Hi @LinoBarreca
Is there any reason to close it?

@fpistm yes. Let me explain it.
Actually, despite the "logic" for me was good, it seems I made a mistake.
I made the assumption that the timer registers could be set even when the timer were not "active" (before resume).

@ABOSTM told me that, if the timer doesn't have a clock, you can't access its registers. I assume he is right because he is way more expert than me.

what was the logic behind my PR:

• user creates an instance of HardwareTimer()
• user calls attachInterrupt() to say "I want to receive the interrupt when the timer will be active"
• attachInterrupt remembers the user callback and sets a register on the timer (this fails according to Alexandre)
• user calls resume()
• resume gives the clock to the timer and starts it. Since the above call failed the register in the timer will not have the information "I must trigger the IRS" and the user callback never gets called.

I thought my logic was working because with the setOverFlow() (actually called before resume()) we still access a register (ARR) of an "unstarted timer" so I thought that should work the same with the DIER register (called by attachInterrupt() ). I didn't really understand why, before resume, we can write to ARR but not DIER but if he says so I believe him (and I closed the PR because it worked in my mind but it will not work in real life)

If I misunderstood something (or you could clarify my doubt), however, let me know :)

Hi @LinoBarreca , Your understanding of the issue is correct,
nevertheless, as I said in #763, I am thinking about enabling timer peripheral clock at HardwareTimer object creation.
Thus if we go for this implementation, your pull request makes sense. So we can keep this PR opened

ok. Thanks Alexandre but...can you please explain me why, with the timer clock disabled, we can write to ARR (setOverflow) and PSC (setPrescaler) but not to DIER?
(again, I trust you, but it seems a bit not logical to me...I don't mean to offend anyone... there must be a reason why....but I'd like to know it)
....if I had a board I could test it..but unfortunately I don't own one...

From hardware point of view it is not possible to write ARR or PSC registres (or any register) with timer clock disabled.
Concerning SetOverflow and setPrescaler, if you look carefully, there is a trick:
that is right thas we try to update registers (in case clock is already present, it allows to update ARR and PSC on the fly while timer is running), but in case clock is not present, we also save the value in structure so that it will be configured later (during the resume):

  __HAL_TIM_SET_AUTORELOAD(&_HardwareTimerObj.handle, ARR_RegisterValue);
  _HardwareTimerObj.handle.Init.Period = ARR_RegisterValue;

  __HAL_TIM_SET_AUTORELOAD(&_HardwareTimerObj.handle, ARR_RegisterValue); <-this fails if the timer doesn't have a clock (=not yet started) but succeeds if the timer is running.
  _HardwareTimerObj.handle.Init.Period = ARR_RegisterValue; <- this saves the value and sets it again at timer start

now I understood. You were very kind.

I add label OnHold as it depends on #763

Hi @LinoBarreca
@ABOSTM implemented #763 and included your PR during this.
I you can provide your feedback, this will be fine, then we could close this one as it is now duplicated in #806.
Thanks in advance

I close this one has integrated in #806.