GNU Go 3.5.5.

GNU Go 3.5.5 is now available at:

http://www.gnu.org/software/gnugo/devel.html.

I am a little worried about the fact that despite the
fact that there should have been no regression changes
since 3.5.5-pre-2 (yesterday), we now have an unexpected
fail in nicklas3:401.

This must have been caused by either arend_5_5.14 or 
arend_5_5.15. Is that possible?

I moved three patches to the ``may not be used''
column. All have been discussed in the list. They are:

inge_5_6.1. Separation of similarly valued moves.
splee_5_5.4. Two edge joseki patterns.
stephane_5_5.1. Connection pattern and tests.

The following patches are still under consideration:

stephane_5_5.2. Move valuation tweaks.
splee_5_5.1b. Revision of break_chain3().
readline_5_2.1a. Revision of readline_5_2.1. Readline support.
large_scale_5_1.1a. Rework of large_scale_5_1.1. Capture on a large scale.

There could be pending patches (probably from since
Mid-February) that have been overlooked. If you've
submitted a patch that is not posted on the web
page, please let us know.

Patches in 3.5.5:

html_pdf_doc_5_5.1a. HTML and PDF documentation support in doc/Makefile.
warning_5_5.1. Moves definition of TRACE_READ_RESULTS to hash.h.
arend_5_5.15. Initialize lunch data for single owl move.
arend_5_5.14. chainlinks[] in compute_eyes_pessimistic() have to be initialized
splee_5_5.2. SGF traces revised.
arend_5_5.13. Use new cache in semeai and connection reading.
arend_5_5.12. After-lunch walk.
arend_5_5.11. Komaster now a private variable in board.c.
paul_5_5.9. Revised edge_block_moves() proposes more move types.
gtp_5_5.1. Don't crash with assertion failure when `undo'ing at stackp > 0.
view_pike_5_5.1. Support for multiple engines in view.pike.
arend_5_5.10. Don't blindly trust owl attack moves against dragons in a semeai.
paul_5_3.3. Be less optimistic about large bulky eyes.
paul_5_5.7a. More aggressive recursive_disconnect2().
html_pdf_doc_5_5.1. HTML and PDF documentation support in doc/Makefile.
arend_5_5.9. Endgame test.
arend_5_5.6. Slowest regression moves.
arend_5_5.7. time_report() and the 21st century.
arend_5_5.8. 3 more tests.
paul_5_5.8. Arbiter functionality for `twogtp.pike'.
arend_5_5.5. Breakin and readconnect fixes.
paul_5_5.6. Renable accidentally disabled parts of paul_5_5_5a.
splee_5_5.5. Slightly generalized A1124a.
paul_5_5.5a. First version of Owl threat combinator.
paul_5_5.4. Pattern attributes reimplementation. Rebroken out of paul_5_5.5.
copyright_5_5.1. A few previously missed copyright notices updated.
paul_5_5.3. Move valuation patch (was listed as paul_5_5.2).
paul_5_5.2a. A few test cases from a match between 3.5.4 and 3.4.
splee_5_5.3. Seki test cases.
gunnar_5_5.6. analyze_eyegraph improved.
gunnar_5_5.5. --level can adjust min_level and max_level.
teun_5_5.2a. Revision of teun_5_5.2. Don't malloc 0 bytes.
teun_5_5.1a. Revision of teun_5_5.1. Warnings fixed.
inge_5_5.1. Debugboard mended.
teun_5_5.3. Bugfix in hashdata_init().
zm_removal_3_5.1. Memory allocation option /Zm1100 taken down in patterns.dsp.
inge_5_5.2. New cache for connection and semeai (mostly disabled for now).
gunnar_5_5.4. Owl tuning.
gunnar_5_5.3. Caching of unbranched owl nodes.
arend_5_5.4. Source code comments for heap in readconnect.c.
stephane_5_1.1a. URLs to regression views.
arend_5_5.3. Complete conversion of break-in code to new cache.
gunnar_5_5.2. Tuning, mostly owl.
gunnar_5_5.1. Test cases for 3-3 invasion.
paul_5_5.1. Yet more features for `twogtp.pike'.
paul_5_4.4. More correct SGF writing.
help_5_5.1. Write `--help' and `--version' messages to stdout, not stderr.
bugfix_5_5.1. Don't overflow in cache size computation.
arend_5_5.2. Fix TRACE_READ_RESULTS.
inge_5_4.2a. Revision of inge_5_4.2. Owl uses new cache.
arend_5_5.1. Tests from the aya game at the CGF meeting.

Dan