Your re/move of ext/hash_map causes a regression

[Hans-Peter Nilsson <hans-peter.nilsson@axis.com>]

> Date: Thu, 11 Oct 2007 16:37:45 -0500
> From: Benjamin Kosnik <bkoz@redhat.com>

> Converts existing uses of hash containers to TR1 containers, as the
> first step in deprecating ext/hash_map, ext/hash_set.

When you (re)moved ext/hash_map, you forgot to adjust the g++
testsuite.  I see a regression for cris-elf (likely everywhere)
between 129434 and 129442:

Running /tmp/hpautotest-gcc1/gcc/gcc/testsuite/g++.old-deja/old-deja.exp ...
FAIL: g++.old-deja/g++.robertl/eb130.C (test for excess errors)

with g++.log:

/tmp/hpautotest-gcc1/gcc/gcc/testsuite/g++.old-deja/g++.robertl/eb130.C:9: error: ext/hash_set: No such file or directory
/tmp/hpautotest-gcc1/gcc/gcc/testsuite/g++.old-deja/g++.robertl/eb130.C:10: error: '__gnu_cxx' has not been declared
/tmp/hpautotest-gcc1/gcc/gcc/testsuite/g++.old-deja/g++.robertl/eb130.C:10: error: expected constructor, destructor, or type conversion before '<' token

The non-comment parts of eb130 are short enough to include:

#include <ext/hash_set>
__gnu_cxx::hash_set<int> foo;

Dunno what's the right fix here; <backward/hash_set> maybe.

brgds, H-P