1999-08-04 libstdc++2.90.6

- testsuite/27_io/istream_extractors_char.cc: 
  Failing on all platforms with -O2, working with -O. Need to look at this.

- _GLIBCPP_USE_CTYPE_ISBIT complete and utter hack to get basic ctype
functionality, such as isspace(char) working on x86/linux. Currently,
this does not work on any other non-linux (non-x86?) host. This needs
to be re-architected, to say the least.

- _GLIBCPP_HAS_BUILTIN_SINF: Need to think of some way to make
bits/std_cmath.h actually take into account the complier intrinsics,
without leading to unresolved symbols at link time. Ideally, this
would take place at configure time.

- bits/std_cwchar.h: Need to figure out what wide string functions are
around at configure time, not with ugly macro hacks.

- there may be one set of remaining string bugs, dependant on final
clarification of the string::find technicalities when finding in an
empty string or using an empty string for an argument. At the very
least, v-3 has interpreted the standard in a way that is in opposition
to other libraries on other platforms.

- trigraphs and keywords a la the iso646 header are not correctly
implemented. It looks like the compiler recognizes them as keywords
but then doesn't translate into the correct bit ops. It is a mystery.

- ostream inserters for floating point types is known to be faulty.

- istream::get(char) is not working for cin: See
http://sourceware.cygnus.com/ml/libstdc++/1999-q3/msg00060.html

- wide strings have not been tested, and may therefore be unusable.

- Chapter 27 io functionality is not finished. As such, there are
known bugs in: ios_base::failure, basic_ios::callbacks, basic_ios::
fmtflags manipulators, basic_streambuf, basic_istream, basic_ostream,
basic_stringbuf, basic_istringstream, and basic_filebuf.

- Many facet implementations are stubs. (22) 

- Almost no optimizations for small-footprint/low-overhead. (22,27) 

- It has not been made thread-safe. 

- There has been some work to wrap the C headers in namespace std::, but 
   it may not be complete yet, and C macros are not shadowed. Please consult 
   the mailing list archives for more information. 


