http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45841
--- Comment #22 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2010-10-05 17:03:09 UTC --- Just a heads-up regarding issue #3. (In reply to comment #19) > Apparently > reading after a write at EOF is not in the tests. Hm, doesn't 27_io/basic_filebuf/sputbackc/char/9425.cc test something like that, or at leas EOF after? It doesn't fail for you? It does for me (cris-axis-elf), seen as a regression with a fixed simulator. Or maybe that's not the reason? Ok, I'll look closer. FWIW, due to <http://subversion.tigris.org/issues/show_bug.cgi?id=2333> I also saw 27_io/basic_filebuf/sync/wchar_t/1.cc (a file removed in r164529) as regressing, which had me confused until I STFW and found that svn up to 1.7.0 (version according to the bug report, but I can confirm it includes 1.6.9) did not include (files in) deleted directories in diffs, like "svn diff -c164529". Note that sync/char/1.cc is included. Looks like lots of libstdc++ file I/O test-cases are missing "// { dg-require-fileio "" }" lines. I'll add them as obvious.