sys/ztimer: auto_init react to possibly missing ztimer_periph_timer

[kfessel]

Contribution description

In #16172 I assumed ztimer_periph_timer to always be available (it was always configured in config.h) but some recent PR by @haukepetersen (#16322 identified by @jue89) showed this is no always the case.
This adapts ztimers auto_init.c to that case.
(if somehow ztimer_periph_timer is not included but required an error message will be shown)

Testing procedure

read and tests/ztimer_xsec

Issues/PRs references

#16172

[kfessel]

@maribu, @kasper may be you want to have a look at this

[maribu]

See inline comments

[maribu]

I think there should be an #else clause added that gives an #error in cause ztimer_msec is used, but neither the periph_rtt nor the periph_timer backend is selected. This might not be possible to configure, but better be safe than sorry.

[kfessel]

error messages for no Backend selected (e.g.: cause there is non) happen in step 3

[maribu]

Suggested change

	# define INIT_ZTIMER_TIMER 1
	# endif
	# define ZTIMER_SEC_CONVERT_LOWER_FREQ ZTIMER_TIMER_FREQ
	# endif
	#endif
	# define INIT_ZTIMER_TIMER 1
	# endif
	# define ZTIMER_SEC_CONVERT_LOWER_FREQ ZTIMER_TIMER_FREQ
	# else
	# error "No backend available to implement ZTIMER_SEC"
	# endif
	#endif

?

[kfessel]

this error message is in l 207 ante commit l 202 post commit

in step 1 a backend is selected (step 1 has all the precedence logic)

if non is (could be) selected a error message is printed in step 3

we can have extra error message in step one but we don't need them there

[fjmolinas]

ping @maribu

[maribu]

The code looks good to me. But I guess it's not a bad idea to test this with all combinations. I think I might have some time for that tomorrow

[fjmolinas]

I think #16553 can fix this as well

[fjmolinas]

@kfessel with #16553 in I think htis is no longer needed, can you confirm?

[kfessel]

@kfessel with #16553 in I think this is no longer needed, can you confirm?

#16553 make the problem not appear: The configuration is as it should be, But since the ztimer auto_init is kind of long and errors may show up in lines that may not show the most simple issue that might cause them.

It also improves the ZTIMER_SEC/MSEC/USEC similarity which makes it more easy to read.

I still would like to see this merged.

#16322 would have been easier to solve if this had been in.
Similar configuration issues may happen in future this provides a better hint.

[kfessel]

#17372 would have been easier to solve if this had been in.

[kfessel]

#17654 got this in