(j3.2006) [Re: Spreadsheet from meeting 167]

[Tobias Burnus]

Hi,

Dan Nagle wrote:
> Do you believe J3 spent significant amounts of time
> on the annex

I cannot comment on that part as I wasn't at the meetings and as I 
cannot judge what J3 would have done otherwise. (As written previously, 
I think that TR is useful even if not many users will read it.)

> or do you dislike warning for real equality?

Well, warnings do not harm, especially if one has to activate them 
explicitly. However, I was surprised that warning for real/complex 
(in)equality turned out to be much less useful than I had expected; the 
number of false positives is really huge. Thus, this is a very special 
warning not suitable for being enabled by default or even by the -Wall 
flag. Note that the limited usefulness of the flag doesn't imply that it 
has to be removed from the annex - after all, the warning can help find 
bugs.

What's your experience with this warning?

> In any case, were you forced to add the warning?

No. I also didn't complain about the suggested warnings. On the 
contrary, I proposed the implementation in one compiler. I only wanted 
to counter your "In any case, the affect on compiler suppliers so far is 
exactly zero" as it has (positively) affected one compiler.

In any case, I don't see anyone forcing compiler vendors to implement 
something due to the vulnerability TR. In addition, as the TR is mostly 
about "language use", the parts directed to compiler vendors and 
language committees are rather small. Besides (vague) suggestions, 
compiler vendors will be mostly affected by those suggestions which end 
up in the Fortran standard. Namely those items starting with "Requiring 
that processors" under "Future standardization efforts should consider".

(Regarding the content of the draft annex, see also the voting comments 
in N1943. I haven't read the new version N1947, only N1929.)

Tobias