Bug#290438: acknowledged by developer
"Kevin B. McCarty" <kmccarty@Princeton.EDU> writes:
> Falk Hueffner wrote:
>> this seems to be the well-known exces precision problem of the
>> i386, so closing.
>
> I (mildly) protest the closing of this bug. The bug is not in fact
> "fixed" -- just "well-known". Even though it is due to a CPU issue
> rather than a compiler problem, unoptimized code returns different
> results than optimized code, which really _is_ a compiler bug.
Unoptimized code returning different results than optimized code is
not a bug per se. This is (or could be regarded) as a bug because it
violates the IEEE standard.
> #290438 also appears to be the only such bug in the BTS filed
> against any version of g77. I suggest re-opening it, maybe
> downgrading to "minor" and tagging "fixed in 3.4/4.0" (if this is in
> fact the case) and/or "sarge", so people can find it in the BTS.
It's not fixed, and never will be. So we can reopen it and tag it as
"wontfix", but honestly, I don't think it will ever keep somebody from
filing a new bug report on this. I'll let others decide.
--
Falk
Reply to: