Strangeness for the record

[Loren James Rittle]

Although I can't find the reference at the moment, I can confirm that
I too now see this failure (again) on i386-unknown-freebsd3.4:

-r 9    0.000   14123   1952    347047  27_io/filebuf_members.cc

[Exact failure occurs on line 58: non-zero return value from close()
 since file descriptor was already closed.  I saw problem last week.]

that had disappeared in my report for the same port yesterday (see
http://gcc.gnu.org/ml/libstdc++/2000-11/msg00429.html ).

This is very strange since I *know* with certainty that these are the
only changes in libstdc++-v3 that I picked up since a build that
didn't display that error:

2000-11-28  Benjamin Kosnik  <bkoz@purist.soma.redhat.com>

        Update for 2.91 release.
        * include/bits/c++config (__GLIBCPP__): Update.
        * docs/17_intro/RELEASE-NOTES: Update.
        * docs/status.html: Update, remove bogus autotools disinformation.
        
2000-11-27  Benjamin Kosnik  <bkoz@redhat.com>

        * tests_flags.in (LTCXX): Change CXX_FLAG to CXXFLAGS.

        * testsuite/22_locale/ctor_copy_dtor.cc: Adjust formatting.
        * testsuite/27_io/istream_extractor_arith.cc (test11): New
        testcase, based on libstdc++/90.

Of course, I now have to wonder if I somehow changed configuration or
if there is a subtle difference between a complete `make bootstrap'
and the rebuilding support libraries during a later reconfigure/make
(which I may have done to get the results I posted yesterday.