Fri 04 Mar 2011 09:59:34 PM UTC, comment #8:
Fix for crash is included in the latest Windows build.
|
Wed 23 Feb 2011 12:13:46 PM UTC, comment #7:
Thanks again, Philippe. I don't think it's a problem that the display now shows it as a double/no double decision, since this is how many players prefer to view such problems since it's actually more challenging to make BOTH player's decision. But if no analysis is present, I agree that changing the Save logic at some point is indicated. But it's not that big a deal to press the Hint key.
|
Tue 22 Feb 2011 11:22:27 PM UTC, comment #6:
The fix is at the open.
Basically, old logic was : if this is a take or pass, use the analysis of the corresponding double and discard what was just read. Except that in the case of a single take or pass there is no preceding double record.
Now, it uses the preceding double if it exists. If not, what was read is not discarded.
The old files can be opened with the fixed version.
Doing something at the save stage (inserting a double record) would be useful since older versions of gnubg would not crash on the file and the display would be (I think) more accurate, with the offered cube shown on the inner board with the right value.
|
Tue 22 Feb 2011 07:46:36 PM UTC, comment #5:
By the way, was the fix to the Save or Open logic? I'm wondering if I should purge all the files that can't currently be opened.
P.S. Sorry for spelling your name wrong. I've seen it so often, I didn't think to double check.
|
Fri 18 Feb 2011 08:45:28 PM UTC, comment #4:
Thanks for the quick work, Phillipe. And your comment about the "go to 1st game" allowed me to update my enhancement requests with better information.
But is it possible that your fix took the position back one step too far? When I use my workaround of saving the double/no double decision, I see the cube offer immediately after I open the position (assuming "go to 1st game" is unchecked).
|
Mon 14 Feb 2011 09:31:12 PM UTC, comment #3:
The crash should be fixed (in the source repository, I don't know when a XP build will be available).
Some quirks remain :
- the display (the position in the move records, really) depends on the value of Settings|Options|Other|Goto first game when loading, which doesn't make much sense for a single position
- the cube should be displayed on the board with the value offered, not on the side with its previous value. Currently, the board looks like it is the double/no double decision which is somewhat misleading.
I intend to look after these, but not necessarily soon.
|
Tue 01 Feb 2011 11:36:44 AM UTC, comment #2:
Thanks for the quick work! If it's not too tricky, it would be best not to back up a half-move before saving the position because I think it will lead to an inconsistent display when viewing positions. For instance, people typically press "Page Up" after opening a position to see what things looked like BEFORE the move/cube decision was made. If you back up a half-move before saving, I think this will make "Page Up" produce unwanted results.
|
Fri 28 Jan 2011 09:46:00 PM UTC, comment #1:
I can reproduce this with the last Windows build (assertion failure at play.c:586).
With a unix build, there is a less brutal failure. The position is loaded as a double/no double decision with (apparently) the taker on roll.
We should probably either save the doubling decision record instead (as suggested by Tom) or both the double and the following take or drop. I think that would do the right thing at least in the unix case.
(file #22551)
|
Fri 28 Jan 2011 11:05:52 AM UTC, original submission:
When you try to view a saved position that was a take/drop decision, GnuGB abends with a "sanity check" error. The workaround is to always save the DOUBLER'S position rather than that of the cube taker. This may be the easiest way to fix the code - always go back a half-move before saving. Also, when there's a failure to open an SGF file, it would be nice if the program didn't abend.
|