===============================================================================
==== XPilot Change Log ====
===============================================================================
This is a summary of changes of release 4.4.3 compared to release 4.4.2.
A bug was fixed regarding asteroids wrapping beyond map_width.
Jude J Anthony submitted a bugfix for robots:
The problem appears to be a confusion between the global Obj[] array (object
*Obj[MAX_TOTAL_SHOTS], declared in server.c, and included through global.h)
and a local array of object pointers.
Within Robot_default_play_check_objects(), at line 1874, we declare a local
array of object pointers (object **obj_list). At line 1907, we get all
items within our visibility with Cell_get_objects(), placing them in the
obj_list. At 1911, we loop j through all the items in obj_list, setting
shot to an object pointer within the list at 1913. At line 1965, we run a
check for items of type OBJ_ITEM. (Thus explaining why the error only
occurs with robots AND items.)
On line 1973, try to find the Importance of Obj[j], but j is actually an
index into obj_list.
===============================================================================
This is a summary of changes of release 4.4.2 compared to release 4.4.1.
Paul Alexander Drees submitted a bugfix for the Windows version of XPilot.
It used to be that, if you typed a message, the score window would go blank
until a new set of scores came in.
Joe Groff submitted a patch to allow more keys to be bound to one
mouse button.
Jarno van der Kolk submitted a patch:
Almost all of the serveroptions can now be "changed at runtime".
Only file-options, like scorefiletable, defautsfile, etc. still can't be
changed at runtime. And the options, timing and teamplay.
Bob Glickstein submitted "flak hack." It changes the shots fired by cannons
to be AAA ("anti-aircraft artillery") rounds. These work like normal shots
but explode into small flak bursts after a short while.
Austin Donnelly submitted a patch:
Also discussed on rec.games.computer.xpilot was an proposal to make the
ball connector non-rigid. This patch adds a new option
"connectorIsString", (default no) which when set allows the connector
to go "slack", so you can't push against the ball.
This makes certain ballruns harder, and others easier. It does take a
little getting used to.
Guido Koopman reworked Bob's flak patch a little. Flak explosions are smaller
and flak can be turned off with an option cannonFlak.
Austin Donelly submitted a patch to allow clients to make some
attempt at backwards compatibility with old servers which can't handle
a turnResistance of 0.0.
Austin Donelly submitted a patch, which adds two new options:
ballCollisions (default no) Do balls collide with other objects?
ballSparkCollisions (default yes) Can balls be blown around by exhaust?
ballSparkCollisions is only active if ballCollisions is also on.
What happens during a ball-object collision depends on the object.
Shots (including missiles of various kinds) are absorbed in an
inelastic collision. Other balls bounce off, unless they were going
too fast (as determined by maxObjectWallBounceSpeed) in which case
both balls pop. Wreckage and debris is also absorbed. Laser pulses,
mines and powerup items pass unhindered through the ball.
I originally had items bouncing off the ball, but this turned out to
look silly: items have low mass and even slightly clipping them
with a ball would send them off at high speed without affecting the
ball much.
I tried to get lasers absorbed, but couldn't work it out - I suspect
the collision detection is slightly different, but I haven't spent
much time on it. Mines I couldn't decide whether they should trigger
or bounce of the ball (so I ignored them).
This is my latest version, and fixes the problem I mentioned earlier
about balls in treasures being blown about.
One of the cute things I've found is that you can blow a ball by
placing and detonating a mine close to it :)
Paul Alexander Drees submitted a patch for the server:
It makes asteroids release items when they are broken.
The probability that an asteroid releases items is given
by the option asteroidItemProb, while the number of items
that are released is between 1 and asteroidMaxItems.
I also streamlined some code in frame.c using a function I
created for the asteroid item determination.
Juha Lindström submitted a patch for the client:
The attached patch draws the object information of the radar to the
map on an area three times the size of the radar. The information
is drawn in the sliding radar way so that you sit in the center
of the radar and the circles show where the opponents are relative
to you. Now if one of the circles suddenly moves quickly towards
you, just drop a few bullets to the direction of the circle, move
behind some cover and BANG the world is again a bit better place
to play xpilot :)
Jarno van der Kolk submitted a version of xp-replay for Windows:
Finally, it works. (i.e. it doesn't crash all the time) :-)
This ZIP contains the source and an executable in the Release-dir.
The files should be unzipped into src/replay/NT, since the project uses
"src/client/recordfmt.h".
Beware that XPreplay loads the entire XPilot recording into memory, so
it doesn't really like the 80 Mb recordings from the summercamp. It CAN
handle the ballrunrecordings from the various ballrunsites just fine
though.
===============================================================================
This is a summary of changes of release 4.4.1 compared to release 4.4.0.
When tuning text options at run time like the password the server
would crash. This has now been fixed.
Clients now support the firewall patch.
For this the client has two new options: -clientPortStart and -clientPortEnd.
When the client is started with e.g.:
-clientPortStart 40000 -clientPortEnd 41000
then the server will use this range for UDP ports to communicate with clients.
This is useful for sites which are behind a firewall.
A new server option -passwordFileName gives the path to a file which
only contains a definition for the server password. The advantage
of this is that you can set the file permissions so that others
cannot read its contents.
Zemyla submitted a patch to fix a small bug in his old blockFriction patch.
The problem was that the option -blockFrictionVisible would not work.
This patch fixes it.
Ben Armstrong fixed a bug where ball ownership for all connected balls
would be reset whenever a player would leave.
===============================================================================
This is a summary of changes of release 4.4.0 compared to release 4.3.2.
New option (bool) logRobots: whether to log robot's comings or not.
Windows only: server config dialog box has a button to allow you to choose a
map and two url links: local Server Options; and Jarno's Server Options Page.
Removed support for CMW and VMS as these have been broke for a long
time and are not expected to be updated anymore.
Removed support for antique 0.x versions of Linux which did not have
support for sending broadcasts to the network and had small socket
buffer limitations.
Add strlcpy and strlcat. Replaced most of the calls to strcpy, strncpy,
strcat and strlcat with calls to strlcpy and strlcat. New code should
preferably only use strlcpy and strlcat.
Removed sourceid strings as they only produced compilation warnings
and haven't ever turned out to be useful.
Paul Alexander Drees, a.k.a. Zemyla, submitted a big patch with a large
number of small new features:
- A small pair of client options (-user and -host) to be abls to set
the user@host in the .xpilotrc file or on the command line.
- A set of options to control ball parameters:
o -ballMass sets the mass of the ball.
o -ballConnectorStringConstant to set the restoring force of the
connector.
0 -ballConnectorDamping to set the damping of the connector's motion.
o -maxBallConnectorRation to set how much the connector can deform
before it breaks.
o -ballConnectorLength to set the default length of the connector.
- -cannonDeadTime to set how long (in seconds) a cannon stays dead once
it is killed.
- -targetDeadTime to set how long (in seconds) a target stays dead once
it is destroyed.
- Random items. These items show no hint of what they are; instead,
they appear to alternate between different items. The items shown
depend on what probabilities there are for items; for example, a map
where the only items are wideangles and armor would have its random
items alternate between wideangles and armors. Dropped items are
never random.
- -randomItemProb to set how oftem items appear random (0.0-1.0).
- -/+captureTheFlag mode. When -captureTheFlag is on, the treasure
that an enemy ball is being deposited in has to be safe in order to
score.
- Friction blocks. The map character for friction blocks is "z".
