Front page | perl.vmsperl |
Postings from June 2002
"Craig A. Berry" <craigberry@mac.com> wrote:
:Urk. I found a case where this isn't right (or actually
:lib/Math/Trig.t found a case where it isn't right):
:
:$ perl -e "my $x = 100.00000000000000001; print $x;"
:100
:
:so far so good, but now add another zero after the decimal:
:
:$ perl -e "my $x = 100.000000000000000001; print $x;"
:0
Could you try the patch below? It doesn't break anything here, but
I can't easily coerce my platform to use the new codepath.
I can't remember if C offers anything closer to a perlish 'redo'
than the <--i; continue;> used below.
Hugo
--- numeric.c.old Sun Jun 9 18:55:23 2002
+++ numeric.c Wed Jun 19 15:24:12 2002
@@ -902,9 +902,22 @@
}
/* combine components of mantissa */
- for (i = 0; i <= ipart; ++i)
- result += S_mulexp10((NV)part[ipart - i],
+ for (i = 0; i <= ipart; ++i) {
+ NV testresult = S_mulexp10((NV)part[ipart - i],
i ? offcount + (i - 1) * PARTSIZE : 0);
+ if (testresult == 0.0 && part[ipart - i] != 0 && expextra < 0) {
+ /* since we calculate 100.001 as '100001 / 1000' we are in
+ danger of rounding to 0.0 rather than 100.0 if the
+ numerator overflows. Catch this, reduce precision, and redo
+ */
+ result /= 10.0;
+ --offcount;
+ ++expextra;
+ --i;
+ continue;
+ }
+ result += testresult;
+ }
if (seendigit && (*s == 'e' || *s == 'E')) {
bool expnegative = 0;
Thread Previous
|
Thread Next