------- Comment #7 from jakub at gcc dot gnu dot org 2006-09-04 12:14 -------
I have downloaded your tarball and it seems you are not building gcc to default
to 128-bit long double on a glibc 2.4+ system. That's very bad idea,
you either shouldn't have upgraded to glibc 2.4+, or should have switched
everything to 128-bit long double.
While --with-long-double-128 libstdc++.so.6 is ABI compatible with older one,
--without-long-double-128 exact ABI of the math stubs depends on whatever
configure tests find out. With older glibc's, log10l etc. is not in system
libm (long double was just 64-bit there and so log10 did the same), but with
glibc 2.4+ it is present in libm (and implements the 128bit long double
logarithm etc.).
Guess something like:
--- libstdc++-v3/src/compatibility.cc 2006-02-23 09:55:01.000000000 +0100
+++ libstdc++-v3/src/compatibility.cc 2006-09-04 14:12:07.000000000 +0200
@@ -398,7 +398,11 @@ GLIBCXX_3.4)
#undef _List_node_base
// gcc-4.1.0
-#ifdef _GLIBCXX_LONG_DOUBLE_COMPAT
+#if defined _GLIBCXX_LONG_DOUBLE_COMPAT \
+ || (defined (__GLIBC__) \
+ && (__GLIBC__ > 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ >= 4)) \
+ && defined __LONG_DOUBLE_MATH_OPTIONAL \
+ && defined __NO_LONG_DOUBLE_MATH)
#define _GLIBCXX_MATHL_WRAPPER(name, argdecl, args, ver) \
extern "C" double \
__ ## name ## l_wrapper argdecl \
in GCC 4.2 and similar thing, but with also the whole _GLIBCXX_MATHL_WRAPPER
and its uses hunks added in GCC 4.1 could cure this (but that doesn't make it a
good idea to mix glibc 2.4+ with DFmode GCC).
--
jakub at gcc dot gnu dot org changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|WAITING |NEW
Ever Confirmed|0 |1
Last reconfirmed|0000-00-00 00:00:00 |2006-09-04 12:14:13
date| |
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28701
