added the support for XML-RPC and JSON-RPC to easier control the server
from other applications;
added web-interface - it is available for download from NZBGet-project's
home page as a separate package "web-interface";
added the automatic cleaning up of the download queue (deletion of unneeded
paused par-files) after successful par-check/repair - new
option "ParCleanupQueue";
added option "DetailTarget" to allow to filter the (not so important)
log-messages from articles' downloads (they have now the type "detail"
instead of "info");
added the gathering of progress-information during par-check; it is
available via XML-RPC or JSON-RPC; it is also showed in web-interface;
improvements in internal decoder: added support for yEnc-files without
ypart-statement (sometimes used for small files); added support for
UU-format;
removed support for uulib-decoder (it did not work well anyway);
replaced the option "decoder (yenc, uulib, none)" with the option
"decode (yes, no)";
added detection of errors "server busy" and "remote server not available"
(special case for NNTPCache-server) to consider them as connect-errors
(and therefore not count as retries);
added check for incomplete articles (also mostly for NNTPCache-server) to
differ such errors from CrcErrors (better error reporting);
improved error-reporting on moving of completed files from tmp- to
dst-directory and added code to move files across drives if renaming fails;
improved handling of nzb-files with multiple collections in par-checker;
improved the parchecker: added the detection and processing of files
splitted after parring;
added the queueing of post-process-scripts and waiting for script's
completion before starting of a next job in postprocessor (par-job or
script) to provide more balanced cpu utilization;
added the redirecting of post-process-script's output to log; new option
"PostLogKind" to specify the default message-kind for unformatted
log-messages;
added the returning of script-output by command "postqueue" via XML-RPC
and JSON-RPC; the script-output is also showed in web-interface;
added the saving and restoring of the post-processor-queue (if server was
stopped before all items were processed); new option "ReloadPostQueue";
added new parameter to postprocess-script to indicate if any of par-jobs
for the same nzb-file failed;
added remote command (switch O/--post) to request the post-processor-queue
from server;
added remote command (switch -W/--write) to write messages to server's log;
added option "DiskSpace" to automatically pause the download on low disk
space;
fixed few incompatibility-issues with unslung-platform on nslu2 (ARM);
fixed: articles with trailing text after binary data caused the decode
failures and the reporting of CRC-errors;
fixed: dupecheck could cause seg.faults when all articles for a file failed;
fixed: by dupe-checking of files contained in nzb-file the files with the
same size were ignored (not deleted from queue);
updated libpar2-patch for msvc to fix a segfault in libpar2 (windows only);
fixed: by registering the service on windows the fullpath to nzbget.exe
was not always added to service's exename, making the registered service
unusable;
fixed: the pausing of a group could cause the start of post-processing for
that group;
fixed: too many info-messages "Need more N blocks" could be printed during
par-check (appeared on posix only);
nzbget-0.3.1 - 1 Feb. 2008
Greatly reduced the memory consumption by keeping articles' info on disk
until the file download starts;
Implemented decode-on-the-fly-technique to reduce disk-io; downloaded
and decoded data can also be saved directly to the destination file
(without any intermediate files at all); this eliminates the necessity
of joining of articles later (option "DirectWrite");
Improved communication with news-servers: connections are now keeped open
until all files are downloaded (or server paused); this eliminates the
need for establishing of connections and authorizations for each
article and improves overal download speed;
Significantly better download speed is now possible on fast connection;
it was limited by delays on starting of new articles' downloads;
the synchronisation mechanism was reworked to fix this issue;
Download speed meter is much more accurate, especially on fast connections;
this also means better speed throttling;
Speed optimisations in internal decoder (up to 25% faster);
CRC-calculation can be bypassed to increase performance on slow CPUs
(option "CrcCheck");
Improved parsing of artcile's subject for better extracting of filename
part from it and implemented a fallback-option if the parsing was incorrect;
Improved dupe check for files from the same nzb-request to detect reposted
files and download only the best from them (if option "DupeCheck" is on);
Articles with incorrect CRC can be treated as "possibly recoverable errors"
and relaunched for download (option "RetryOnCrcError"), it is useful if
multiple servers are available;
Improved error-check for downloaded articles (crc-check and check for
received message-id) decreases the number of broken files;
Extensions in curses-outputmode: added group-view-mode (key "G") to show
items in download queue as groups, where one group represents all files
from the same nzb-file; the editing of queue works also in group-mode
(for all files in this group): pause/unpause/delete/move of groups;
Other extensions in curses-outputmode: key "T" toggles timestamps in log;
added output of statistical data: uptime, download-time, average session
download speed;
Edit-command accepts more than one ID or range of IDs.
