Is this an info or a warning? There are other goto err including ones with malloc issues.

This depends on how msg_identify_magnetic_space_group_type is meant to be used and if any issue is recoverable, expected, etc.

This is an unreachable situation without something floating-point error. So, I guess this will be info.

If it's unreachable like an assert it should be warning. Assuming the floating point error is not a dynamic tolerance

I will leave whether info or warning to you, but let me put more context.
If the parameters symprec and mag_symprec are sufficiently small, this is like the assert. But if these are too large, this assertion may fail. In that case, I recommend to decrease these parameter values.

Since these are tied to a user input, these would not be classified as assert (assert is only for debug if you are sure that users would not encounter it, that is why it is gated by NDEBUG), but as runtime try-except-like functionality.

It seems it could be either info or warning, but these are quite hard to detangle, so I am not sure how to classify either, since some are in a for loop, some are not. Ultimately we are mostly limited by the C language that we cannot use libraries like spdlog, which would make these issues obsolete (since we can dynamically enable/disable/redirect logging, including having python native logs)

Let's go with info.

For these tolerance, I will move them to debug and have only one info when it exits

Comment above seems to indicate it is a proper warning that at least we should know about if it happens

This code is introduced for layer group, but we will not want to expose the feature for users since we should reconsider the algorithm, and the implementation will be redesigned. This will be a work not only the code redesign but require some scientific discussion. The code lines for the layer group implementation were inserted in many places and the refactoring will need some effort and take time. So I just have been leaving it untouched.

This part is conflicting with the comment above

This function transforms the basis but when the distortion of the input crystal structure is large and with naive choice of the tolerance, this fails. count == size is a minimum check that should satisfy the condition. So the comment says Sanity check. Could you recommend a better word?

Maybe this is a debug_print?

So if I see this message, I will see that the crystal structure is distorted. But as you commented, it is not possible to see it from the local piece of the code.