diff options
| author | enge <enge@211d60ee-9f03-0410-a15a-8952a2c7a4e4> | 2010-07-15 15:38:54 +0000 |
|---|---|---|
| committer | enge <enge@211d60ee-9f03-0410-a15a-8952a2c7a4e4> | 2010-07-15 15:38:54 +0000 |
| commit | 28ee86b61c7eacf03545df2dbb170c8a3c4ca6c2 (patch) | |
| tree | eca3e9f8540b252fed3976ee9ae9bf5ece1c43a4 /BUGS | |
| parent | 809ef11df2efae9573e1b6aef12544bf8551f1c8 (diff) | |
| download | mpc-28ee86b61c7eacf03545df2dbb170c8a3c4ca6c2.tar.gz | |
tan.c: reverted r795 and r796 and implemented simple fix
now, an overflow in sin and cos leads to NaN+i*NaN even if
the correct result is representable
git-svn-id: svn://scm.gforge.inria.fr/svn/mpc/trunk@808 211d60ee-9f03-0410-a15a-8952a2c7a4e4
Diffstat (limited to 'BUGS')
| -rw-r--r-- | BUGS | 3 |
1 files changed, 3 insertions, 0 deletions
@@ -7,4 +7,7 @@ sin(x)*sinh(y) is representable. If furthermore an underflow occurred in sin(x) (which has not been observed in practice), then the return value would be NaN*(+-inf)=NaN. + As another example, tan is computed as sin/cos; if there is an overflow + in both sin and cos, then NaN+i*NaN is returned even if the result + may be representable. |