The mainline still has __builtin_fsqrt, where the 3.0 branch changed to __builtin_sqrt. This sort of regression from branch to mainline should be avoided - but has been present since the 28th of May. http://gcc.gnu.org/ml/gcc-patches/2001-06/msg00394.html Release: 3.1 20010609 (experimental) Environment: System: Linux digraph 2.2.19 #1 Wed Mar 28 16:01:38 UTC 2001 i686 unknown Architecture: i686 host: i686-pc-linux-gnu build: i686-pc-linux-gnu target: i686-pc-linux-gnu configured with: ../gcc-cvs/configure --prefix=/opt/gcc/mainline --disable-shared --enable-threads=posix --with-system-zlib How-To-Repeat: Work on the GCC manual and observe a bogus difference branch-to-mainline, because this patch (including the doc changes) has not been applied to the mainline.
Fix: Adapt http://gcc.gnu.org/ml/gcc-patches/2001-05/msg01939.html for the mainline and apply it.
Responsible-Changed-From-To: unassigned->gdr Responsible-Changed-Why: You applied the patch to branch only.
State-Changed-From-To: open->analyzed State-Changed-Why: Confirming my PR. This is a regression from 3.0 to the mainline.
State-Changed-From-To: analyzed->closed State-Changed-Why: This bug has been fixed by a patch from Roger that has just been committed to CVS, Thanks for the report, Andreas 2001-12-14 Roger Sayle <roger@eyesopen.com> * builtins.def: Rename BUILT_IN_FSQRT to BUILT_IN_SQRT and __builtin_fsqrt to __builtin_sqrt. * builtins.c (expand_builtin_mathfn,expand_builtin): Same. * doc/extend.texi: Simplify documentation to match patch. * f/com-rt.def: Use __builtin_sqrt instead of __builtin_fsqrt. * f/com.c (ffecom_init_0): Same, and fixed enumeration usage. * libstdc++-v3/acconfig.h: Test for __builtin_sqrt instead of __builtin_fsqrt. * libstdc++-v3/acinclude.m4: Same. * libstdc++-v3/include/c_shadow/bits/std_cmath.h: Same.