[Welcome to Sensei's Library!]

StartingPoints
ReferenceSection
About


Paths
KGSWishlist

Sub-pages
Fixed
Discussion

Referenced by
KGSTheKiseidoGoSe...
KGSUsersGuide
KGSFirstTimeUserE...
MetaDiscussion
KGSWishlist/Gener...
KGSBugs/Fixed
CompostHeap
KGSBugs/Discussion

 

KGS Bugs
    Keywords: Online Go

This page can be used to file bug reports about anything which is KGS related. Usually it will be the client. When filing bugs, please make sure that you include enough information so that it is possible to reproduce the bug. This is usually easiest if you explain step by step what you did, what happened, and what you expected to happen.

The [ext] Jago bug report checklist is a useful guide for filing bug reports.

Version gathering tips:

  • The CGoban2 version is displayed on the main six-button window. Also, please specify whether you are using the downloaded version or the Web Start version.
  • To find out which version of Java you are using, open an xterm or a DOS window and type 'java -version' (without the quotes) and note the build version.
  • The versions of your major operating system components may be useful, eg XP Pro SP1, Slackware Linux 9.1 kernel 2.4.22 XFree 4.2.0 KDE 3.2.

To see fixed bugs go to /Fixed.

Please, do not enter wishes here. Instead go to KGS Wishlist.


wms: Should I remove things that simply are not bugs from this page? There are a bunch. In the wishlist, I never remove things no matter what I think of them, but here maybe I should?

Cheyenne: If the bug has been fixed and is not likely to be seen again, it should be okay to remove the bug report. May want to have a section for "Bugs currently known" and a section "Working as designed" for those things that are commonly reported as a bug

wms: OK, it's been 24 hours. Here I go. :-) I removed all non-bugs, all bugs I fixed, and bugs that I'm fairly sure are caused by a user's particular system. Some of the bugs removed are not yet in any release, but I'm taking them out anyway.

rubilia: Is this referred to as the ladder ones will be fixed in future, anyway? It feels somewhat confusing to know of bugs being mentioned neither here nor at /Fixed, although still appearing. Supposedly, you don't want us to report them again?

wms: Yes, I probably should have put them in the "fixed" file. Sorry. If you know a bug was here before, please do not re-file, either I don't consider it a bug (in which case it belongs in the wishlist) or else I have already fixed it.


Barthoze: To help beta testing of future versions, I'm preparing a KGS features and behaviour list, that also gives the expectable behaviour of the client/server.


If anyone has an idea how to put order into this page, it would be appreciated :)


  • When trying to download a forfeited game, you get the unfinished one. It is possible to load a forfeited game in a room though.
    • wms: What does this even mean? You get the unfinished game? What other one is there?
  • wms: Sometimes at the end of a game, comments get screwed up. This is because when the game ends, players are sent the comments again, so that they will see what observers said. The client tries to "hide" these new comments, but sometimes it fails and you see them out of order or duplicated.
  • Harleqin: When I successively open a bunch of sgf files, the game and the edit window will be a bit smaller than before each time. So, after a few files I will have to re-enlargen the windows. This is perhaps just an annoyance, but there might be an easy fix :). I am using Java 1.4.2 and KDE 2 on XFree86 4.2.1. I have not observed a similar behaviour with other java applications.
    • wms: Not an easy fix. It's a java bug. Long explanation here removed - trust me, it is too hard to fix this for me to do it.
    • os?: Let me try to explain the algorithm that the previous contributor to the discussion may have had in mind (the problem being that the window comes up differently than desired).