When a ship is in a friction block, the frictional force on it is
equal to -blockFriction instead of -friction. There is also the
-/+frictionVisible option to change whether the friction blocks
appear as decor or not at all.
- -coriolis determines through what angle (in degrees) a ship's
velocity vector rotates in a frame.
- -robotRealName, -robotHostName, -tankRealName, and -tankHostName.
These options set the realnames and hostnames for robots and tanks.
- -tankScoreDecrement. This option describes what the difference
is between the player's score and the score of his released tanks.
- -/+turnThrust. This option behaves exactly like the former
compile-time option of the same name -- turning uses fuel and produces
sparks.
- -/+selfImmunity. This option determines whether a player and
his tanks are immune to his own shots, mines, sparks, etc.
- -defaultShipShape and -tankShipShape. This sets the shipshape
for people who don't define a shipshape (or everyone when
-allowShipShapes is off), and the shipshape for released tanks.
- An enhancement to the Tune_option function that lets it set
valSec and valPerSec values (so the /set command can change
these values).
- A refinement to the /set command to show the value after tuning
in the server message.
- A /get command to get the current value of an option. It does not
work for password and mapData (on purpose -- I do not want any Joe
Blow to be able to get the password, and the mapData far too long.)
- A refinement to the /help command. When the command is given with
no arguments, it iterates through the list of commands and displays
each one in turn instead of showing a fixed list.
- Empty treasures. These act like regular treasures, but they do not
own a ball, and are not counted in the number of treasures a team
owns. Thus, they are useful for places that are inaccessible to
players, but that balls can be dropped into (like Worm Mountain).
- A refinement to the Detonate_item() function. If the victim is a
tank, the items belong to the player who owns the tank.
- A refinement to shoving. Shoves now count as kills, robots can
declare war on people who shoved them, and robots can undeclare war
on people they shove.
Guido Koopman submitted a patch:
- treasures are allowed in non-team mode.
balls can not be saved because they don't belong to a team.
number of frames loose is not reported when a ball is destroyed.
- robots can declare war when killed by a ball.
- small changes in the client to paint treasures without team correctly.
Fixed the saving of recordings to use the proper window dimensions.
Added Ben Jackson's fix for not drawing the border of the active view
with window scale factors which are equal to one or less.
The order in which radar information about players is sent is now randomized.
This is to prevent cheat hacks which take advantage of the fixed order
some information is sent in.
Paul Forman submitted code which adds options to prevent players
from placing stationary mines and prevent players from throwing
mines of the speed at which they fly is below a limit.
This patch was modified and simplified to only add one new option
to the server -minMineSpeed which controls the minimum speed at which
mines move. The default is zero, which allows for stationary mines.
Non-zero values will give the mine an initial speed based on the
player velocity and direction.
Fixed bug which caused lowering of number of robots not to work.
Guido Koopman submitted a really nice patch which implements asteroids:
- a new object type: asteroid
- version magic number has been increased to 4400 (was 4300)
- there are currently only two shapes for asteroids, but more
can easily be added.
- sounds for asteroids: asteroid_hit and asteroid_break
- a new packet type: PKT_ASTEROID (number 80)
- two new files in src/server/: asteroid.[ch]
- constants like mass, size etc. are #defined in asteroid.h
- points are awarded for hitting asteroids with balls
- four new server options:
option name: description: default:
asteroidsWallBounce Do asteroids bounce off walls? true
asteroidCollisionMayKill Are asteroids deadly? true
asteroidsOnRadar Are asteroids visible on radar? false
asteroidScoring Points for shooting asteroids? true
asteroidProb Asteroid create probability 5e-7
maxAsteroidDensity Max. [0.0-1.0] asteroid density 0
Behavior of asteroids:
Asteroids are created with a random size from 1 to the maximum size (4).
Bigger asteroids can take more hits before breaking than smaller ones.
Asteroids size 1 can take no hits (are destroyed when hit).
When an asteroid breaks, two asteroids of size N - 1 are created, plus
some debris and wreckage.
If asteroidScoring is on, one point is awarded per asteroid broken (or
destroyed for size 1), up to a maximum of 101 points. This is the same
as for cannons.
Clients with versions less than 4400 are not sent the asteroid objects,
although they can still see them on the radar if asteroidsOnRadar is on.
Sometimes when a client was slow to join in an almost full game a robot
could be started which would cause the client login process to fail.
This has now been fixed.
Guido submitted a new patch to let asteroids collide with many objects:
Improved asteroid - object collisions. Asteroids now collide
with wreckage, debris and sparks from players or cannons; no
longer collide twice with other asteroids; trigger mines and
torpedoes instead of hitting them. Also, different objects
do different amounts of damage to asteroid. Players and
asteroids bounce elastically.
New algorithms have been implemented in the XPilot server which speed up
the handling of objects. This will especially be noticable on servers
with more players, or servers with lots of robots and objects. These
servers will see reduced CPU usage, which benefits network I/O processing.
A client visible feature related to these improvements is that packets
received will hardly ever be larger than one kilobyte. What happens
when a player is near a very large number of objects is that at most
the 500 objects closest to the player are transmitted over the network.
The option parser internals have been overhauled. Now for each option we
can specify from which sources it can be set (command line, defaults file,
or the map file). And we can control which options can be seen by clients.
When an option list was requested options with type valSec or valPerSec
were not corrected for FPS. This made them look very different from
their values in the map file. This has now been fixed.
Now a few options are not listed when a client requests to see
the list of server options. These include: robotFile, mapFileName,
defaultsFileName, scoreTableFileName, and password.
Added a -motdFileName server option. This option specifies which
file to use for the MOTD which is displayed when clients join.
Its default value is determined at compile time, or gotten from
the environment variable XPILOTSERVERMOTD.
When cannons were configured to use items this could crash a server
when a full nuclear cluster mine would explode.
Now option define statements which are set in the defaults file
can be expanded in the map file and by doing so set options which
could not have been set in the map file explicitly. This sounds
very complicated, but it will be easy to understand with an example:
When in the defaults file I specify these two define statements:
define: PrivateSettings \multiline: EndPrivateSettings
idleRun: no
noQuit: no
contactPort: 15000
serverHost: localhost
reportToMetaServer: no
EndPrivateSettings
define: PublicSettings \multiline: EndPublicSettings
idleRun: yes
noQuit: yes
contactPort: 15345
reportToMetaServer: yes
EndPublicSettings
Then I can expand these defines in any map file with:
expand: PrivateSettings
or:
expand: PublicSettings
This does allow me to choose between these option values in
each map file. While the options themselves are of a kind
which normally cannot be set in mapfiles, but only in the
defaults file or on the command line.
Some changes have been made to accomodate maps with hundreds
of robots. The most important of these changes is the addition
of a new packet type to send radar information to clients much
more compactly. This is packet type is called fastradar packet.
Ben Armstrong updated the xpilots manual page.
===============================================================================
This is a summary of changes of release 4.3.2 compared to release 4.3.1.
There is now a Visual C++ XPilot.dsp in the project root directory to
allow easier building of Windows code.
Nasty shots now use the shot_size variable.
Since at least version 4.2.1 there was a bug in saving keys to the XPilot
client config file. This has now been fixed.
Paul Alexander Drees submitted an updated robots file.
===============================================================================
This is a summary of changes of release 4.3.1 compared to release 4.3.0.
Some .mak files in the NT subdirectories were removed because it was too
hard too keep them synchronised with the rest of the source tree.
The file src/common/socklib.c was rewritten from scratch to get rid
of some old non-GPL copyright stuff. Now everything in XPilot should
be under the GPL copyright.
