jFP religious wars

[Gabriel Dos Reis]

Robert Dewar <dewar@gnat.com> writes:

| Zack Weinberg wrote:
| > "R. D. Flowers" <rd@chatta.us> writes:
| >
| >>would not some flag like -fp-pedantic (or however to fit in the
| >>namespace) be a good idea (for both decent sets of folks: the
| >>fast-and-a-little-dirty and the clean-and-a-little-slow people)?
| > Maybe.  What would it do?
| 
| I don't think this would work. There is nothing "pedantic" in
| the carefully formulated rules that make it easier/possible
| to write accurate floating-point code. It is far better to
| have the existing flags for -ffast-math etc, which clearly
| warn that the resulting semantics may be unreliable.

Fully agreed.

| The use of -pedantic to cover little used language features
| or rules is one thing, the use of a flag like this to give
| incorrect semantics.

In fact, the flag "-pedantic" is misnamed -- at least, for what it does
currently in C++ (I think the same is true for C).  It is there only
for historical accident.  Multiplying that naming by a factor greater
than 1 would be misguided.

-- Gaby