      The size error Java makes when creating the new window can be calculated by comparing the getsize() results of the old and the new window, and the window sizes can then be adjusted accordingly. In pseudocode:
      • Previous run
        (x, y) := getsize()
      • Next run
        create_window(x, y)
        (xx, yy) := getsize()
        increase_window_size(x-xx, y-yy)
        // alternatively: set_window_size(x+(x-xx), y+(y-yy))
    • TimK: If you get the window size at shutdown time, and restore the window to that size on startup, the window shrinks because the APIs are lying to you, right? What if you only changed the size that is saved between sessions when the user resizes the window? Then the first time a restart happens, the window would shrink a bit, but as long as the user didn't resize the window during that session, the window size would be the same the next time the app started.
  • mgoetze: When looking at a game on KGS, if I have my mouse over the "rewind to beginning of game" button, then start autoplay (with the keyboard), then click the mouse, it rewinds to the beginning of the game even though it now looks like the mouse is over the "stop autoplay" button. (Easy to work around, but potentially annoying/confusing.)
  • Reuven - This happens when I use the java web start ver under the latest java on win2k - At times (I can't reproduce it at will - Seems a bit random - But does seem to occure often when oppenning windows of deactivated games) my whole screen gets grey, and the rest(bars) are reversed... Closing doesn't always do the trick, restart is nessaccery at times... If u need any further details..
    • wms: Sounds like a bug in your java implementation, your OS, etc...probably not fixable by me, but if you find something that makes it happen every time, then I may be able to add a workaround.
  • When watching a game on KGS and you browse back a number of moves, the time counter will continue to go down. It either should not do this, or should stop at the time when the next move was actually played.
  • RBerenguel: If in a demonstration game you create a ko, and then erase all stones, the little square mark doesn't disappear. (That's just a minor issue, the whole program is superb I think)
  • mgoetze: When cloning a game with freely placed handicap stones (NZ Rules), japanese-style handicap stones appear in addition to the freely placed ones.
  • WarEagle A minor bug, or at least non-obvious design feature. When browsing back in an open game, the move counter in both the game window and KGS room counts backward. I would have thought that the KGS room information would have reflected the current game state, while the game window reflects the move number shown.
    • I don't really think that counts as a bug. The room information is displaying (for each game) what board position you can currently observe. It's more of a feature.
  • misterb@KGS: Using Java 1.4.2, Windows XP Pro SP1 and CGOBan 2 2.5.7 (The version that you download but keeps somekind of Java Schedule Process to download any upgrades). The problem that I encountered was with 2 things. First of all. Changing font-size. It said that I had to log out and re-login to make it work. When in fact I had to close down the entire program. Just annoying and confusing. Either it should say "Restart the program" or make it re-read the settings. Secondly, the buddies list. I have several users on this machine. One for the Go-Club and one for myself. However, both users share the buddylist. Which they shouldn't.
    • savra?: Same nasty problem here. Maybe it should be in the wishlist, because it's a missing feature and not a bug?
  • Small bug. When you finish a game, unless you close down the window the system thinks you are still playing. Therefore you cannot challenge a new opponent.
    • Hu: The subtlety of the UI issue is that for the majority of usage, it needs to proceed simply and fluidly, but additionally, for craftwork usage, it needs to be able to do sophisticated production with minimal extra motion. I think that automatically spawning extra nodes would be better than the current situation in the circumstances yoyoma illustrates, but with the reservation that he indicates. It could be the regular mode of interaction, but a craftworker could have available an extra placement tool (edit tool) mode where stones would be added to the existing placement node. In the new UI, it could be triggered by a ctrl-F2 or holding control down while selecting the Edit tool item, or both.

Ansgar?: cgoban doesn't handle the AB/AW properties correctly. Try the following sgf:

 (;FF[4]GM[1]AP[Firefox:0.8]SZ[9]AB[aa:ac];AW[ba:bd][ad])

In the second node, the black stones in the corner are removed, although the sgf standard says: "Adding a stone doesn't make any prisoners nor any other captures (e.g. suicide). Thus it's possible to create illegal board positions."