Applied Jarno van der Kolk's fix for improving sliding radar performance
under Windows.
Replaced the existing map generator with wildmap, which generates much
better maps if the default map could not be found, or if "-map wild"
was specified on the command line.
If the client stopped receiving frames, you couldn't disable
mouse control unless you managed to free the mouse pointer with
something like WM key shortcuts. Fixed this problem thans to Uoti.
Ben Jackson submitted code which adds a button to the welcome screen
with which you can ping all the servers. The sort criteria for the
list of servers is then augmented with the ping times.
Ben Jackson submitted partially finished code to measure the speed
of different graphics features. He didn't commit yet to any
further work on this.
Juha Lindström submitted code to fix the new socklib library
for Windows peculiarities. He adds the following remark:
It seems that the current 4.3.1alpha doesn't compile on windows due to a few
issues in socklib.c and wildmap.c. Attached is a patch I used to get the
windows version working. Below is a short explanation of the socklib issue
as it was more complex.
Currently the windows part of function sock_get_host_by_name expects that it
is called from a thread that does not own the window notifyWnd. This is the
case when the function is being used by the server but as the client usually
contains only one thread it will block forever in the function. In previous
xpilot versions this was handled so that the client never used GetAddrByName
(the old name for sock_get_host_by_name) but instead used gethostbyname
directly. In this new version every host name lookup is done with
sock_get_host_by_name and this causes the client to freeze.
The patch fixes this problem in the following manner. The function still
uses WSAAsyncGetHostByName to do the lookup to solve the long timeout
problem with gethostbyname. After calling WSAAsyncGetHostByName the calling
thread enters a loop and starts polling its message queue for a
WM_GETHOSTNAME message indicating that the lookup is ready. If the calling
thread is the owner of notifyWnd (as is the case with the client), windows
will deliver this message to the queue. If the calling thread is the worker
thread of the server , windows will deliver the message to the message queue
of the main thread because it is the owner of notifyWnd. Upon receiving this
message, the main thread will forward it to the worker thread using
PostThreadMessage thereby letting the worker thread exit the loop.
As I'm not at all experienced in windows programming, I'd suggest that
someone who is reviews that stuff before any of it is included in the
release.
-- Juha
Ben Armstrong submitted a patch to limit the UDP ports which the
server uses for client connections in a predefined range.
For this the server has two new options: -clientPortStart and -clientPortEnd.
When the server is started with e.g.:
-clientPortStart 40000 -clientPortEnd 41000
then the server will use this range for UDP ports to communicate with clients.
This is useful for sites which are behind a firewall.
Fixed compilation problems for Linux with glibc 2.2.2.
Ben Jackson supplied a patch to the server, which adds a new -serverHost
option. With this option you can specify the hostname and IP address
the server will use if the machine has more than one network interfaces.
Fixed a bug which caused the client to crash when the shipshape file was
specified in the ~/.xpilotrc file, but the shipshape file did not exist.
Ben Armstrong submitted an updated manual page for the xpilot client.
The default value for maxColors has been changed from 4 to 8.
On some old and cheap graphics hardware this may slow down the
client a bit, so in that case players may need to explicitly set
the value of maxColors to 4 in their XPilot defaults file.
Ben Jackson submitted a patch to have XPilot compile under Sun Solaris.
Ben Armstrong submitted a patch which adds an option to the client
which controls the color of old messages. The option is -oldMessagesColor.
It can be configured in the "config" window.
The doc/README.sounds file has been updated to the current state of affairs.
The XPilot sounds file is now located as LIBDIR/sounds instead of
the erroneous definition of SOUNDDIR/sounds.
===============================================================================
This is a summary of changes of release 4.3.0 compared to release 4.2.1.
Changes to be able to support many more colors.
Support for RGB color cubes to be able to map true color pictures
to screens which have less colors.
Claus Leth Gregersen submitted a busload of new code which adds fancy
graphics to the client. The entire world and all objects and ships
can now be drawn in full color pixmaps. He also submitted code
to be able to support the old client graphics and the new pixmap
graphics in the same client.
Jarrod Miller updated mapedit to include all the 4.2.1 server options.
Changes to maps2image in Makefile, bdf2c.c and maps2image.c for portability.
One fix in maps2image.c for closing a file too often, which caused a crash on
some platforms.
Fix for emergency shields. When maxShieldedWallBounceSpeed was more than 100
bouncing with emergency shields would lower the max allowable bounce speed
to 100.
Added the Mersenne Twister pseudo random number generator to XPilot.
Changed the pseudo random number code to use the Mersenne Twister instead
of the stock rand(3).
The server log file for messages to god now must exist and be writable
and be less than 100 KB before messages to god will be logged.
Peder Herborg submitted a patch to make sliding radar also work under Windows.
Ben Jackson submitted a patch containing several fixes:
altScaleFactor doesn't need a special update function.
Changing it doesn't change the display.
textured walls need WINSCALE'd texture offsets to stay in sync
with the walls
split the sending of phased/cloaked for new clients
new clients have enough info in guiobjects.c to paint phased
ships differently, but for now it calls the cloaked ship
drawing. the code should probably be reorganized so that
phasing and cloaking aren't mutually exclusive.
I think there was a bug that would have made cloaked ships on
your team show up in the wrong color, so I fixed it. No
current maps get into that situation (unless you count bloods
with items).
misc cleanup in guiobjects.c
fix another excess Send_display() bug. we need an xpilot
protocol dumper to check this sort of thing.
swapping scalefactors calls Config_redraw() to swap the values
in the config panel
fix the xpilot.domain.com search code. I might be the only
server using it these days...
servers send radar team information to clients running the
patch version that was first to have the feature, since the
data is sent the same way.
Now the server only tries to do a DNS lookup for xpilot.yourdomain.com
when the server option searchDomainForXPilot is explicitly turned on.
The reason to make this an option lies in the fact that for some sites
DNS lookups are slow, especially lookups for nonexistent names.
===============================================================================
This is a summary of changes of release 4.2.1 compared to release 4.2.0.
The server option -armorHits was inadvertently mentioned in the ChangeLog
for version 4.2.0. It was removed from 4.2.0 before the public release.
Clients with -maxColors set to less than 8 would not remove the high bit
from the size element of a radar packet. When the -maxColors was set to 4
and if the client was connected to a new modified server then this bug
caused teammates not to be visible on the radar.
The server now sets the high bit of the size element of a radar packet
when the player to be shown is a teammate. The server only does this
for clients which have a version of 4.2.1 or higher.
Clients who have the -maxColors option to 8 or higher can now show
teammates in green on the radar when playing on a 4.2.1 server.
Changing the sparkProb setting in the client while playing would
not always succeed. This should be improved now.
Support for displaying more than 16 colors has been added for
clients who do not play in colorSwitching mode. This feature
notably will be useful for clients which have been configured
to run in multibuffer or doublebuffer mode.
Erase mode now also works in multibuffer and doublebuffer mode.
The -altScaleFactor client option now works even if not configured.
The carriage returns in a few client source files have been removed.
Some features of the DeathCup 2000 server are now standard in XPilot.
These include:
- The useWreckage option to turn off the wreckage that
destroyed ships leave. The default value is on.
- The roundsToPlay option can be used to specify the number of
rounds to play. By default this is set to 0, meaning no limit.
- The timerResolution option can be used to specify the operating
system specific timer resolution. Setting this option to the
correct value can improve the accuracy at which the server
transmits frame updates to the clients.
