wait locks deepsleep, Thread::wait does not

[janjongboom]

Description

In our docs (https://os.mbed.com/docs/v5.7/reference/sleep-manager.html) we state that when you call wait we automatically put the device in deep sleep. This is not true. For some reason wait() blocks deepsleep, but if you change it to Thread::wait it does not. Why?! This is super confusing. Why would there be two different behaviors when pausing a thread?

Then on non-RTOS mode wait actually can do deepsleep, so that makes it even more confusing.

All wait should behave equally and should not block deepsleep.

Issue request type

[ ] Question
[X] Enhancement
[ ] Bug

[c1728p9]

There was a PR for this change but it was rejected since it breaks backwards compatibility:
#6969
[Mirrored to Jira]

[kjbracey]

I missed #6969 - I would have approved it with a mod to split long and short delay behaviour on the float version.

I don't think this long-running API glitch can be justified on backwards compatibility grounds - we do change bigger things in minor releases.

The most likely breakage would be people doing wait_ms(1) to get 1 milllisecond. If we change that to just ticks, they could get as little as 0 milliseconds, which might break some hardware timing.

I wonder if following the 5.9 lpticker changes we could do a compromise solution which uses the lpticker to achieve better-than-rtos-tick precision without the lock?

[Mirrored to Jira]

[kjbracey]

Then on non-RTOS mode wait actually can do deepsleep,

No it can't, it spins and uses 100% CPU.

[Mirrored to Jira]

[klaas019]

Could someone update this example https://os.mbed.com/docs/v5.6/reference/sleep-manager.html . It is broken.
[Mirrored to Jira]

[adbridge]

Internal Jira reference: https://jira.arm.com/browse/IOTCORE-23