Commit 5a7c4f3 broke the web interface. See comment on the commit for more details.

I finally tried gerbera on a real machine. Is anyone having a problem importing videos & pictures using the default import.js / common.js scripting? My gerbera/config.xml has the virtual-layout type="js". My gerbera log file says:

JS: /usr/share/gerbera/js/import.js:154: ReferenceError: object_root_path is not defined

I tried using some print commands in import.js and it appears that the object_root_path is indeed empty (obj.location is fine). Just wondering if either I did something odd when I compiled or if something isn't working right. I can run some debug cases - just let me know what to do so I can get the data needed to help.

config/config.xsd only validates configuration version 1, but not config v2 as it is generated by gerbera.

A current XSD file should be shipped.

• Default configuration as generated by gerbera: http://p.cweiske.de/433
• validate with: $ xmllint --schema config/config.xsd ~/.gerbera/config.xml
• XSD primer: https://www.w3.org/TR/xmlschema-0/

Our spidermonkey code is very old, as is the version of spidermonkey it needs.

Its also way over the top for our scripting needs.

Lets switch to something smaller and designed to be embedded such as Duktape or MuJS.

Hi,

I've got one question regarding the master branch and ubuntu 16.10.
I am not able to build it errorfree!
Although ffmpeg and libav there is the following error with cmake:

-- Could NOT find LibAV (missing:  LIBAV_ROOT_DIR LIBAV_AVDEVICE_LIBRARY LIBAV_AVFILTER_LIBRARY) (found version "ffmpeg version 3.0.7-0ubuntu0.16.10.1 Copyright (c) 2000-2017 the FFmpeg developers
built with gcc 6.2.0 (Ubuntu 6.2.0-5ubuntu12) 20161005

CMake Error at CMakeLists.txt:481 (message):
  FFMpeg/LibAV not found

Is there any kind of solution which already came up?

Thanks a lot in advance and BR,
count

We need it, especially for dc:creator, which is currently set to the first tracks artist. Which can sometimes look odd, especially on compilations.

I changed my import script, stopped gerbera, emptied all SQL database tables and started gerbera:

INFO: database doesn't seem to exist. automatically creating database...
ERROR: Mysql: error while creating db: mysql_error (1050): "Table 'mt_cds_object' already exists"
Exception raised in [src/storage/mysql/mysql_storage.cc:226] init(): Mysql: error while creating db: mysql_error (1050): "Table 'mt_cds_object' already exists"

Gerbera should not assume that no tables exist when they are simply empty.

I removed -DWITH_AVCODEC=1 from the cmake command and still got an error:

CMake Error at CMakeLists.txt:494 (message):
FFMpeg/LibAV not found

I had to pass the parameter with value 0 to overcome this. If it's really optional, it should default to 0 in my eyes.

tools.cc: swap interfaceToIP to ipToInterface
server.cc: call UpnpInit2 with an interface name

Done?

The second time in a week I had the following error in my logs:

Mai 05 18:31:34 dojo gerbera[15713]: Exception raised in [src/storage/mysql/mysql_storage.cc:365] exec(): Mysql: mysql_real_query() failed: mysql_error (2006): "MySQL server has gone away"; query: UPDATE mt_autoscan SET last_modified=1486286336 WHERE id=1

Maybe a timeout. Gerbera should reconnect then.

Need to update the install paths etc.

Do we install a mediatomb symlink perhaps?

Hello All,

I am running a recent local build of Gerbera on my local and hitting a deadlock consistently when the SessionManager.checkTimer() clears the stale sessions.

Here is the scenario.

1. Open Gerbera web UI @ http://localhost:49152
2. Establish a session (with SID and logged in)
3. Wait for SessionManager::timerNotify to fire, and cleanup the stale sessions

It looks like the Timer mutex is AutoLock 2 times. First on the Timer::notify, to call the SessionManager cleanup, then again on the Timer::removeTimerSubscriber

void  Timer::notify()
{
    AutoLock lock(mutex);

Sessions are removed via SessionManager::timerNotify and the SessionManager::checkTimer is called to clear the Timer.

void Timer::removeTimerSubscriber(Subscriber* timerSubscriber, zmm::Ref<Parameter> parameter, bool dontFail)
{
    log_debug("Removing subscriber...\n");

    AutoLock lock(mutex);

   ^^^^^ DEADLOCK ^^^^^

The deadlock consistently happens on the second call to the AutoLock lock(mutex) for the Timer instance.

My guess is that I have missed some simple configuration or runtime difference, but cannot figure it out. Likely due to inexperience with C++.

Any helpful direction would be appreciated. If anyone else sees similar issue and has ideas it would be great!.

Thanks,

Eamonn

On a new scan of the music files with a 11dea73 build, all the artist names, album titles, and years show up as "Unknown". This happens when import.js is used.

My import.js tries to obtain the artist from obj.meta[M_ARTIST] with fall-back to obj.aux["TPE2"]. The config.xml file contains <add-data tag="TPE2"/> for both id3 and libextractor.

If you would prefer that we do not use Issues to ask questions, feel free to delete this.

