[2.90.9] planning for the next release

[Benjamin Kosnik <bkoz at cygnus dot com>]

In an ongoing effort to make libstdc++ development easier, I thought
I'd post some tentative ideas about what might be attempted for the
next snapshot. . . here are some things that are outstanding. OK, a
lot of things that are outstanding. Feel free to laugh outright, or
add bits as the situation warrants. If I've missed especially critical
bits, please let me know. If you plan on working on a specific item,
please pipe up. I suppose this stuff should be in the TODO list...

1) libio merge
  - merge v3/libio into Uli's current tree
  - add iconv stuff
  - hack around it for hosts that don't support it... or something??
    generic layer?
  - wchar_t?
  - -fnew-abi vs. libio
 
2) basic_string<wchar_t>
  - take 21_string tests and port to wstring?
  - conformance issues

3) complex merge / unify with Uli

4) SGI stl merge, addition of testsuites
  - ?? to be determined
  - absolutely positively need a volunteer

5) linker research
  -  ie get -ffunction-sections -fvtable-gc -Wl,--gc-sections working??

6) host-related ctype problems (pending resolution of compiler problems)
  - hpux11
  - irix6.5?
  - aix 4.3, aix 4.3

7) C/C++ sync_with_stdio problems
  - sss and his c++/write/c++ bit getting lost

  - redirecting stdin/stdout/stderr--how to test for this?
  http://sourceware.cygnus.com/ml/libstdc++/2000-q1/msg00022.html

  - I believe there is still one outstanding sss bug from christmas

8) filebuf::putbackfail -- make standards-conformant

9) config work
  - flatten directory structure by removing the stl dir?
  - fixincludes for math
  - fix problems building out of source directory
  - fix problems with 
	--prefix=/usr/local screwing up and including things like "exception"
  - think of a more intelligent way to organize the ctype stuff 
    (reduce the duplication)
  - thread-safe stuff
    - ?? work on the c++threads.h interface ie config/threads-posix.h
    - ?? atomicity cleanups
    - ?? maybe do some testing of the current MT implementation?