- The password option can be used to set the password which is
needed to obtain operator privileges while playing.
- Using talk messages starting with a slash players can now
interact with the server. The /help command lists the
available commands. Some of these require operator
privileges which can be obtained by giving the correct
password to the /password command.
When sending frame updates to the clients the server now first packs
information about ships before shots. It used to be the reserved.
On some servers with lots of items this could cause ship information
not to be included because the frame update packet was already full
with item data.
On maps with robots and balls the server could crash at startup
when -idleRun was also set as soon as a robot would be searching
for the nearest ball. This caused a server crash on Solaris,
but not on Linux.
===============================================================================
This is a summary of changes of release 4.2.0 compared to release 4.1.0.
Sven Mascheck submitted two bugfixes. One is a problem with locking on
players during teammode. The other with lock banks.
Teemu Elias Hirsim$ki fixed a bug in xp-replay related to the recent
addition of the mirror item.
Some changes were derived from the XPilot summercamp patches webpage:
- client: an extra check for line style being solid in paintobjects.c.
- server: extra checks to keep player direction within a valid range.
- server: If you were near your treasure box and couldn't connect to the
unmoved ball there, you sometimes were unable to connect to any other
balls either.
- server: if the target was destroyed of a team without an active player
then no message would be displayed.
- server: loadlock could cause a stale lock to be loaded. This could
result in a server crash.
Sometimes a player could not join over a slow connection because a robot would
take his homebase. The message xpilots would print would look like:
xpilots: Not enough bases for players
Ben Jackson submitted a large number of bugfixes for client display:
These patches fix lots of bugs with the client display when using a
scaleFactor other than 1.0. The client looks significantly better at
both very large scale (< 0.5) and very small scale (>= 2.0). I believe
I have fixed all of the font position bugs, some of which were even
visible with no scaling. The client looks better both with much bigger
and much smaller fonts for gameFont. All of the changes were tested
with and without -erase, so there should be no new pixel dust.
Some of the most significant improvements are in the HUD, which had
many layout bugs due to scaling. Also, a longstanding bug with the
duration of the fuel display has been fixed.
src/client/painthud.c:
Make Paint_meter() consistent about what is WINSCALE()d and
what is not, and avoid scaling twice for some things.
Position scoreobjects properly.
Position lock range correctly.
Omit mini version of locked ship if it would be too small to
identify anyway (smaller than 1/3rd original size overall,
which makes ships like the LARDBUS 10x10 pixels).
Revamp item display to avoid overlaps and other problems.
Center "Autopilot" light correctly.
Center hud messages correctly. Also, if the first HUD message
is too wide to fit in the gap at the bottom of the HUD, move it
down a line so it doesn't overlap the fuel/vert hud lines (if any).
Move fuelCount decrement so that the fuel number actually
disappears eventually if fuelMeter is off. See also the comment
about shorter FUEL_NOTIFY time below.
src/common/const.h:
Decreased FUEL_NOTIFY (time the number stays on) from 16 seconds
to 3. If you liked the old behavior, just set fuelNotify (no
relation to that constant) to a very low number and it will stay
on all the time, just like before.
src/client/paintmap.c:
Properly center `F' in fuel squares. Fully erase it if the font
is bigger than one square.
Change display of base teams/names to center the info below the
base rather than left-justifying it. Make it work for larger
fonts. Make it display team numbers/player names for bases facing
right without overlapping them when scaling is on.
Target team numbers are properly centered.
Treasure team numbers are centered and just above the ball.
src/client/paintobjects.c:
Properly center `P' in paused player blocks. Huge fonts still
lose for -erase, though.
Properly center bomb owners under bombs.
Properly center player name under players. Added 1 pixel more
separation between the name and the ship.
Properly center the round delay.
Fix ball size with ballTexture and scaling. You'll still have
to supply a ball.xpm that's the right size to get the expected
results.
Scale mines, properly erase them at all scales.
src/client/xinit.c,
src/client/configure.c:
Scale dashes/cdashes to scaleFactor. This affects all dashed
lines: (visible) cloaked ships, shields, tractor beams, hud
lines, partially conencted treasures, refueling lines...
src/client/netclient.c:
Clarify the draw_width/draw_height role. Rescaling and resizing
both now have the proper effect, and the client doesn't send
resizes for every frame.
Ben Jackson also submitted a patch to improve the accuracy with which lock
distance is calculated:
src/server/update.c:
Move computation of lock distance to a second loop after the
player position update loop. The direction is computed during
the frame update, so the distance and direction were out of
sync, but only for players with lower indices than the player
they were locked onto.
This affected the HUD display for players and messed up some
robot computations.
Ball and player positions were not calculated at `the same time'.
So it was not possible to connect at very high speeds.
This is called the `connector bug'. Fix taken from
Uoti Urpala's server patch.
New client option `selectionAndHistory' for the unix version:
* You can place the cursor in the talk window with the left pointer
button. This works also with `Left'/`Right'.
* You can cut from the player talk messages (at the left top edge).
* You can cut from the talk window.
* At both places, the selected text is drawn emphasized (blue)
(but not until you finish the selection).
* You can paste into the talk window.
Unprintables (according to iso-latin-1) are converted to space.
* Your own messages are stored in a cyclic history (default ist 32 lines)
you can browse through it in the talk window with `Up' and `Down'.
Number of saved lines is `saveLinesInHistory'.
* The `Up',`Down',`Left',`Right' from above are configurable with
`keyTalkCursor[Up|Down|Left|Right]'.
* You can print all current server and player messages to the `tty' with
`Print' (`keyPrintMessagesStdout').
Details:
* As long as a cut from the messages is pending (button not released),
no new messages will be drawn, but they get buffered and are drawn after
you release the button.
* Emphasized text in the talk window gets deleted when typing on.
* The cut will notice if you wanted to include a final `newline'.
* About the `history':
- When you type some text and then start browsing through the history
(without pressing enter) your text is stored and still available
by browsing. Purpose: looking at/resending/modifying older messages.
- The history line currently drawn is automatically stored in the
selection, means you can paste it immediately.
* Cutting in common:
- The selected text is stored the way like terminalemulators do.
It stores selections in both `cut buffer' and `primary selection'
(in contrast to netscape f.i. using only the primary selection).
[Some versions of netscapes sometimes delete(!) the `primary selection'
when starting (why did they call it `mozilla'?)]
* Cutting from the player messages:
- As long as there is a selection, older messages will get white but then
their `expire' won't be counted down further. Thus they won't disappear
(except when new messages come in).
- Cutting doesn't include the last character - like in
a terminalemulator.
* A blinking cursor being over an unemphasized underscore is drawn
emphasized to be able to still see that underscore.
* About finishing a cut (releasing 1st button) outside the talk window:
if you do it right from where you started, it's considered as
cutting all the rest of the line. Left: analogous.
* note: only one XClient `owns' the `primary selection' at a time.
Fixed auto-idle-pause bug
(client auto shield when restarting from the base was interpreted as keystroke)
Lasers disappear less frequently when wrapping (from Uoti Urpala).
Fixed turnspeed bug (from Uoti Urpala).
Team locking fixed again.
If FPS is odd, the client might have gotten no frames
at all (from Uoti Urpala).
Wreckage color was changed to red (pointed out by Espen Skoglund).
But deadly wreckages are still painted in white.
When the client has set the sparkProb option to 0 then the
server will not send wreckages when they are not deadly.