I have been trying to get mediatomb working and have hit a problem, so I am wondering if this also occurs in gerbera (it isn't mentioned on the list of differences).

I have a large (500 tracks) playlist, which is in a specific order. All the clients I have tried with mediatomb (and the mediatomb web interface itself) show the list in alphabetical order, even though the playlist.js is setting the item.playlistOrder correctly. I can't be certain whether this is mediatomb's problem, or the clients are re-sorting.

Does anyone use playlists with specific ordering in gerbera and do they work? If they do, I will make the effort to build this on debian testing and give it a go.

My config.xml file specifies that server port 49152 shall be used. When restarting gerbera, it switches to port 49153. When restarting again, it goes to 49152 again. When waiting some time until starting gerbera after stopping it, the port is not switched to 49153.

This was already a problem with mediatomb. Maybe you could fix this for gerbera.

@whorfin

What is the "correct" way to get the currently hardcoded solaris flags? pkg-config or something else? Do you know any other CMake-using packages that detect these sorts of things?

Hello, I've just came across this project and found out it's quite useful so I'm planning to use it on my Raspberry Pi which is running ARM arch, the compiling process of this require a lot of unnecessary dependency that needs to be installed so I think it's a good idea to add arm* to ppa builds.

I tried to use add a cover.jpg image item with the web interface.

Data:

• Select folder
• "add Item"
• type: Item
• title: cover.jpg
• location: /var/spool/music/Hörspiele/cover.jpg
• class: object.item.imageItem
• description: (none)
• mime type: image/jpeg

When clicking "add item", gerbera crashes:

2017-05-05 23:17:33   ERROR: Exception caught: Script: failed to execute script
Exception raised in [src/scripting/script.cc:304] _execute(): Script: failed to execute script
2017-05-05 23:17:33   ERROR: Exception caught: Script: failed to execute script
Exception raised in [src/scripting/script.cc:304] _execute(): Script: failed to execute script
gerbera: /usr/local/src/gerbera/src/storage/sql_storage.cc:223: zmm::Ref<CdsObject> SQLStorage::checkRefID(zmm::Ref<CdsObject>): Zusicherung »0« nicht erfüllt.
systemd[1]: gerbera.service: main process exited, code=killed, status=6/ABRT
systemd[1]: Unit gerbera.service entered failed state.

"Zusicherung »0« nicht erfüllt" is german and emeans "assertion 0 not fulfilled".

I didn't notice this at first as i've been manually installing the binary, but finally traced this down:

in the build dir,
ldd gerbera
[all the lovely paths are correct]

ldd CMakeFiles/CMakeRelink.dir/gerbera
libupnp and libixml show as (file not found)

The CMakeFiles/gerbera.dir/relink.txt file is missing the "-Wl,-R..." paths found in link.txt, and thus the
loader paths are missing

It's not obvious to me how the cmake config is causing this to happen, but it is, and it's a problem.

So having spoken to the original authors of MediaTomb they would rather keep the name mediatomb associated with them, which is fair enough.

So we need a new name for the project going forward....

Sometimes SIGHUP is completely ignored.
Often, SIGHUP messes with the configured port.

[anecdotal reports over in #32 ]

In quick summary, one can observe in the logfile, and confirm as behavior, that the "Initialized port" is incremented, and not the actual requested port, each time a SIGHUP is sent to mediatomb. Eventually the requested port will be selected, if luck smiles upon one.

Also src/storage/sql_storage.cc:936 sometimes throws an exception

Reformat the source to match WebKit C++ style guidelines and document it.

• Its sensible.
• Its is well documented.
• Its well supported by tooling.
• It mostly matches what is already being used.

We should also provide a .clang-format file, or set of astyle parameters.

I've added the PPA, but it's unclear what to do next. apt install gerbera didn't work. apt install mediatomb seems to install the old fork. After that, launching MediaTomb from the shortcut that gets installed, navigates me to file:///var/lib/mediatomb/mediatomb.html, which is missing. Launching mediatomb from a terminal seems to launch the old project.

I incorrectly associated this with SIGHUP because of bundled changes.
While I'm not positive, since the reported version in logging doesn't include git revision [maybe it should?] I believe this showed up with 49a9e4d

I know that I started seeing the bad behavior along with the "event is ok" message which was introduced very shortly after that commit. 49a9e4d changed a bunch of
things related to the autoscan timer, hence my suspicion.

I've created a test case where a subdir of an autoscan directory has a permission problem as far as the mediatomb server process is concerned. So I can see a heartbeat of
ERROR: Failed to stat
those heartbeats only happen once at the start of mediatomb now, rather than approximately every autoscan interval

Current git (f82bbf5) gives me an SQL error when I try to browse the UPnP server with either BubbleUPnP or ControlDLNA (both android apps):

Apr 13 23:46:19 dojo gerbera[688]: 2017-04-13 23:46:19    
INFO: Exception: Mysql: mysql_real_query() failed: mysql_error (1054):
"Unknown column 'cover.jp%' in 'where clause'"; query:
SELECT `id`
 FROM `mt_cds_object`
 WHERE ( lower(`dc_title`) LIKE`cover.jp%`
   OR lower(`dc_title`) LIKE`albumart%.jp%`
   OR lower(`dc_title`) LIKE`album.jp%`
   OR lower(`dc_title`) LIKE`front.jp%`
   OR lower(`dc_title`) LIKE`folder.jp%` )
  AND `upnp_class`=`object.item.imageItem`
  AND `parent_id` IN (SELECT `parent_id` FROM `mt_cds_object` WHERE `parent_id`=`564` AND `object_type`=`2`)
 ORDER BY `dc_title` DESC

This is with mysql 5.5.