  • wms: Right you are. CGoban 2 is incapable of having boards with zero-liberty stones. When CGoban loads an SGF file that would require this, it inserts "AE" properties as needed. This is one of the places where CGoban 2 does leave the SGF standard - there are one or two other places, always in the form of standards-compliant files that KGS does not show correctly. I do try very hard to make sure that any file saved by KGS is 100% SGF standard compliant, but the standard has a few places where the work to fully support the standard vs. how often the feature is used seemed a poor tradeoff. In this case, it wouldn't have been super hard to support the standard, except that the standard is incomplete in this case and I didn't even know how to meet the standard! The problem is this: Let's say you have AB/AW that creates zero-liberty stones still on the board. This is no longer a valid go position, but the SGF spec says it should be shown anyway. What happens when a B[] or W[] property shows up? When a move is made on a board, do you remove all zero-liberty stones on the board, or do you only remove the stones that have been reduced to zero liberties? For that matter, does it even make sense to have B/W properties when the board is not legal? Is it possible to make a legal move on an illegal board position? Anyway, I hope you see my point. Anybody can sit there and say, "It should work like thus-and-so," but that doesn't help - if you do a lot of work to fit a standard, the work is only helpful if you end up fitting the standard precisely and unambiguously; I could not find a way to do so in this case, so I chose to change the file to make it unambiguous in how it is displayed and manipulated by SGF viewers/editors.
    • Ansgar?: In my opinion the the standard isn't incomplete here. B/W properties should simply be executed, even when illegal. So I guess you can make illegal moves on an illegal board, too. After a move, only groups adjacent to the placed stone and the group the stone itself belongs to should be captured.
    • wms: PS - The other place where KGS doesn't fit the standard that I can think of now is in SGF files that game game setup properties that aren't in the root node. It is legal to have an SGF file that combines 9x9 and 19x19 games, or handicap-3 and handicap-5 games, or whatever, by moving the SZ[] or HA[] into variations. Like zero-liberty AB/AW stones, this raises a lot of ambiguity issues; what if a B[]/W[] property would capture a stone were the board 9x9, but wouldn't capture if the board is 19x19, and different variations from this position have different board size - what to do then? So, similar to the AB/AW leaving dead stones on the board, I decided that it just isn't possible to sensibly implement this feature, and CGoban will refuse to load such files.
    • Ansgar?: SZ[] must only appear in the root node, so the size of the board is never ambigious. As for HA[]... the handicap stones should be added with AB[] within the same node, but then they will not show up from the beginning. So I don't see a see a sensible way to support this either.
    • wms: PPS - If somebody were to argue that it is wrong for CGoban to silently add "AE" properties to erase the zero-liberty stones, I can see the point. Perhaps I should just print a message and refuse to load the file?
    • wms: PPPS - So firefox now includes an SGF editor? News to me, I'm using firefox right now, but I never found the SGF portion of it! :-)
      • Ansgar?: Well, I was using firefox to create the sgf :-)
    • IanDavis: Am I the only one who doesn't get this? AB/W are usually used to place handicap stones in normal games - so why shouldn't they be captured. Obviously the designer must have intended them to only be used for diagrams, in which case you have to wonder why they are usually described as stones.
      • Ansgar?: Placed stones may be captured by later _moves_, but not by other setup stones.
    • axd: Talking about the SGF standard, I'd like to draw your attention to a remark I made in KGSWishlist/FileHandling about cGoban's misuse of the circle marker.
      • IanDavis: Do we need misuse in bold here? I actually quite like the optional last move marking by using a circle. However should I get tired of it I can just turn it off, like the board coordinates. I don't see the point of wms creating a seperate token for this mark up that would just be banal
  • rubilia: I haven't checked this from the other side yet, but yesterday, an opponent of mine apparently still saw the result of our actual game, although the variation I had edited resulted in a very different score.
    • MarkD: I have seen the same error. The score was wrong after the game, when the commentator pressed "edit" and added a test stone half of the people were seeing the correct score, the other half were seeing the wrong numbers. Both groups were using the current client version.
    • wms: Will be fixed in the next release. Only the teacher sees the score update properly when working with demo or review games.
  • sni: On the KGS archive webpage, the filenames are often linked wrong, when the same game-pairings appear more than once on a page. The filename should be incremented, like for example player1-player2-3.sgf for the 3rd game.
    • wms: Can you give a specific example? I just checked (see [ext] http://kgs.kiseido.com/en_US/gameArchives.jsp?user=imaginarie&month=8&year=2004 ), the ones I looked at properly increment the revision code.
    • sni: Sure. For example [ext] http://kgs.kiseido.com/de_DE/gameArchives.jsp?user=germanchmp there is the first and the third game from the top with the same filename, since they are in different directories on the server its not that easy?!
    • wms: Ah. So they aren't linked wrong at all, as the original comment stated - each link leads to a separate file as they should. They simply have the same name, but they are in different directories. This is not a bug, the "revision code" of a file is only there to separate files on the same day, not all files ever. I'll take these comments down in a few days, meanwhile sni if you want to request to have this changed just add it to one of the wishlist pages, although I'm pretty happy as things are so I doubt that I'll make a change like this.
  • wms: Got this in email: When you are spectating a game (say the game is 100 moves long) and you are looking at a previous move (say move 50), and a player who is playing the game gets an undo, then when you press forward to the end of the game, it will show the move that the player played before the undo (the move that was taken back), and it will also not grey out the "next move" button, even though there is nothing after that. you can click next many times, and it will switch back and forth between the undone move (old move 100), and the board state after the undo was done (move 99). After a player moves though (new move 100), it will let you cycle to the end of the game and normal functionality returns.
  • mgoetze: On OS X with Webstart: When you enter a game where both players have a picture, then try to enlarge both pictures, you will always get the one you clicked on first. I was unable to reproduce this under Win2k?.
  • AshleyF: SGF representing the game has undone moves as the main line and the actual main line of the game as a variation. It makes it a bit of a pain to run the game through tools such as SmartGo's Analyze Game which normally only follows the main line - it stops short at the undone move.
  • RobAnybody: F6 and F8 don't work as shortcuts while editing a SGF file (possibly also in a demonstration game). They are supposed to switch the current tool to 'circle' and 'number' tools.
    • wms: Works for me. I just tried it.
      • RobAnybody I reconfirm, here are some details: CGoban v2.6.5, win xp, java -version on the command line gives the following:
 java version "1.4.2_04"
 Java(TM) 2 Runtime Environment, Standard Edition (build 1.4.2_04-b05)
 Java HotSpot?(TM) Client VM (build 1.4.2_04-b05, mixed mode)

