incorrect rounding in scaled_rounding with scale of 1

[dhebbeker]

Most rounding functions in scaled_rounding don't work as intended if SCALING==1. I suggest to

• either forbid to use a scaling of 1:
  • note in the documentation
  • static_assert in the code
• or use template meta-programming to return value if SCALING==1

Can you please clarify which behaviour is intended? Then somebody could start a pull request...

_{Based on ETL version 20.40.0.}

[jwellbelove]

It's a long time since I last looked at these.

[jwellbelove]

As I look at this again after 7 years, I can't help thinking that scale factors that are not a power of 10 do not really make a lot of sense.

[dhebbeker]

The SCALING could be the size of a container or frame. The following example would make sense:

const unsigned numberOfNecessaryFrames = etl::round_infinity_unscaled<sizeof(Frame)>(payloadLength);

[jwellbelove]

When I run that I get

for 8 bytes and frames of size 8: 2 frame(s) are necessary

Shouldn't that be 1 frame?

[dhebbeker]

I agree. That is subject to issue #1041