Mbed-os: STM32F4: lp ticker failures on master

[0xc0170]

Description

• Type: Bug
• Priority: Major

---

Bug

Reference: https://github.com/ARMmbed/mbed-os/pull/5768#issuecomment-355425482

There are at least 2 failures from yesterday, the report provides binaries and test failures

@ARMmbed/team-st-mcd please review

[jeromecoutant]

Hi
Do you have several test benches with NUCLEO_F401 ?
To be sure, can you unplug all these boards and make the test again ?

[0xc0170]

To be sure, can you unplug all these boards and make the test again ?

We can. How that would help (what is causing these failures) ?

[jeromecoutant]

RTC configuration has been updated in a recent PR.
But as long as the board is powered, the new configuration is not applied as we checked before RTC init procedure if RTC has already been configured.

[0xc0170]

Do you have several test benches with NUCLEO_F401 ?
To be sure, can you unplug all these boards and make the test again ?

@studavekar @kegilbert all of the boards were reset previously?

[0xc0170]

Affected PRs:

• https://github.com/ARMmbed/mbed-os/pull/5708
  
  
  
  • https://github.com/ARMmbed/mbed-os/pull/5768
  
  

[0xc0170]

RTC configuration has been updated in a recent PR.

What is the PR, can you reference it ? The git log for the RTC API: https://github.com/ARMmbed/mbed-os/commits/master/targets/TARGET_STM/rtc_api.c . I recall we had similar error a while ago, and devices were reseted . Is there a way to make this backward compatible ? (no hard reset target needed).
Un-powering device is not a solution that we seek

To me, this seems like the problem is back, thus might be a bug somewhere?

[jeromecoutant]

Is it possible that CI is sometimes testing MBED branch started from an "old" commit ?

[0xc0170]

Is it possible that CI is sometimes testing MBED branch started from an "old" commit ?

That can be (although rebase is always good 😈 ). The affected PR were based on the recent master.

In case there was PR on the old commit, should not lead to a breakage like this? One test just finished few minutes ago, the tests were green

[0xc0170]

Just got two F4 lp ticker failures here https://github.com/ARMmbed/mbed-os/pull/5597#issuecomment-355995753

[studavekar]

Ticker failure on F7 not sure if its related https://github.com/ARMmbed/mbed-os/pull/5795#issuecomment-357342685 & https://github.com/ARMmbed/mbed-os/pull/5760#issuecomment-357329201

[0xc0170]

5841 also

@jeromecoutant I'll reproduce today this ticker tests locally with the latests tests that run.
We might need to disable lp ticker for those devices if not fixed soon.

[mprse]

I tried to reproduce failure of the lp ticker tests on NUCLEO-F429ZI board without success.

Tried to run the following tests:

• tests-mbed_hal-lp_ticker,
• tests-mbed_drivers-lp_timer

several times using GCC_ARM, ARM and IAR tool-chains and develop profile with the following additional flags (tests on CI are compiled with these flags):

• "-DMBED_TRAP_ERRORS_ENABLED=1",
• "-DMBED_HEAP_STATS_ENABLED=1",
• "-DMBED_STACK_STATS_ENABLED=1".

No failure has been detected.

Performed also the following test:

• created temp branch with head commit from 5th September,
• pulled out usb cable and connect it back,
• run lp ticker test (passed),
• switch back to master and run the test again (passed x3).

[jeromecoutant]

@0xc0170
I think I could find some optimization in the code execution.
Tests on going

[jeromecoutant]

What is the status after 1 week test ?
Thx

[0xc0170]

Looks like it is fixed, I'll close it