Ben Jackson submitted fixes for better (in-client) handling of
changing power/turnspeed and the Config Window (reflecting them):
src/client/configure.c
Add Config_redraw() and associated support, for updating the
config display when parameters change. Used below.
Change constraints on maxFPS from 0..30 to FPS/2..FPS. The
code was always keeping the variable maxFPS in that range, but
the display could go one past, which was annoying.
src/client/client.h
src/client/client.c
src/client/paintdata.c
src/client/painthud.c
Client always ignores PACKET_SELF power/turnspeed for purposes
of configuration. Now it just displays it in the meter, but
maybe even that is misleading? Maybe it should only do that
if snooping, and otherwise display the correct values, assuming
autopilot is messing with them.
src/client/paintdata.h
src/client/paintdata.c
New global `snooping' which is true whenever we're watching
another player instead of ourselves.
Since server never calls Send_eyes() (?!?!?!) the client now
makes a best guess -- it assumes that the player at the center
of the display is the player we're watching, especially if it's
the player himself!
src/client/xevent.c
KEY_SWAP_SETTINGS uses a heuristic to guess if the server will
accept this keystroke, and if it thinks so it sends the keystroke
and ALSO swaps the settings locally, and updates the config
display.
KEY_INCREASE_POWER, KEY_DECREASE_POWER, KEY_INCREASE_TURNSPEED
and KEY_DECREASE_TURNSPEED are all done in-client now. The client
sends the new absolute value. This is probably the Right Thing
anyway, and it's necessary since we now ignore the PACKET_SELF
power/turnspeed values.
src/common/pack.h
src/server/netserver.c
New magic version 0x4200.
old clients->new server: power/turnspeed while autopiloting are ignored
by server to avoid having people set values
relative to the autopilot values
new clients->old server: power/turnspeed are not corrupted by autopilot
but the old server will lose the value. I
tried having the new client send the values
whenever it sees the autopilot light go off,
but due to a combination of bugs in the
OLD server this is disasterous.
new clients->new server: power/turnspeed sent while autopiloting set
the saved values and are used as soon as
autopilot is off
The disasterous bug mentioned above is caused by being able to
set pl->turnspeed while autopiloting in older servers. That
can't happen anymore because it's either ignored (old clients)
or put into the save values (new clients).
src/server/object.h
src/server/update.c
Autopilot no longer saves/restores turnacc -- was irrelevant anyway,
since it is immediately recomputed.
A frame counter for balls was added (from Uoti Urpala's patch).
Parking on the base is easier now (also due to harder debris) : the allowed
maximum speed is increased.
Ignore client maxFPS request if it is 20 (the client default setting)
and if serveroption `ignore20MaxFPS' is true (default), by Uoti Urpala.
The client knows about talk macros now (Jonny Svärling, Adam Lorentzon)
modified: macros work also properly in teammode when one is dead.
See doc/README.talkmacros.
Guido Koopman submitted a patch for a new armor item:
- one new item: armor. If shields are not used/available, an armor item
can absorb a specified number of hits before the player is killed.
The idea is to bridge the gap between un- and shielded games.
- Three new options: initialArmor, maxArmor and itemArmorProb.
With defaults of 0, 10 and 1e-9.
Only check for robot leaving when it is not transported to its homebase.
This leaves the shots from a dead robot in the game a bit after it dies.
TkXpInterface had a problem accessing temp_dir in a subroutine where
it was not declared as a global variable.
The Makefile of maps2image had some problems.
Fixed incorrect use of rand throughout the server. The problem with
most rand implementations is that the low order bits are not very
random at all. Therefore almost all uses of rand have been replaced
with rfrac.
Now larger map areas can be viewed when using larger scaleFactors on
big screens. Try "xpilot -scaleFactor 2.5 -geometry 1280x1024".
When using pointercontrol, configuring turning settings is easier now:
Setting it to zero makes the server completely ignoring it.
A first cut at a graphical user interface to the meta server.
This is still very very alpha. It works, but not everything.
Ship shize increased from 14 to 16 due to `analytical collision detection'
which inspects a real circle and not just a square anymore.
Claus Leth Gregersen submitted a big patch:
- Introduced a new key for swapping scale factor settings.
- Fixed a slight bug in a MS-Windows paletterealization.
- Moved all the painting logic expect the linedrawing for the blocks
into guimap.c, added a border around the updated area too.
- Moved all functionality out from the key_press switch into seperate
functions.
- Extracted the function from xevent and put them in a new file
xeventhandlers.c, those functions only used in this new
file was also moved.
- Fixed the windows mouse bug with a hack, fixed a few compile problems.
To prevent cheating the order in which information is send from the server
to the client about players and objects is now permuted.
Guido Koopman submitted a patch for team cannons hitting cannons of another
team:
- added some extra checks to FROMCANNON objects hitting a cannon to allow
cannons to be killed by cannons from opposing teams. They don't actually
aim or fire at the other cannons, but a shot could accidently hit them when
it misses a player.
Jarrod Miller submitted a patch which fixes the orientation of bases that sit
on the edge of a map when edge wrap is on and a base orientor on the opposite
edge should be accounted for.
Claus Leth Gregersen fixed portability problems for Windows and
submitted several patches to improve the way the graphics drawing
code is organized.
===============================================================================
This is a summary of changes of release 4.1.0 compared to release 4.0.0.
Fixed a bug which caused some valid hostnames to be rejected.
Changed the Imakefile for better compilation on Solaris 2.6 with gcc.
Sven Mascheck fixed a bug which could cause the playing window to remain
blue after being ECMed. This bug only occurred in erase mode.
Guido Koopman contributed a bit patch with several new features:
adds/fixes the following:
- new item: mirror. reflects laser pulses. usual options with usual defaults.
doesn't need turning on or off and uses no fuel. each additional mirror
halves the chance a pulse gets through.
- cannons now use all offensive and semi-offensive items. see cannon.[ch] for
lots of comments.
- option cannonsUseItems replaces cannonsPickupItems. old option now used as
alternative spelling. if on, cannons don't pickup, get or use items.
default is off, to keep the conservative majority from complaining.
- new option cannonItemProbMult. if 0.0, cannons don't get items, except for
the initial items, but can still pickup or steal them. 1.0 means each
cannon gets an item once per minute on average. default is 1.0.
- resubmission of min/maxRobots options. see previous patch for info.
- new option resetOnHuman. if >0, the round/race is restarted when the < Nth
human player logs on. handy if you don't want to wait for the robots to
finish the round/race.
- new option teamCannons. when on, cannons don't activate on teammembers.
with teamImmunity on, players can't be hit by fire from their team's cannons
and can also fly through them. default off.
- new option teamFuel. when on, players can't refuel at fuelstations owned by
other teams. default off.
- new option keepShots. when on, mines/missiles/shots are made anonymous when
a player leaves the game, instead of being deleted. default off.
- new option wormTime. gives the number of seconds wormholes remain stable
ie. connected to the same out or inout wormhole. if 0, a new wormhole is
searched every time a player passes through it. other objects are only warped
while the wormhole is stable. if 0, a wormhole remains stable for other
objects until a new player passes through it. default is 0.
- alternate version of the recent (few months ago) nasty looking shots hack.
shots look like '+' or 'x' (chosen randomly).
- removed maxima for lasers and tractorbeams. maximum can now be set by
max<item>s up to 99.
- no more gaining lots of points by killing your own tanks. killing your own
tank now counts as killing a teammate.
