This is the mail archive of the
glibc-bugs@sourceware.org
mailing list for the glibc project.
[Bug math/15749] cbrtl fails for arguments > 1e308
- From: "joseph at codesourcery dot com" <sourceware-bugzilla at sourceware dot org>
- To: glibc-bugs at sourceware dot org
- Date: Wed, 17 Jul 2013 12:12:06 +0000
- Subject: [Bug math/15749] cbrtl fails for arguments > 1e308
- Auto-submitted: auto-generated
- References: <bug-15749-131 at http dot sourceware dot org/bugzilla/>
http://sourceware.org/bugzilla/show_bug.cgi?id=15749
--- Comment #2 from joseph at codesourcery dot com <joseph at codesourcery dot com> ---
The problem is that ldbl-96 cbrtl calls fabs when it should call fabsl.
This will also result in spurious underflows for arguments that underflow
doubles. (It shouldn't cause significantly inaccurate results for other
cases - the Newton-Raphson iterations should eliminate any significant
consequences of xm being slightly inaccurate.)
There's a standing issue in the wiki todo list to improve libm-test.inc
coverage of large and small input values including subnormals, which would
have detected this issue (and I suspect various other bugs not yet found)
- currently all finite nonzero test arguments to cbrt have absolute values
between 0.001 and 27.
--
You are receiving this mail because:
You are on the CC list for the bug.