C++ regression between 107732 and 107796

[Geoff Keating <geoffk at geoffk dot org>]

The new regression tester isn't up to sending mail yet, but it did  
notice a regression introduced between

Date: 2005-11-30T19:35:49Z
Revision: 107732
URL: svn://gcc.gnu.org/svn/gcc/trunk
Status: regress-2

and

Date: 2005-12-01T03:01:06Z
Revision: 107796
URL: svn://gcc.gnu.org/svn/gcc/trunk
Status: regress-4

(Actually, two regressions, one of the Java tests seems to be failing  
intermittently.)

The other new regression is:

libstdc++.sum ext/pb_assoc/example/tree_intervals.cc

It appears it can only be due to one of these:

r107733 | reichelt | 2005-11-30 11:59:50 -0800 (Wed, 30 Nov 2005) |  
15 lines

        * tree.h (CASE_FLT_FN, CASE_INT_FN): New macros.
        * builtins.c (expand_builtin_mathfn): Use them.
        (expand_builtin_mathfn_2): Likewise.
        (expand_builtin_mathfn_3): Likewise.
        (expand_builtin_int_roundingfn): Likewise.
        (expand_builtin): Likewise.
        (integer_valued_real_p): Likewise.
        (fold_builtin_int_roundingfn): Likewise.
        (fold_builtin_bitop): Likewise.
        (fold_builtin_logarithm): Likewise.
        (fold_builtin_1): Likewise.
        * convert.c (convert_to_integer): Likewise.
        * fold-const.c (negate_mathfn_p): Likewise.
        (tree_expr_nonnegative_p): Likewise.

------------------------------------------------------------------------
r107737 | rguenth | 2005-11-30 12:55:41 -0800 (Wed, 30 Nov 2005) | 8  
lines

2005-11-30  Richard Guenther  <rguenther@suse.de>

        PR tree-optimization/22501
        * tree-ssa-forwprop.c (forward_propagate_addr_expr_1): New
        function split out from ...
        (forward_propagate_addr_expr): ... here.  Use it to propagate
        ADDR_EXPRs to all uses.

------------------------------------------------------------------------
r107738 | jason | 2005-11-30 12:58:27 -0800 (Wed, 30 Nov 2005) | 4 lines

        PR c++/21123
        * cp-gimplify.c (cp_genericize_r): Don't dereference  
invisible reference
        parms in a thunk.

------------------------------------------------------------------------
r107762 | ghazi | 2005-11-30 18:31:49 -0800 (Wed, 30 Nov 2005) | 5 lines

        PR middle-end/25158
        * builtins.c (fold_builtin_fputs): Defer check for missing
        replacement functions.


The error is

/Users/regress/tbox/svn-gcc/libstdc++-v3/testsuite/ext/pb_assoc/ 
example/tree_intervals.cc:217: failed assertion `overlaps(c, interval 
(145, 165)) == true'

You can find the full log file at <http://gcc.gnu.org/regtest/HEAD/>  
with all the other regression tester stuff.

I think the patch that most likely caused it is 107738, followed by  
107737, 107733, and least likely 107762, since this testcase doesn't  
seem to use any math builtins and certainly not fputs.  It does use  
lots of C++ references.

Attachment: smime.p7s
Description: S/MIME cryptographic signature