drivers/periph_spi: let spi_acquire return void

[maribu]

Contribution description

• Changed signature from int spi_acquire(...) to void spi_acquire(...)
  • Most SPI implementations and 99.9% of all users treat it like this anyway
  • By adding assert()s to verify parameters, this actually increases error checking compared to before, where both SPI implementations and user mostly didn't care about errors
• Updated soft_spi to match the periph_spi API
• Updated implementations
• Updated the few users, which actually checked the return value
  • Most checked the return value only during <foo>_init(), but afterwards never again
  • This is updated to add a pair of spi_acquire() and spi_release() if and only if NDEBUG is not defined.
  • The reasoning is: If assert()ions are disabled anyway, we won't trigger them no matter how hard we try :-)

Testing procedure

This should be mostly unscary. However, without assert()ions enabled, timing behavior changes a bit during initilization, as a useless pair of spi_acquire() and spi_release() is simply dropped.

Issues/PRs references

Depends on and includes:

• drivers/periph/spi: clean up error codes and doc #15896
• sys/arduino: Bug and style fixes in SPI #15910

[maribu]

Let's see if the CI finds any implementations / users that I missed.

Beware: While the WIP label is still present, I will squash without further notice ;-) I'll drop the WIP label once the CI is green.

[maribu]

The remaining failed build is an CI hiccup, so this is no longer WIP.

[maribu]

Cherry-picked review fixes that address @hugueslarrive's review in #15904 into this PR as well

[dylad]

LGTM.
Please squash !

[maribu]

Note that if an argument is used by an assert() only, it will be unused once the assert is disabled via NDEBUG. The reason is that assert() is (as required by the C standard) implement as a preprocessor macro.

[fjmolinas]

Go!

[maribu]

Thanks!

[gschorcht]

Shouldn't we do the same for i2c_acquire? All implementations just return 0 with only one exception

RIOT/cpu/cc2538/periph/i2c.c

Lines 234 to 242 in 97758b8

	int i2c_acquire(i2c_t dev)
	{
	DEBUG("%s\n", __FUNCTION__);
	if (dev < I2C_NUMOF) {
	mutex_lock(&lock);
	return 0;
	}
	return -1;
	}

which is a typical case for an assert. This was already the topic of a discussion in issue #10673, but after we removed the return value of spi_acquire, there is no argument to keep it for i2c_acquire.

[maribu]

I totally agree!

[gschorcht]

Ok, I will try to provide a PR in next weeks.