for better workarounds, I have notified the SDK team and they might have some ideas. In any case I have registered a feature request for solving the general use case, for instance by making nrfx_*_uninit() behave similarly to nrfx_*_init() and return NRFX_ERROR_INVALID_STATE if already uninitialized. IMHO that would be the best and most consistent way to solve this within our driver APIs.

Regards,
Terje