What to do if you find a bug!

Please email me botlog.txt and a description of the problem and what was going on when the problem occurred!

Please DO NOT telegram me about it because I do not want to drop everything I'm doing to fix the at that moment, and if you send me an email then I will be sure to remember to look into the problem later!

Or, you can now submit bug reports to the Turtleflight Forum!

Known bugs:

• If an error occurs when reloading macros (due to a syntax error in the macros file), the program may become unstable. (The workaround is to restart the program if an error occurs when reloading macros.)
• The $nextnear function sometimes returns a blank string. (The workaround is to use the $nearby function instead.)
•  If two instances are started at the same time, the Delete button on the Instances window may get grayed out improperly. (The workaround is to use the Connection/Delete Instance menu item,or better yet don't try to start two instances at exactly the same time!)
• Error messages don't always properly indicate the point in the input string where the error occurred. In particular, if the error is within a macro, then the indicator will show an erroneous error location.
• The timestamp macro (that replaced the function version when I added the @eventtime function) isn't entirely accurate. Better to use @eventtime instead, anyway, since @timestamp is only accurate to the second. I'm keeping @timestamp around for backward compatibility only.
• A very rare and puzzling bug has something to do with the log window's RichEdit component. I haven't seen it for many versions. The program crashes with a "RichEdit error" message. Please let me know if you experience this, and the circumstances. If possible email your botlog.txt file too. :) NOTE: I suspect this may have something to do with conflict between threads. Update 6/3/02: I haven't seen this in a long time, I think it went away. :)
• Flicker: I've tried to reduce the flicker that occurs when the log window gets full (1000 lines) and lines are removed from the top. The Actions/Advanced/Anti-Flicker menu item toggles this. However when anti-flicker is on, sometime garbage will be left in the log window from overlapping windows that are closed just at the same moment that the log window scrolls. It's ugly, but harmless.
• Scroll Lock: this works, but somewhat awkwardly. If you click anywhere in the log window after turning on scroll lock, it effectively cancels it. An alternative is to instead use Chat/Mute Log which greys out the log window and stops new chat from appearing.  New chat will still be sent to the log file if it is opened.

• The @nextnum and @nextnum_ functions will return the wrong number if they are called while the global variable list window is opened, and if the destination list name comes alphabetically before the source list name. (This is because the global variable list is automatically sorted when the variable list window is open, so a new destination list item gets inserted before the most recent index point, making the index incorrect.)

• The program would sometimes locks up if the user tried to flush the Action queue.
  

• When run continuously over a period of days, Magsbot will slowly use more and more memory. Restarting it every few days is the workaround. This is a major bug, yes, and I'm trying to fix it as soon as possible. Memory leaks are notoriously tough to pin down though. (Update: as of version 4.7 Magsbot uses a new memory manager, and the memory leak seems to be much less, but I'm not sure it's entirely gone...still testing. It may be that the memory loss only occurs after a disconnection. At any rate, Magsbot can run at least for a week or more without a restart.) Update: I believe this problem no longer exists, practically speaking. Only very tiny amounts of memory may still get "lost" under rare circumstances, such as forceably closing a thread that was started using the THREAD {} clause, by using the THREADEND command. In that case, the variables in the thread's local variable list may not get freed.
  

• Starting a new instance with AutoEnter checked, when there is already another instance running, will cause an error message and the new instance will fail to enter the world, because the Location tab gets cleared (no world name) when it starts up. (The workaround is to start new instances with AutoEnter unchecked, then type the desired world name and coordinates and click Set to enter the world.) (Fixed in 4.0 b8.)
• Client list sometimes gets blank rows in it.
• If the same IP is connected more than once as a client, the IP in the client window should not be grayed until all connections from that IP are closed.
• @wordinlist and macros derived from it should not have been case-sensitive. Fixed in 4.0 b3.
• An error occurs if  the behavior table is cleared completely, then the user attempts to paste rows into it. (The workaround is to add a blank row before pasting, then delete the blank row afterward.) Fixed in 4.0 a47.
• The wrong instance could have its status button grayed out when an instance failed to enter a world. Fixed in 4.0 a45.
• The wrong variable list was being passed to the FTN command, in version 4.0 a44. That's fixed in 4.0 a45.
• Concatenation of strings in a proc name argument (e.g. in CLICKBTN) wasn't working right, so with a command like CLICKBTN $a+$b the $b part would be lost.
• List index error when Goto Bookmark, fixed.
• SAVELISTTOPD wasn't working right, sometimes saved 0 lines to file. Fixed 1/26/03.
• EJECTADD, etc commands weren't working right. Fixed 1/26/03.
• The alias system didn't work right if no aliases were declared for a particular bot name. Alias are now handled differently, by defining global variables (Alias buttons on Actions panel, main tab) instead of the userdefs.udf file.
• AWCoords button didn't translate altitude properly.
• DELETEINSTANCE deleted the wrong instance under certain circumstances.
• Sometimes the wrong instance got greyed on the Instance List when an instance failed to enter a world.
• An error could occur if Magsbot was closed or a button or behavior table entry deleted while code in the button or in the behavior table entry was still executing.

Fixed in 3.0:

• Fixed the arrow icon on survey dialog (it looked like a diamond on some computers, depending on the fonts installed.)
• bug: saving a button category changed the default button file path
• Fixed the distance macros @dist, etc.
• bug: $fmt wouldn't take constant strings as arguments
• bug: @uns wasn't working right
• bug: @btncount_[@tab] didn't return correct number
• The Nearby List only showed one entry for multiple avatars that have the same name. That could only happen with bots, since citizens can't have the same name, of course. But the bug was fixed and some new functions were added to allow avatars to be selected by session number, if multiple avs of the same name are present.  (Selecting avatars by their session number instead of name is actually the normal way the SDK does it, but I originally used selection by name because I thought it would be simpler for users. You can still do it that way, but bear in mind that it is possible for there to be several avs with the same name in an area, if they are bots.)
• bug: clicking a button highlights it without unhighlighting previously selected button. fixed.

• Reloading the macros file was resulting in the program crashing. I believe this is fixed now, but if you experience any strange errors in Magsbot, it may be this bug. Workaround: if you make changes to the macros file and need to reload it, restart the program instead of reloading the macros file from the menu. Or if you reload buttons and behavior after reloading the macros, it will also prevent the problem.
• A negative sign preceeding a variable is dropped when the expression is packed. (i.e., -@x becomes @x) A workaround is to use @x*-1 instead of -@x, until I fix this.
• The main timer sometimes pauses when you open a modal window, such as the edit window.
• Communications with the world server sometimes stall. This is a result of the "fix" (ooops) I did to try to prevent a pause when opening a modal window. Until I fix it, you can restart the communications simply by opening the command window and closing it again (F5 then Esc).
• If an object's data has an embedded carriage return (if someone typed Ctrl-Enter to start a new line when typing something in the description field for the object, for instance), then a propdump created using the AW utility for that purpose will place an ASCII 127 following each carriage return, but Magsbot survey routines don't currently do that.
• This is a new one within the last beta version or two: sometimes the bot will disconnect without any sign of being disconnected. It will disappear from the world but not show on the instance list that it's been disconnected. You need to restart to program to reconnect; apparently the universe connection is being lost. The mysterious "Reason 10043" (not listed in the SDK reason codes!) is reported. This seems to happen after a "wsf" (waiting-for-server) situation in the world. I don't know what I changed that started this happening! Arrgh.
• Sometimes when you double-click on the "On" or "Off" in the "Active" column of the behavior table, it doesn't change like it should. The workaround is to just select the "On/Off" with the mouse then hit the Enter key to toggle the On/Off.