Opened 14 years ago
Last modified 12 years ago
#2769 new Bugs
x86_64 + g++: wrong results for sin()
| Reported by: | anonymous | Owned by: | Boris Gubenko |
|---|---|---|---|
| Milestone: | Component: | interval | |
| Version: | Boost 1.34.1 | Severity: | Problem |
| Keywords: | x86_64 interval transcendental | Cc: |
Description
When using boost's interval library under x86_64-Linux with g++ we encountered strange results for sin(): For example for an interval [-2.1,-2.1] sin() returns [nan,nan].
Attachments (1)
Change History (4)
by , 14 years ago
| Attachment: | interval_test.cpp added |
|---|
comment:1 by , 12 years ago
Seems to still be a problem in Boost 1.42: see the report from Debian: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=584325
follow-up: 3 comment:2 by , 12 years ago
The underlying C cos function has problems as well, even outside Boost.Interval. This bug appears to be a manifestation of something like http://sourceware.org/bugzilla/show_bug.cgi?id=3976 (which also applies to many other functions). According to that discussion, though, the standard library is not required to respect round-down or round-up for transcendental functions at all.
comment:3 by , 12 years ago
Replying to jewillco:
According to that discussion, though, the standard library is not required to respect round-down or round-up for transcendental functions at all.
Looks like this should be marked as wontfix?
c++-File depicting the problem