Andrew Gierth <andrew@tao11.riddles.org.uk> writes:
> The problem is that if we're relying on -fexcess-precision=standard
> semantics in places besides infinity checks, then we won't get those
> semantics on clang/i386/no-sse2 since it has no comparable option. (What
> are we doing about compilers for x86-32 other than clang and gcc?)
Well, the bigger problem is that we don't really know exactly where
nonstandard precision semantics might cause visible behavior changes.
It's just about a dead certainty, IMO, that there are some other
places we don't know about because the regression tests don't expose
them; or that future code rearrangements might create new problems.
Given that (if I understand Andres correctly) -fexcess-precision=standard
behavior is required by C99, it's hard to get excited about expending
a whole lot of work here. I'm fine with adding some compiler options
or telling the user to do so, but I don't think we should do anything
much beyond that.
regards, tom lane