* do not use std::abs() because math.h includes one, that converts ints
to double. Use own iabs(), which directly uses labs() instead.
* don't use the word "gamma" as name for variables
* don't use a homebrew round() function
It bases mainly on 98d01d1f34
If a filename does not contain at least mode, map and timestamp (due to
messed up format strings), the default format is used to ensure unique
filenames.
server side: added server parameters
demo_max_number
number of demo files to hold in RAM
demo_save
enable/disable writing of demos to file
demo_path
demo file path and filename prefix
demo_filenameformat
format string to assemble demo information to a filename
%g : gamemode (int)
%G : gamemode (chr)
%F : gamemode (full)
%m : minutes remaining
%M : minutes played
%s : seconds remaining
%S : seconds played
%h : IP of server
%H : hostname of server (client only)
%n : mapName
%w : timestamp (formatted by demo_timestampformat)
demo_timestampformat
timestamp format string (see strftime() for format options)
if the string starts with 'U', UTC is used - otherwise local time
client side:
changed demonameformat and demotimeformat to SVARs
demonameformat
like demo_filenameformat
demotimeformat
like demo_timestampformat
removed demotimelocal getdemonameformat getdemotimeformat getdemotimelocal
DOCREF demonameformat demotimeformat
DOCREF demotimelocal getdemonameformat getdemotimeformat getdemotimelocal
The server keeps track of every player ID that ever connected successfully.
This database stores the last names and IPs used by the player and the
last countries the connections came from. Some stats are counted, like
"time connected" and "total tks". Most importantly, the database stores
rights (like admin) and restrictions (like ban). All rights and
restrictions are kept as dates (or timecodes), so, for example, if your
ban date is in the future, you are banned - and if it's in the past, you
can play. The collected stats are meant to help server owners at managing
players - especially, if they want to give someone admin rights.
The database is periodically saved to a file and read from that file on
startup. The default file is config/servervita.cfg and one backup is kept
in config/servervita_bak.cfg. During server start, if the regular file
can't be loaded, the backup is tried instead.
While the server is running, it checks periodically for a file named
config/servervita_update.cfg, and if the file exists, it is renamed to
a backup filename including a timecode and read and merged into the
database. The automatic rename makes sure, that a file is only read once.
Using such updates, a server owner can easily give several people admin
rights or for example, add clan tags for a whole list of IDs. (Yes, the
database also stores a clan tag for every ID...)
The database also stores two short comments for every ID. One can only be
read by server admins, but the other one is public. The public reason, for
example, can state a reason for a ban.
One database record takes up 264 bytes of RAM. During autosave, accounts,
that have not been used for a specified time (default 12 months), will
not be written. After a server restart, those old records will be gone.
Currently, there is no process in place, to also clean up old records
from RAM. Maybe, that'll be added later...
Uses own thread to actually write to logs, so the main thread can't be
blocked by a slow logfile. Log entries are buffered in two ringbuffers,
one for the main thread and one for all other threads. The main thread
will copy the second buffer to the main buffer periodically. The main
thread buffer is read from the log worker thread.
Since two buffers are kept, there are now several functions to add a log
entry:
mlog()
for main thread functions only: will never block
tlog()
other threads only: may block, until main thread copied some entries
xlog()
automatically determines, which thread it's been called from and
then writes to the proper buffer (will block accordingly)
To avoid improper usage, mlog and tlog verify, if they are used correctly.
(Only for debug builds. An assert triggers, if necessary.)
"readserverconfigsthread" gets triggered by poll_serverthreads() once a
minute to check the filesizes of all server config files. Altered files
get loaded into memory.
poll_serverthreads() in the main thread then processes the buffered files
one at a time. During processing, a semaphore is locked, so that other
threads can use the data structures derived from config files - if they
use the semaphore to lock the ressource.
Using the data from main thread can be done without using the semaphore,
since _all_ changes are done in the main thread.
If the main thread wants to update data that is locked by a thread, the
main thread will defer the changes for a minute. The main thread will
never lock.
almost 3000 lines of "errors" contained one actual memleak, one actually
unused variable and three actual ambiguous code parts
from cppcheck website: "The goal is to detect only real errors in the
code (i.e. have zero false positives)." :)
thx for finding five problems in our code
menuitemmapload mapname action [hoveraction [desc]]
create menu entry for a map file
menudirlistsub action dots
adds subdirectory entries to a menu with a files list
has to come after the menudirlist command
<action> is the command to be executed, when the menu entry is chosen
if <dots> is 1, an entry ".." is added
enumfiles path [extension]
lists files in a directory
if extension is specified, only files of that extension are listed
if the extension is "dir", only directories are listed
DOCREF menuitemmapload menudirlistsub enumfiles
SHA512 was adapted from a few public domain sources and pseudocode from
wikipedia. The entropy collector gathers entropy from keyboard events and
mouse movement - and, just for fun, from screenshots.
The entropy pool is saved at program exit and reloaded at program start.
Entropy is used to re-seed the pseudorandom generator (not that it would
make a difference).
* do not use std::abs() because math.h includes one, that converts ints
to double. Use own iabs(), which directly uses labs() instead.
* don't use the word "gamma" as name for variables
* don't use a homebrew round() function
According to SDL documentation, the SDL implementation of timedwait may
suck on some platforms, so it should only be used, where high CPU usage
during the wait is no issue.
getpckserver() will now only wait for one second for access to the
package server list and return an error instead of waiting indefinitely.
getmod modname
Try to download mods/modname.zip from all configured package servers.
Does not automatically activate the mod. The zip files are limited to
one megabyte total.
DOCREF getmod
* own source file for autodownload
* use cleaned up zip functions
* use new http get function (no more curl)
* enforce strict file naming rules and file size limits
* ping all package servers on every game start to determine ping times
* support media file update lists (also kept on package servers)
* allow users to manually prioritise package servers
* automatically select server with newest media file revision
new or changed commands:
addpckserver hostname [priority]
Adds a host name to the list of package servers. If a priority is given,
it influences the sorting of servers. Servers with higher priority are
queried first. If servers have the same priority, they are sorted by
ping. Default priority is zero.
getpackageserver
Returns a table with four columns of all configured package servers. The
columns are: host name, priority, ping and resolved.
"resolved" is a flag, indicating, if the server answered the ping during
game start.
removed cubescript idents: sortpckservers gamestarts
DOCREF addpckserver getpckserver
* change priority of zip packages from "highest" to "just below home dir"
* strict naming rules for zip packages
* require all zips to be below directory "mods/"
* only allow zips to intoduce files below "packages/" (or with a special
zip package name also below "config/")
* make zip package config persistent (saved to init.cfg)
WARNING: intermediate commit - autodownload system defunct!
DOCREF (removed) addzip removezip
Check printf-like functions like conoutf, if format strings and argument
lists match.
This commit also fixes a lot of code parts that actually had mismatching
argument lists - or did things, that this check does not like.
numtime() was rewritten in an ugly but very portable fashion (it's
recommended to stay away from "long long" format entries, because errors
usually only come up at runtime).
(some code was copied from sauer)
"findfilelocation" contains the location, where the file was found.
Not a nice solution, but nice enough :)
Note: if the location is FFL_WORKDIR, the existence of the file has not
actually been checked yet.