[Cfp-interest] Annex F specification for Part 2

[Jim Thomas]

>> 8.2 Conversions among decimal floating types, and between decimal
>> floating types and generic floating types
>> 
>> PDF page 12, last paragraph: Remove "is in the range of values that
>> can be represented but".
> 
> do we assume conformance to part 1? seems not - is this ok?
> without part 1 (or annex F conformance) we don't know generic floating types have infinities - so need to retain words or equivalent - REVISIT

Several parts of Annex F contain specification that is required by IEC 60559 but that is not included in Part 2, e.g.,
F.4 Floating to integer conversion
F.6 The return statement
F.7 Contracted expressions
F.8 Floating-point environment
F.9 Optimization

There is some support in our study group for allowing an implementation to conform to Part 2 even if it doesn't conform to Part 1, or even to C11 Annex F. Here are some alternatives for how we might proceed:

0.	Go ahead and  make Part 1 conformance a requirement for Part 2 conformance. An implementation could still say it "adheres to the specification in Part 2 for decimal floating types", or some such, without claiming formal conformance to Part 2. This approach would allow for a simpler spec - easier to write and maintain and easier to understand.

1.	Incorporate by reference the needed numbered clauses from Annex F as modified by Part 1 into Part 2. Where only part of a clause is needed, replicate the specification in Part 2. This would allow decimal-only conformance. The specification would become somewhat more complicated.

2.	Replicate all needed parts from Annex F as modified by Part 1 into Part 2. This would allow decimal-only conformance. The spec would be self conned, hence potentially clearer. Maintenance would be an ongoing problem.

3.	Modify the changes to Annex F in Part 1 to separate the general specification from the basic binary specification. Make conformance to Part 2 dependent on adherence to the general specification in Annex F as modified by Part 1 (but not dependent on conformance to Annex F as would be indicated by __STDC_IEC_60559_BFP__). This would allow decimal-only conformance. The spec would become more complicated/obscure. I fear this approach would require a lot of changes to Part 1.

Your thoughts?

-Jim