- reimplemented deflector code. shots fired horizontally or vertically used
to gain extra speed from deflectors (no longer so) and more than ~15
deflectors were clipped by the collision checking code (also solved).
deflectors might feel a little different, but not much has changed.
- fixed yet another bug in Pick_team function which caused TEAM_NOT_SET to
be returned even if free bases were available.
- stun lasers now disable thrust, shields and shots for 5 frames. enforced
because robots could otherwise not be stunned (they turn those on again
every frame).
- having emergency shields on now allows you to fly against cannons without
dying. the cannon still dies and points are awarded as usual for killing
the cannon.
- heatseekers now target the engine instead of the center of the ship. this
seemed more realistic, but you don't see much of it.
- implosive mines/missiles are now as powerful as normal ones. they used to
be only half as powerful.
- hyperjumping leaves behind two wormholes, one inhole and one outhole. these
wormholes are removed after wormTime seconds or 1 second if wormTime is 0.
- robots try to thrust through walls less often. this was caused by a bug in
the goto cannon/fuelstation code.
- robots now go after loose balls too. before, they only navigated to treasure
boxes.
- Ben Aveling added a hack to make robots consider only 1000 objects to prevent
server lag when a nuke went off. this caused problems on some maps.
increased this number to 2000.
- robots now also use lasers on other players than the ones they are at war
with.
bugs left:
- cannon mode 3 doesn't check for walls between cannon and victim.
- temporary wormholes don't show up on screen even if wormholeVisible is on,
because they are not transmitted to the client.
- emergency don't protect you from treasure boxes.
- maximum number of tanks is still 8.
- team owned fuelstations or cannons are not displayed differently from normal
fuelstations/cannons. problem: the current way maps are encoded before
transmission to the client has not enough unused bits.
- lifetime of laser pulses is too short for reflected pulses to be harmful to
others.
- new deflector code seems to slow down the server when lots of objects are
present.
- cannons don't use defensive items yet.
- the new 'nasty' shot look is hardcoded, ie. there's no way to turn it off.
fun things to do:
- tell me what you think about everything I changed since v3.6.2. please. the
amount of feedback I get from others than the XPilot developers makes me
think I'm the only one who is actually interested in the time I put into
hacking XPilot.
- well, just try out everything this patch adds. I had a lot of fun coding it,
so I expect you (the people who play XPilot for a living) to have at least
as much fun using it.
Guido Koopman added some improvements over his big 4.0.0 patch:
adds/fixes the following:
- used RES/3 in Cannon_throw_items instead of CANNON_SPREAD.
- resetOnHuman only reset the game when the =< N-1st player entered
works as it should now, <= Nth.
- the object-out-of-map bug was apparently caused by items not having the
correct direction (tsin/tcos use lookup tables!). fixed in both Place_item
and Cannon_throw_items.
- Do_deflector was a little slow. rearranging the various tests and adding
a new one to get rid of unwanted objects quickly helps (profiling shows
50% less time spent in Do_deflector).
- tried to speed up robot code.
- check on BIT(pl->have, ITEM_DEFLECTOR) caused robots to almost always turn
on deflectors (even if itemDeflectorProb == 0). replaced ITEM_DEFLECTOR with
OBJ_DEFLECTOR.
- ROB_LOOK_AH was 2, even though robots check for possible collisions every
frame. removed ROB_LOOK_AH.
- possible use of ECMs and tanks as defense against missiles/mines was only
checked for when object was already within shield_range. checking is now
done earlier.
- replaced LENGTH() macro by QUICK_LENGTH() or VECTOR_LENGTH() in some
places.
fun things to do:
- use a profiler more often.
Guido Koopman contributed more small improvements:
adds/fixes the following:
- in various places, both Wrap_length and Wrap_findDir were used on the same
values.
replaced with WRAP_DX, WRAP_DY, LENGTH and findDir to eliminate overhead.
- removed some unused/unnecessary stuff from Make_debris and Make_wreckage.
- in cmdline.c, wormTime was declared as valSec, but it was used in update.c
as meaning ticks.
added '* FPS'.
- with wormTime set to 0, temporary wormholes were given a countdown of 1
tick. this caused problems when a wormhole was removed between a player
hitting it and being warped.
don't add wormholes when wormTime == 0 and remove both wormholes at once.
- added QUICK_LENGTH check to robot collision checking loop to get rid of
out-of-range objects quickly.
- robot collision checking on laser pulses could go wrong at map edges
because LENGTH was used instead of wrapped LENGTH.
- hyperjumping could cause problems when no empty space was found in 20
samples. w.x, w.y would then point to a wall or so.
hyperjumps to original location in that case now.
bugs left:
- there are still some strange tests in Make_debris/Make_wreckage which seem
unnecessary.
- should I have left that last 'problem' (^^) in as a feature?
fun things to do:
- run a profiler to see if XPilot is any faster now.
Added a debugaudio.c to the client to debug sound problems.
Changed the test for the existence of a xpilot host in the local DNS domain.
This slowed down startup considerably when DNS searches weren't limited to
one domain. Now if the local domain can be found only xpilot for that single
domain is tried.
Sven Mascheck contributed code to make the ERASE hack a client commandline option.
He also found out why sometimes the screen remains blue after an ECM.
Added Rakk & Murx's analytical collision detection patch.
Fixed a bug which would allow dead players to lock on any other player
even if lockOtherTeam was false.
Don't lock closest on paused players. Thanks to Sven Mascheck.
Added a new client option -ignoreWindowManager. It does what it says.
Handy if your window manager tends to uverrule your geometry value,
or if you think it spends to much time checking for window manager keys,
or if it doesn't handle colormap switching well.
===============================================================================
This is a summary of changes of release 4.0.0 compared to release 3.7.1
The contributed packages maps2image and TkXpInterface have been updated.
TkXpInterface has been updated to version 3.8.0 which supports the
Tcl/Tk versions 8.0 patchlevel 2.
Guido Koopman contributed a patch for automatically pausing idle players:
- Idle players are put in pause mode when they get killed. A player is idle
when there has been no keypress or -release for 60 seconds.
Guido Koopman contributed a patch for preventing base mining:
- a new server option: baseMineRange
When set, players are not allowed to drop or throw mines within this many
blocks from someone's homebase. The default is 0, which allows basemining.
Guido Koopman contributed a patch for a new item called deflector:
- a new item: deflector
When turned on, it pushes away objects that are moving towards your ship.
With more items, both the range and the force increase. The fuel cost does
not increase, just like with cloaks.
- a new server option: itemDeflectorProb
Default is "1e-9".
- a new server option: initialDeflectors
Default is "0".
- a new server option: maxDeflectors
Default is "10".
- a new client option: keyDeflector
Press to turn deflectors on. Press again to turn off.
The default key is 'o'.
- incompatibility with older (i.e. plain 3.7.1 and earlier) versions
This comes from a different packet for item counts and affects both the
server and the client. The new version number is 3503 (v3.5.0 and later
used 3500, 3501 is used by the hyperjump patch and 3502 by the phasing
device patch).
Guido Koopman contributed a patch for a new item called hyperjump:
- a new item: hyperjump
When used, it teleports the player to a random map location.
- a new server option: itemHyperJumpProb
Default is 1e-9.
- a new server option: initialHyperJumps
Default is 0.
- a new server option: maxHyperJumps
Default is 10.
- a new client option: keyHyperJump
Press key to activate and use a hyperjump.
Default key is 'q'.
- incompatibility with older (i.e. plain 3.7.1 and earlier) versions
This comes from a different packet for item counts and affects both the
server and the client. The new version number is 3501 (3.5.0 and later
used 3500).
