I’ve uploaded 3.1.4-svn5489, which contains fixes for the following problems:
- Bug 2731738: Memory test fails on Windows 98, Windows ME
- Bug 2718948: v3.0+ DoesNotStack trait bug
- Replace Crtl-Shift-Click Selection filter with Alt-Click selection filter
- Allow Alt-lasso to select Alt-click selection filter units
- Allow Ctrl-lasso to deselect Does Not Stack units.
- Double-click on Does Not Stack units causes selection status to toggle
- Bug: Bind to loopback ports only to avoid collisions and firewall problems
- Bug: Module Manager fails to start if port in prefs is already in use
- Bug: Race condition when setting Module Manager port and key
- Bug: GUI calls by Module Manager server executed off the EDT
- Bug: Memory test inaccurate on Linux
- Bug: Deadlock when Module Manager is closed before children
Other than the DoesNotStack problem, these fall into three categories:
-
Problems with checking the amount of RAM on the system (new to 3.1.3,
in order to fix another bug),
-
Problems with the Module Manager finding a free port to use, and
-
Problems with threading in the Module Manager.
If all of these fixes work properly, you shouldn’t notice any difference
from normal operation (unless you were having a problem before).
I’d like to release this as 3.1.4 sometime in the next few days; please
give this a look.
The build I’m now proposing as 3.1.4 is 3.1.4-svn5499.
There are two extra debugging features I added, as of this build:
-
Rather than being assigned a random long as an id, processes are given
ids which count up from zero. This way, you’ll see smaller (and easier to
read) numbers identifying processes in the errorLog.
-
Module Manager threads which exist to handle communication with a Player
or Editor are now labeled with the id of that child process. This will help
us identify threads when running in a debugger.
Has anyone had a chance to try svn5499? Should I release it?
I tested the new DoesNotStack trait. I like the Alt Combo restriction. It works fine for me. Though there is a little side effect
To select multiple game pieces with that trait configured to restrict selection using the Alt combo, you have to use either the Alt+Lasso (works like a charm) or the Alt+Shift+Click combo. And I noticed that on my vista laptop & desktop, the Alt-Shift combo does not change locale to EN/US.
I tried to select one unit with Alt+Shift then check locale in the chat window. Nothing changed. I selected another one, still nothing. Now I did the Alt+Shift combo while the main Vassal Control/chat window had focus and there it changed the locale.
That’s fine by me: it’s as if the fact that the Alt+Shift combo as a special meaning in the map window, the OS does not get to see it (intercepted?). Whereas in the control/chat windows, it has no use and then the OS gets it and changes the locale.
I have yet to check what would happen if my main map windows was “docked” in the control windows instead of being located in a separate window.
I tried to dock my windows by removing the button to hide & show the map window.
Then I tried different stuff: A+S select 1 piece and checking locale by typing characters in the chat windows that would change with local (typically a & q).
A+S click won’t change locale. A+S without any click in the following seconds and it would change locale wether mouse over map or not.
I don’t know if you’ll have more luck than I did to check if there is any consistency to the the “trigger” of this A+S combo changing locale or not within Vassal.
But as it is now, that suits me
I fixed one more bug:
- Bug 2696618: ImageSaver fails when map has nonpositive dimensions
So now the release candidate is 3.1.4-svn5503, which I’m uploading now.
I fixed another bug:
- Bug 2721306: IllegalArgumentException in PanelWidget.getLayout()
This would happen if you tried to add a Panel with < 1 column to a Piece Palette.
Now the release candidate is 3.1.4-svn5506, which I’m uploading now.
Fixed another bug:
- Bug: Java 1.5 ImageIO fails to load PNGs with iTXt chunks
This problem is kind of insidious, as Macs which have Java 6 still have Java
5 as the default until the user changes it. There was quite a long discussion
at ConsimWorld which led to isolating this bug.
That’s the last bug fix from me for 3.1.4.
Since nobody’s complained, I’m going to release it soon.