I also noticed that F6 brings the 'Help' button into focus (a magenta rectangle appears around the word Help). F8 has no visible effect. If there's anything else I could do to see what causes that situation - write on my homepage lest I forget to check this page.

  • pwaldron?: Open a user's game list and pick a game that is currently open in a room (i.e. one with spectators). Try and save the game to disk. An error message comes back.
    • wms: Are you sure that it isn't a case where the game simply is still in progress? What is the error that you saw?
      • pwaldron?: Unless I misunderstand you, I don't think the game was in progress--I had just resigned it and left the game, as had my opponent (still spectators left though). I decided afterwards that I'd like to get a copy of the game record, so I opened up the game list from my user data window, and selected "Copy to disk". In the game list the game was boldfaced still; does that mean it's in progress (even though the game was over?). The error was a Java IO exception of some sort; I sent you an email with the exact phrase.

  • Rengobot scoring is broken, stones marked as dead don't contibute to the score. Workaround is to use SE.
  • Room Owners - try changing your room type from open to private, you'll find you have to log in and out again to be able to use the edit button anymore.

  • Chris Hayashida: When creating SGF files, if you use "Save As" to save a newly created file, it still prompts to save on close.


This is a copy of the living page "KGS Bugs" at Sensei's Library.
(OC) 2004 the Authors, published under the OpenContent License V1.0.