Guido Koopman contributed a patch for programmer privileges:
- the owner of a server can always watch other players in pause mode, not
just when they are on the same team.
- the owner of a server always gets the full framerate, not 1/3.
- these are not meant for cheating, but for robot developers to see how the
robots actually behave.
- ownership of the server is determined by checking the login name and the
network address of the client. The login name should be the same as the
one used to start the server and the network address should be 127.0.0.1
ie. localhost.
Guido Koopman contributed a patch for a new item called phasing:
- a new item: phasing device
This item gives you four seconds in which you can fly through literally
everything. This in turn makes you invulnerable. On the other hand, you
also can't pick up items or balls, refuel, repair, fire, ecm, transport
or whatever, except fly around. Think of it as being in a different
dimension as the rest of the game. Nothing that happens there can affect
you, but you can't affect the game either. Any balls connected to your
ship when you turn on the phasing device are also phased. This only works
if they are connected by a solid line. The dashed connector lines are
broken by phasing. Any items you drop while phased are destroyed, whatever
the value of the loseItemdestroys server option. When the four second timer
(displayed in the same way as the emergency shield/thrust, shutdown and
selfdestruct timers) runs out, you drop back to normal space. Anything that
is still (partially) inside a wall when the timer runs out, is destroyed by
a wallcrash. A phased ship is drawn with dashed lines, like a cloaked ship,
but visible. You can still use shields (not useful) andcloaks when phased.
- a new server option: itemPhasingProb
Default is 1e-9.
- a new server option: initialPhasings
Default is 0.
- a new server option: maxPhasings
Default is 10.
- a new client option: keyPhasing
Pree key to activate and use some of your phasing.
Default key is 'p'.
- incompatibility with older (i.e. plain 3.7.0beta and earlier) versions
This comes from a different packet for item counts and affects both the
server and the client. The new version number is 3502 (v3.5.0 and later
used 3500 and 3501 is used by the hyperjump patch).
Guido Koopman contributed a patch for racemode:
- a new server option: roundDelay
This option gives the number of seconds the server waits before a new
round is started. This applies also to the first round. If a new player
logs in before the round has begun, the delay counter is restarted. The
default is "0". This keeps the first players to log in from starting
before the last one logs in.
- a new server option: maxRoundTime
This option sets the maximum duration of a round, in seconds. If the round
has not ended naturally before the timer expires, it is terminated. Scores
are updated as usual and a new round is started. The default is "0", which
means unlimited. This takes care of extremely slow racers (eg. robots), who
keep the others waiting too long.
- Both of these options work in unlimited lives mode too.
- Base order (race mode) is used when players log in the first time too.
Before, it was only used after each round.
Changed some default key mappings:
The keys moved do not affect actual gameplay. They are setup oriented.
This opens up some space on the keyboard for new "action" events.
Name old key(s) new key
Pause p Pause Pause
SelfDestruct q End
ToggleCompass y KP_7
ToggleOwnedItems o KP_8
ToggleMessages 0 KP_9
Deflector N/A o
HyperJump N/A q
Phasing N/A p
Rodney R. Korte ported XPilot to OS/2.
The hardcoded maxima for some player item limits have
been raised to values suggested by Rodney R. Korte.
Sven Mascheck contributed improvements for drawing the screen scaled
while the ERASE hack is active, including a bugfix for the WINSCALE macro.
Lewis W. Beard has released XPShipEditor, a Win32 ship shape editor written
in Visual Basic. As of XPilot-12, XPShipEditor will be included with the
Windows binaries distribution. XPShipEditor itself included the GPLed source
can be found at http://lwb.org/xpilot/
Guido Koopman contributed several more bugfixes and improvements:
- players could protect their teams' treasure by pausing while attached to it
pausing now releases the ball
- in unlimited races, all the players had to finish before the race was over
instead of just one player
replaced check on alive players by check on finished players
- robots sometimes chased an item too slowly
added check on harvest mode before slowing down
- robots didn't do anything with treasures
robots now try to connect to every ball they encounter
- robots didn't repair targets
actually they tried to, but didn't find any targets to repair, because
the distance was calculated the wrong way
distance now calculated correctly
- robots didn't dodge cannon bullets
because trying to dodge single bullets seemed overdone
now they do dodge them, because cannon bullets are quite dangerous
- robots often got stuck hanging under walls
I still don't know why they did, but I commented out some code that caused it
robot behavior didn't change, so it wasn't very important apparently
- someone (I believe is was Murx) complained that robots should be more
aggressive against players dragging balls
robots are now more likely to relock on those players
- robots attacked their own teammates
they don't anymore
- robots wasted time shooting targets of teams with no players
added check on playerlessness
- robots wasted time refueling when they didn't need to
problem only occurs when there's nothing to attack nearby
they now only refuel up to 90%
- players could get killed by cannon fire before a round started
disabled collision checking if the start is delayed
- robots detected cloaked ships even when cloakedExhaust was off
added check on cloakedExhaust
Guido Koopman contributed a patch for having robots in teammode:
- new server option: robotTeam
indicates which team to use for robots. default 0
- new server option: restrictRobots
when on, keeps robots from joining teams other than the one indicated with
the robotTeam option. default on
- new server option: reserveRobotTeam
when on, keeps human players from joining the team indicated by robotTeam.
default on
- before this patch, robots always left one base free for new human players.
they can now (if maxRobots is high enough) fill a server completely
that's not a problem, because robots are automatically kicked out when
a human player wants to join the game, even if other teams have free
bases left
- robots are kicked (if necessary) before paused humans
- if a human didn't indicate which team to join, it is selected for her
if teamAssign is on. in this selection, the robot team is avoided if
possible
cool things to do with this patch:
- practice your teamplay even when you're alone
unset restrictRobots and set maxRobots to an odd number
then join your favorite team (robots are autokicked if that team happens
to be full)
- find out what it is like to be on team 0
unset reserveRobotTeam and join team 0
- find out how good these new (since the fix380-3 patch) robots really are
set robotTeam to something other than 0 (on most maps, team 0 is either
not available or very different from the other teams) and turn on
restrictRobots to keep them from joining other teams, then join an other
team
- find bugs in the teampicking and autokicking code
try out weird settings for robotTeam, restrictRobots, reserveRobotTeam,
maxRobots, teamAssign and mapFile in combination with different numbers
of human players
Guido Koopman contributed several bugfixes:
- robots weren't kicked out of team games when a human player didn't specify
a team to join
- while firing, press phasingdevice key. when phasing ends, firing resumes
bug caused by keyrelease event being ignored while phased
only keypresses are ignored now
- robots that were killed in this round (limitedlives mode) are removed from
the game before the next round starts
I find it confusing to have different robots in my team every round, so
I disabled removal of robots in teamplay mode. use robotsleave options to
get robots out if you want
- robots weren't allowed to completely fill a server (ie. occupy all homebases)
I fixed that in the last fixpatch I submitted, but it was put back, probably
because of #1 (see ^)
- robots tried to kill phased ships
- robots picked up fuel items they didn't need
- robots didn't turn off cloaks when putting up shields
caused them to get shot/crash into wall when cloakedShield was off
- robots reconnected to balls they had just thrown
added 3 secs delay between detach and connect to any ball
- robots with large shipshapes got stuck on their bases
fixed by forcing thrust on when in homebase block
- phasingdevice/em.thrust/em.shield were turned off after each item ran out
next item is now automatically used
- implemented ball_handler in robotdef.c
bugs left:
- robots fly too fast when stealing treasures
- robots ignore balls outside treasureboxes
- robots ignore targets and other players when stealing treasures
- robots don't avoid balls (ie. they crash into them a lot)
- set rounddelay > 0 and before the counter reaches 0, turn on pause when
ships are replaced by numbers. for the rest of the round, all ships are that
number
- robot navigation still sucks (now also used in ballstealing code)
- I probably forgot some bugs
fun things to try:
- start server with a map with lots of open space and easy-to-get-to treasures
add a few (1..3) robots and enter pause mode. now watch the robots steal the
treasures and throw them into their own treasures. neat eh? works on maps
with less open space too, but you don't get long throws then
Guido Koopman (Big Dog)
Espen Skoglund contributed a patch for displaying ship wreckage parts when a
ship explodes.