E.g: "nzbget -E P 2,6-10,33-39"; The switch "-I" is not used anymore;
Move-actions in Edit-command affect files in a smart order to guarantee
that the relative order of files in queue is not changed after the moving;
Extended syntax of edit-command to edit groups (pause/unpause/delete/move
of groups). E.g: "nzbget -E G P 2";
Added option "DaemonUserName" to set the user that the daemon (POSIX only)
normally runs at. This allows nzbget daemon to be launched in rc.local
(at boot), and download items as a specific user id; Thanks to Thierry
MERLE (merlum@users.sourceforge.net) for the patch;
Added option "UMask" to specify permissions for newly created files and dirs
(POSIX only);
Communication protocol used between server and client was revised to define
the byte order for transferred data. This allows hosts with different
endianness to communicate with each other;
Added options "CursesNzbName", "CursesGroup" and "CursesTime" to define
initial state of curses-outputmode;
Added option "UpdateInterval" to adjust update interval for Frontend-output
(useful in remote-mode to reduce network usage);
Added option "WriteBufferSize" to reduce disk-io (but it could slightly
increase memory usage and therefore disabled by default);
List-command prints additional statistical info: uptime, download-time,
total amount of downloaded data and average session download speed;
The creation of necessary directories on program's start was extended
with automatic creation of all parent directories or error reporting
if it was not possible;
Printed messages are now translated to oem-codepage to correctly print
filenames with non-english characters (windows only);
Added remote-command "-V (--serverversion)" to print the server's version;
Added option "ThreadLimit" to prevent program from crash if it wants to
create too many threads (sometimes could occur in special cases);
Added options "NzbDirInterval" and "NzbDirFileAge" to adjust interval and
delay by monitoring of incoming-directory for new nzb-files;
Fixed error on parsing of nzb-files containing percent and other special
characters in their names (bug appeared on windows only);
Reformated sample configuration file and changed default optionnames
from lowercase to MixedCase for better readability;
Few bugs (seg faults and memory leaks) were fixed.
nzbget-0.3.0 - 23 Nov. 2007
The download queue now contains newsgroup-files to be downloaded instead of
nzb-jobs. By adding a new job, the nzb-file is immediately parsed and each
newsgroup-file is added to download queue. Each file can therefore be
managed separately (paused, deleted or moved);
Current queue state is saved after every change (file is completed or the
queue is changed - entries paused, deleted or moved). The state is saved on
disk using internal format, which allows fast loading on next program start
(no need to parse xml-files again);
The remaining download-size is updated after every article is completed to
indicate the correct remaining size and time for total files in queue;
Downloaded articles, which are saved in temp-directory, can be reused on
next program start, if the file was not completed (option "continuepartial"
in config-file);
Along with uulib the program has internal decoder for yEnc-format. This
decoder was necessary, because uulib is so slow, that it prevents using of
the program on not so powerful systems like linux-routers (MIPSEL CPU 200
MHz). The new decoder is very fast. It is controlled over option "decoder"
in config-file;
The decoder can be completely disabled. In this case all downloaded articles
are saved in unaltered form and can be joined with an external program;
UUDeview is one of them;
If download of article fails, the program attempts to download it again so
many times, what the option "retries" in config-file says. This works even
if no servers with level higher than "0" defined. After each retry the next
server-level is used, if there are no more levels, the program switches to
level "0" again. The pause between retries can be set with config-option
"retryinterval";
If despite of a stated connection-timeout (it can be changed via
config-option "connectiontimeout") connection hangs, the program tries to
cancel the connection (after "terminatetimeout" seconds). If it doesn't
work the download thread is killed and the article will be redownloaded in
a new thread. This ensures, that there are no long-time hanging connections
and all articles are downloaded, when a time to rejoin file comes;
Automatic par-checking and repairing. Only required par-files are downloaded.
The program uses libpar2 and does not require any external tools. The big
advantage of library is, that it allows to continue par-check after new
par-blocks were downloaded. This were not possible with external
par2cmdline-tool. New configuration options "loadpars", "parcheck" and
"parrepair";
There is a daemon-mode now (command-line switch "-D" (--daemon)). In this
mode a lock-file (default location "/tmp/nzbget.lock", can be changed via
option "lockfile") contains PID of daemon;
The format of configuration-file was changed from xml to more common
text-format. It allows also using of variables like
"tempdir=${MAINDIR}/tmp";
Any option of config-file can be overwritten via command-line switch
"-o" (--option). This includes also the definition of servers.
This means that the program can now be started without a configuration-file
at all (all required options must be passed via command-line);
The command-line switches were revised. The dedicated switches to change
options in config-file were eliminated, since any option can now be changed
via switch "-o" (--option);
If the name of configuration-file was not passed via command-line the
program search it in following locations: "~/.nzbget", "/etc/nzbget.conf",
"/usr/etc/nzbget.conf", "/usr/local/etc/nzbget.conf",
"/opt/etc/nzbget.conf";
The new command-line switch "-n" (--noconfigfile) prevents the loading of
a config-file. All required config-options must be passed via command-line
(switch "-o" (--option));
To start the program in server mode either "-s" (--server) or
"-D" (--daemon) switch must be used. If the program started without any
parameters it prints help-screen. There is no a dedicated switch to start
in a standalone mode. If switches "-s" and "-D" are omitted and none of
client-request-switches used the standalone mode is default. This usage
of switches is more common to programs like "wget". To add a file to
server's download queue use switch "-A" (--append) and a name of nzb-file
as last command-line parameter;
There is a new switch "-Q" (--quit) to gracefully stop server. BTW the
SIGKIL-signal is now handled appropriately, so "killall nzbget" is also OK,
where "killall -9 nzbget" terminates server immediately (helpful if it
hangs, but it shouldn't);
With new switch "-T" (--top) the file will be added to the top of download
queue. Use it with switch "-A" (--append);
The download queue can be edited via switch "-E" (--edit). It is possible
to pause, unpause, delete and move files in queue. The IDs of file(s)
to be affected are passed via switch "-I" (fileid), either one ID or a
range in a form "IDForm-IDTo". This also means, that every file in queue
have ID now;
The switch "-L" (--list) prints IDs of files consequently. It prints also
name, size, percentage of completing and state (paused or not) of each file.
