status and priorities

[Benjamin Kosnik <bkoz at redhat dot com>]

At the beginning of the year, I proposed some development priorities.
For reference, this message (and follow-ups) can be found here:

http://gcc.gnu.org/ml/libstdc++/2003-01/msg00003.html

First, I'm pleased to report that many of these items have already been
done! The progress has been fast and furious, and I think everybody can
be truly pleased with the progress. A lot of hard work by many people
has made a noticeable improvement. Some issues remain, however.

Paolo has suggested it might make sense to revisit this list, update,
and prioritize. A great idea. At the bottom of this email is such an
attempt, comments and suggestions, as well as volunteers, are welcome.

Please note that the following list is for mainline development, which
may or may not be the library used in gcc-3.4, depending on ABI
decisions made by the gcc SC. To clarify, mainline libstdc++ is assumed
to be the library used with -fabi-version=2, and the latest stable
release branch is assumed to be the library used with -fabi-version=1.
When ABI decisions are made by the SC I'm assuming this list will be
notified, till then it's pointless to speculate on what/should happen.
Development efforts should focus on mainline at the moment.


optimization

- Paolo/Nathan work on filebuf structure and optimizing writes

- Jerry/bkoz work on caching locales. (I expect for this to be mostly
  done by the beginning of next week.)

- Jerry(?)/Paolo(?) work on __float_to_char.

-  relocations, work on getting these down

-  Jerry Quinn
    issues with __builtin_memcpy and std::copy
    http://gcc.gnu.org/ml/libstdc++/2003-02/msg00076.html

-  look ahead for unbuffered io, so know when multiple putc's can be
   coalesced.

-  unlocked __basic_file + new mutext class


debug

(high)
- Apple work on debug modes in STL (very high priority)


configuration/make/build

(high)
- looking at the  _GLIBCXX_USE_WCHAR_T issues for HPUX, cygwin, and Solaris

- dependency tracking for includes

(high)
- moving to current auto tools


io    

(high)
- wchar_t fixups variable encoding work (working wfilebuf).
  including wchar_t positioning and seeking.

(high)
- 27_io testsuite for wchar_t


locale

-  icu integration as ext facets? ways of integrating xerces into locales.

-  solaris named facets work (unresolved, lack of current sun hardware issues)
  
-  generic named facets work (Loren were you working on this?)


string

(long-term)
- solution for COW vs. deep copy issues, MT scalability


allocator

(medium)
- perhaps find a better way of selecting default allocators.

- move to mt_alloc for --enable-threads=yes type configs?

-  need persistent, extern allocators


docs

- this seems to be progressing nicely, but perhaps Phil and Jonathan
will have some ideas. One thing I will note, is that if there was some
kind of information on how to mark stuff up (doxygen style guide, etc)
it might make it less intimidating. Although, I suppose it's not so hard
now, given that so much of this is doxygenated already. Thoughts?


testsuite

-  valgrind hooks into make check so can tell memory leakage
  Some commentary on the valgrind users list

-  add hooks for qmtest for visual diffs 

- automatic testing of interactive tests

- diffing generated output files

- make check-abi needs to have full symbol checking. Scope the LSB
testsuite, see what's going on with the typeinfo etc. bits.


include

(medium)
- get glibc-2.3 std/c99 namespaces working under linux for "C" compat


compiler-side

(high)
- compression for wide versions of basic types

(high)
- wider use of the fixed demangler