The official XPilot map filename extension is now .xp.
The .map extension still works but is deprecated.
The fuel drainfactor feature from 3.7.0 is removed.
The client now strips any control characters from server status reports.
This should prevent garbage output on xterms (Winfried Truemper).
The server now better checks client input parameters like user names
and hostnames.
Servers report the number of free homebases for each team to the meta server.
The mouse pointer is now better centered in the talk window (Sven Mascheck).
Guido Koopman contributed more code:
adds/fixes the following:
- new client option: refreshMotd. When on, a fresh version of the server MotD
is requested every time it is displayed. Default off. Handy for those
servers with ranking tables in the MotD.
- color for wreckage changed from RED to WHITE. Since wreckage consists of
parts of ships, I thought it should look like that, and ships are white.
- targets are now correctly displayed in targetRadarColor (was wallRadarColor).
- removed constraint on radar colors which kept me from using red and white.
- removed PAUSEd check from TEAM() macro. It caused negative points for kills
made while paused (eg. with mines).
- new server option: allowViewing. When on, every player is allowed to watch
any other player when paused, waiting or dead. Default off. Because this is
meant for inexperienced players, the framerate is not lowered to 1/3.
- new server option: wreckageCollisionMayKill. When on, being hit by wreckage
unshielded is deadly. Default off. Don't kill players from too close when
shields are not allowed and this option is on.
- new server options to tune scoring for various kinds of killing:
option: desc.: default:
shotKillScoreMult ordinary bullet kills 1.0
torpedoKillScoreMult direct torpedo hits 1.0
smartKillScoreMult direct smart missile hits 1.0
heatKillScoreMult direct heatseeker hits 1.0
clusterKillScoreMult kills by cluster debris 1.0
laserKillScoreMult laser kills 1.0
tankKillScoreMult tank runovers 0.44
runoverKillScoreMult personal runovers 0.33
ballKillScoreMult ball kills 1.0
explosionKillScoreMult 'succumbed to explosion' 0.33
shoveKillScoreMult assistance in wallcrashes 0.5
The actual score is calculated by multiplying the rated score (10 points for
players with equal score, less for players with score lower than yours, etc.)
with this factor. The factor for tank kills is an approximation of what it
used to be (tank's score = player's score - 500).
- new server option: crashScoreMult. Default 0.33. For crashing with another
player.
- new server option: mineScoreMult. Default 0.17. For hitting a mine.
- new server option: cannonSmartness. Can take four values. 0 gives really
dumb cannons that only fire straight ahead when a player is near. 1 is the
default and gives normal cannons that fire in a random direction when a
player is near. 2 gives cannons that fire more or less (small random error)
in the direction of the player who activated the cannon (ie. the first player
who is within range). If the player is not in front of the cannon (cannondir
+/- 60 degrees) it fires as close as possible. 3 gives the smartest cannons
(for now) that fire at the nearest player in front of the cannon. If there
is no player in front of the cannon (defined as above), the cannon won't
fire. Also, it doesn't aim at the player directly, but calculates where that
player will be when the shot reaches her. The default is 1.
- new server option: cannonsPickupItems. When on, items that hit a cannon are
added to that cannon's inventory, instead of just passing through it. Default
is off.
- For now the only items useful to cannons are wideangles and sensors. Sensors
are used to increase the activation range for one cannon. Cannons get the
same items as players initially. The smarter the cannon, the narrower the
stream of bullets it fires with wideangles.
- ecms disable cannons temporarily. The formula used is the same as that for
players. Having more sensors decreases the time a cannon is disabled.
- added check on rdelay > 0 in LaserCollision. Not executing LaserCollision
caused laser pulses to remain in the air between rounds.
- half-used phasing devices are now counted as defensive items too. Before,
only unused phasing devices were counted.
- added check on rdelay == 0 to KEY_SHIELD release to keep players from
lowering shields before a round starts.
- moved cannon activating code from frame.c to update.c. Cannons didn't
activate on robot players which gave them an unfair advantage, especially
with smartcannons.
- removed NOT_CONNECTED check in round delay countdown in the hope of curing
the problem where going into pause causes ships to remain digits forever.
- changed maximum speed of cluster debris when DRAINFACTOR is defined to
5 * minimum speed. This lowers the drain from one cluster particle from
+/- 1275 to +/- 25 times the drain of a standard bullet. Still a lot, but
that's cluster debris.
- cannon explosions now also generate wreckage.
- added longterm_mode to default robot struct. The robot_mode was only used
for calculations in one frame. longterm_mode is supposed to give them a
little longer memory.
- added some extra checks to keep robots from going after useless items.
- robots now also hyperjump for torpedoes.
bugs left:
- I'm not sure I cured the ships turning into digits bug.
- I'm not sure about the radar colors (I have only tested it on a truecolor
display without DB).
- no options to tune scoring for player vs. world (ie. wallcrashes,
cannonkills, etc.) yet.
- smartcannons don't use anything fancy yet. The problem is that most items
can only be used by players.
- aiming of smartcannons in mode 3 is off by about 1/2 ship size, or 1/2
block_sz, or 3 degrees, or .. (ie. I don't know what causes it). It's most
noticeable with multiple wideangles.
- smartcannons don't activate correctly across map edges.
fun things to try:
- any help in making the storing of things like ecms, transporters and pulses
more general so it can be used for cannons, would be appreciated.
Guido Koopman contributed more bug fixes:
- activation distance for cannons changed
from 12*BLOCK_SZ to 0.5*VISIBILITY_DISTANCE
- smartcannons in mode 3 no longer fire on phased players
- semantics for server option maxRobots changed
maxRobots now specifies how many players there should be in total
if there are less then maxRobots players, robots will be added
- new server option: minRobots
specifies how many robots there should be at least
-1 sets minRobots to the same value as maxRobots
default is -1, which gives the old behaviour for maxRobots
- phasing device was not turned off when hoverpausing
fixed
- on some maps, when the robot manager tried to add a robot, the Pick_team()
function returned TEAM_NOT_SET even though there were free bases left
bug was caused by teamless bases being given to team 0 without increasing
the total number of bases for team 0
fixed
- deflectors are now turned on automagically when you have them
- Pick_team() used to avoid returning robotTeam if possible
Bert removed this, causing team 0 to be returned as first choice
put back old behaviour
- autopilot is now allowed when phased
bugs left:
- deflector code doesn't work OK with many (>10 or so) deflectors
- autopilot turns on emergency thrust, but doesn't turn it off again
- man pages are outdated
===============================================================================
==== $Id: ChangeLog,v 5.38.2.1 2001/11/04 21:05:19 dik Exp $
===============================================================================
