NRF52: Stall on SPI Master + RTC IRQs and compile fixes for non SOFTDEVICE_PRESENT target

[drahnr]

Description

Contains a set of minor fixes to avoid a crash if IRQ of SPI master and RTC fire due to the sequence of IRQ locking.

The bug is reproducable with a certain binary (priv) and jumper.io, with real hardware this causes spurious hangs.

It also adds app_fifo./h from nordic SDK to make it build without SOFTDEVICE_PRESENT flag.

Pull request type

[x] Fix
[ ] Refactor
[ ] New target
[ ] Feature
[ ] Breaking change

[0xc0170]

This is targeting 5.8 as this is not applicable to master (because of the recent refactors) ?

[marcuschangarm]

All these changes are for NRF51. Have you tried the new NRF52 implementation currently in master and to be released in 5.9?

[drahnr]

Without those fixes and removing the SOFTDEVICE_PRESENT macro it will not compile my project, which is NRF52_DK derived.

I accidentally aplied the fix to the NRF51 fix though I meant to do that for NRF52, the fix itself is necessary for both.

I can not confirm or deny that the bug is fixed in master, I did not look into the source yet and only a certain test can reveal it which fails to build on master to many breaking changes...

[cmonr]

@drahnr Looks like this slipped our attention due to release activities. The fix would have to come into master instead of directly targeting the release branch itself.

However, the last 5.8 minor release is about to be made in 12 hours, and all other PRs will be targeting 5.9. I would suggest trying to compile your project against the 5.9 branch and if the issue still exists, open a PR targeting master. Once the PR lands into the repo, we have a process for bringing the changes into the release branch.

[drahnr]

@cmonr so this PR can essentially be closed I guess, correct? Hopefully I can create another test tomorrow / early next week.

[cmonr]

@drahnr Correct.

Sounds good. Looking forward to it!