Plus summary info: number of files, total remaining size and size of
paused files, server state (paused or running), number of threads on
server, current speed limit;
With new switch "-G" (--log) the last N lines printed to server's
screen-log, can be printed on client. The max number of lines which can
be returned from servers depends on option "logbuffersize";
The redesigned Frontends (known as outputmodes "loggable", "colored" and
"curses") can connect to (remote) server and behave as if you were running
server-instance of program itself (command-line switch "-C" (--connect)).
The log-output updates constantly and even all control-functions in
ncurses-mode works: pause/unpause server, set download rate limit, edit of
queue (pause/unpause, delete, move entries). The number of connected
clients is not limited. The "outputmode" on a client can be set
independently from server. The client-mode is especially useful if the
server runs as a daemon;
The writing to log-file can be disabled via option "createlog".
The location of log-file controls the option "log-file";
Switch "-p" (--printconfig) prints the name of configuration file being
used and all option/value-pairs, taking into account all used
"-o" (--option) - switches;
The communication protocol between server and client was optimized to
minimize the size of transferred data. Instead of fixing the size for
Filenames in data-structures to 512 bytes only in fact used data
are transferred;
Extensions in ncurses-outputmode: scrolling in queue-list works better,
navigation in queue with keys Up, Down, PgUp, PgDn, Home, End.
Keys to move entries are "U" (move up), "N" (move down), "T" (move to top),
"B" (move to bottom). "P" to pause/unpause file. The size, percentage
of completing and state (paused or not) for every file is printed.
The header of queue shows number of total files, number of unpaused
files and size for all and unpaused files. Better using of screen estate
space – no more empty lines and separate header for status (total seven
lines gain). The messages are printed on several lines (if they not fill
in one line) without trimming now;
configure.ac-file updated to work with recent versions of autoconf/automake.
There are new configure-options now: "--disable-uulib" to compile the
program without uulib; "--disable-ncurses" to disable ncurses-support
(eliminates necessity of ncurses-libs), useful on embedded systems with
little resources; "--disable-parcheck" to compile without par-check;
The algorithm for parsing of nzb-files now uses XMLReader instead of
DOM-Parser to minimize memory usage (no more needs to build complete DOM-tree
in memory). Thanks to Thierry MERLE <merlum@users.sourceforge.net> for
the patch;
The log-file contains now thread-ID for all entry-types and additionally
for debug-entries: filename, line number and function's name of source
code, where the message was printed. Debug-messages can be disabled in
config-file (option "debugtarget") like other messages;
The program is now compatible with windows. Project file for MS Visual
C++ 2005 is included. Use "nzbget -install" and "nzbget -remove" to
install/remove NZBGet-Service. Servers and clients can run on diferrent
operating systems;
Improved compatibility with POSIX systems; Tested on:
- Linux Debian 3.1 on x86;
- Linux BusyBox with uClibc on MIPSEL;
- PC-BSD 1.4 (based on FreeBSD 6.2) on x86;
- Solaris 10 on x86;
Many memory-leaks and thread issues were fixed;
The program was thoroughly worked over. Almost every line of code was
revised;
nzbget-0.2.3
Fixed problem with losing connection to newsserver after too long idle time
Added functionality for dumping lots of debug info
nzbget-0.2.2
Added Florian Penzkofers fix for FreeBSD, exchanging base functionality in
SingleServerPool.cpp with a more elegant solution
Added functionality for forcing answer to reloading queue upon startup of
server
+ use -y option to force from command-line
+ use "reloadqueue" option in nzbget.cfg to control behavior
Added nzbget.cfg options to control where info, warnings and errors get
directed to (either screen, log or both)
Added option "createbrokenfilelog" in nzbget.cfg
nzbget-0.2.1
Changed and extended the TCP client/server interface
Added timeout on sockets which prevents certain types of nzbget hanging
Added Kristian Hermansen's patch for renaming broken files
nzbget-0.2.0
Moved 0.1.2-alt4 to a official release as 0.2.0
Small fixes
nzbget-0.1.2-alt4
implemented tcp/ip communication between client & server (removing the
rather defunct System V IPC)
added queue editing functionality in server-mode
nzbget-0.1.2-alt1
added new ncurses frontend
added server/client-mode (using System V IPC)
added functionality for queueing download requests