colossus-tracker Mailing List for Colossus (Page 42)

SourceForge Headquarters 1320 Columbia Street Suite 310 San Diego, CA 92101 +1 (858) 422-6466

Bugs item #2097824, was opened at 2008-09-07 04:52
Message generated for change (Comment added) made by cleka
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=101939&aid=2097824&group_id=1939

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: None
Group: None
Status: Open
Resolution: None
Priority: 5
Private: No
Submitted By: Babbling Balrog (nsillito)
>Assigned to: Clemens Katzer (cleka)
Summary: Great error message.

Initial Comment:
So the game hung with a battle mat open (see other bug report). So I started a new game, but the battle mat still didn't close when the game setup dialogue box opened. So I thought, I'll just close the battle mat and carry on. Well, I clicked on the X in the upper right corner and tried quit, nothing, tried close, nothing, so I tried something else and then got this interesting error message:

Arggh!! Seems the standard Quit procedure does not work. Doing System.exit() now the hard way. 

OK

This was such a great error message I had to post it.

After clicking OK the game shut down, including the new game setup dialogue. 

Oh well, I just start a new game.

The Aug 22 build seems to have introduced a lot of bugs.

----------------------------------------------------------------------

>Comment By: Clemens Katzer (cleka)
Date: 2008-09-07 08:27

Message:
Logged In: YES 
user_id=1717697
Originator: NO

Yeah, it's a nice message, isn't it?

Whatever it is worth, here what it means:

Earlier (2 or 3 public builds ago), upon starting a new game, Colossus did
not clean up much of the current game - threads did stay running, thus all
the memory they referenced kept used; it just created more objects and more
threads. So with every new game inside an once started Colossus
application, the used memory grew and grew.
When user did Quit, it just did "System.exit()" - tell the Java Virtual
machine to exit no matter what.
And, for multiboard games (hot seat), if one player was dead, you still
could not close his masterboard - doing so, the whole Colossus application
terminated - game gone.

I rewrote that to the following: 
When user selects new or load or quit, it stores somehwere what to do
next, and then initiated a clean shutdown phase. The client which owns the
board where you used the menu, did send a message to the server, so that
that one initiated the shutdown. It informed all clients to go down, close
their boards (if there were, e.g. hot seat games), then theu close their
sockets, and exit their "wait until something comes from server" loop. Out
of their loop, those threads cede to exist.
The server thread did some more cleanup, closed it's own sockets (play &
file serving), freed some static variables, and exited it's loop. 
This made him return to it's main loop, where it found what next to do:
new, load, or quit.
On load or new, it would do now same way, and basically same preconditions
as if done fresh after game start; for quit, it would just quit it's own
main loop and cease to exist.
No user thread running any more, the Java Virtual Machine would peacefully
just terminate itself. No System.exit() needed at all. I consider this to
be a clean exit, and a good thing.

And when in a hot seat game one board does "close", it can cleanly close,
leave the rest of game intact; if this is the last board, the server will
do same as if one did "new game" - back to the game setup.

Now, the message above: You say on client side once you want to exit. It
sets a flag for that, and then initiates the "normal" shutdown as said
above. As you noticed, this clean shutdown depends on the fact, that client
can still send to and receive from server. If somewhat later you have
opportunity to use this menu option again, it implies, the first attempt
failed - it sees you tried it because the flag is set. Not much use to try
again, is there? So, you want quit, clean way does not work, so it does it
the hard way, just as earlier. But it tells about it. Well mostly this
message was for me, when I implemented and made this working nicely, and,
under normal circumstances, the standard user is not supposed to see it.

Yes, then it kicks you totally out, have to start Colossus again. As
earlier. Without that (e.g. if started via web start), would have to go for
task list or ps -ef on unix cmdline, find which java ws process it might
be, and kill it with tasks manager or kill command.
(otherwise it stays running, using memory and perhaps blocking the port.
Wouldn't do good to anyone, now would it?).

And I think, this was working well when I committed it. Didn't get that
for a long time...

Around that time, Peter initiated some architectural cleanup, re-using
code from server on client side instead of duplicating, work more straight
with objects instead of everywhere lookup by markerId, for() instead of
iterators, which lead to eventually change to Java 1.5, and some more.
A good goal, but risky...

Especially this "instead of always lookup by ID" has broken many things; I
have fixed quite a number of those in recent months (the load game problem
is caused by it too), but it seems there are still some left. Now, if such
lookup problems occur, it might lead to exceptions which throw the client
thread out of it's read loop - and then the client-server communication is
broken.

((ah, and one thing: this clean shutodwn is also needed, and in fact I
started it due to it, because of the "Web Client" thing. I wanted it to be
possible to discard the current game and clean board and all, but the web
client stays up. i.e. had to get rid off all the system.exit().
*Sigh*, noone is using WebServer and WebClient since we don't run a
server, but perhaps one day... ))

I hope someone agrees with me that this exit procedure is an improvement
against the old way...

-Clemens

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=101939&aid=2097824&group_id=1939

2007	Jan	Feb	Mar	Apr	May	Jun	Jul (23)	Aug (6)	Sep (15)	Oct (31)	Nov (22)	Dec (47)
2008	Jan (21)	Feb (19)	Mar (10)	Apr (15)	May (25)	Jun	Jul (6)	Aug (91)	Sep (85)	Oct (192)	Nov (14)	Dec (11)
2009	Jan (19)	Feb (8)	Mar (14)	Apr (1)	May (5)	Jun (6)	Jul (54)	Aug (115)	Sep (56)	Oct (21)	Nov (39)	Dec (7)
2010	Jan (54)	Feb (15)	Mar (15)	Apr (14)	May (4)	Jun (10)	Jul (6)	Aug (6)	Sep (10)	Oct (3)	Nov (1)	Dec (16)
2011	Jan (13)	Feb (92)	Mar (11)	Apr (9)	May (5)	Jun	Jul (5)	Aug (7)	Sep (2)	Oct (3)	Nov (10)	Dec (5)
2012	Jan (4)	Feb (19)	Mar (5)	Apr (6)	May (10)	Jun (2)	Jul (1)	Aug (8)	Sep (2)	Oct (2)	Nov (5)	Dec (3)
2013	Jan	Feb (32)	Mar	Apr	May	Jun	Jul	Aug	Sep	Oct	Nov	Dec

colossus-tracker Mailing List for Colossus (Page 42)

colossus-tracker — notifications from Colossus trackers