There is an appetite for a modern UI, its high on the wish list.

In terms to technologies ideally for simplicity I would like a SPA that has no compilation requirement, BS4 is ideal, and uses jquery so perhaps that would be enough in terms of libraries for this case.

Because everyone loves colored build output, right?

When importing content using the import.js script, before being passed to the script, every file is first converted into a JS object called orig which contains all the file metadata.

When importing content using the playlists.js script, all we have to go on is the path to the media file. For paths to local files, it would be useful if there was a way to convert that file path into a JS object with metadata (similar to how the import.js script works). That would make it possible to get to artist and album data without resorting to extracting that info out of the path string.

Ex:

var audio = getCdsObject(path_to_audio_file);
print(audio.meta[M_ARTIST] + " " + audio.meta[M_ALBUM]);

I tried to build gerbera on macOS (10.12.4). It fails because not all dependencies can be satisfied. Please have a look.

$ brew install libexif libmagic lzlib mp4v2 spidermonkey sqlite taglib ffmpeg ffmpegthumbnailer id3lib expat libupnp ossp-uuid mysql

$ git clone https://github.com/v00d00/gerbera.git
$ mkdir build
$ cd build
$ cmake ../ -DWITH_MAGIC=1 -DWITH_MYSQL=1 -DWITH_CURL=1 -DWITH_JS=1 \
-DWITH_TAGLIB=1 -DWITH_AVCODEC=1 -DWITH_EXIF=1 -DWITH_LASTFM=0
-- The C compiler identification is AppleClang 8.1.0.8020042
-- The CXX compiler identification is AppleClang 8.1.0.8020042
-- Check for working C compiler: /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/cc
-- Check for working C compiler: /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/cc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Check for working CXX compiler: /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/c++
-- Check for working CXX compiler: /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/c++ -- works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Looking for pthread.h
-- Looking for pthread.h - found
-- Looking for pthread_create
-- Looking for pthread_create - found
-- Found Threads: TRUE
-- Looking for nl_langinfo
-- Looking for nl_langinfo - found
-- Looking for setlocale
-- Looking for setlocale - found
-- Could NOT find INOTIFY (missing:  INOTIFY_INCLUDE_DIR)
-- Performing Test ICONV_SECOND_ARGUMENT_IS_CONST
-- Performing Test ICONV_SECOND_ARGUMENT_IS_CONST - Failed
-- Found Iconv: /usr/lib/libiconv.dylib
-- Found PkgConfig: /usr/local/bin/pkg-config (found version "0.29.2")
-- Checking for one of the modules 'uuid'
-- Could NOT find UPnP (missing:  UPNP_UPNP_LIBRARY UPNP_INCLUDE_DIR)
CMake Error at CMakeLists.txt:389 (message):
  gerbera requires libupnp 1.8.0 or above.

-- Configuring incomplete, errors occurred!
See also "/Users/tristank/Documents/Code/git/gerbera/build/CMakeFiles/CMakeOutput.log".
See also "/Users/tristank/Documents/Code/git/gerbera/build/CMakeFiles/CMakeError.log".

Hi! First time poster so please be gentle. I'm trying to get mediatomb_v00d00 working on a headless debian-stretch.
Part 1 - I tried cross compiling the ubuntu PPA (dated 3/14) into debian but it had a problem. I tried compiling with the method on the main project page here and the error seems the be the same. See the attached cmakeerror.log file.
CMakeError.log.txt

Part 2 - Late last year, Debian changed their mysqlclient-dev to default-mysqlclient-dev. There were some BTS reported on the debian version of mediatomb saying that the debian/control file should be updated. Can you please confirm which is the proper one I should be using?

Part 3 - Not sure if this is an repacked ubuntu compiling artifact or not. A few weeks back I successfully compiled the ubuntu ppa into debian. I was able to get the mediatomb-daemon to work - but it only worked after I installed exim4. The daemon stil worked after I removed exim4. Any clue on what that is? I compiled the debian MT and it worked without the exim4 installed.

Part 4 - Is it worthwhile to add net-tools as a dependency? The daemon needs it - and a CLI machine doesn't include net-tools by default.

Sorry so many questions.
Tommyp

And include git rev hash.

I have several albums with songs sung by multiple artists. The ogg and mp3 files have multiple artist fields.

In the "Artists" tree, entries named "artist1 artist2" are created instead of 2 artist entries that contain both the same songs.

Screenshot:

Those with combined names are duets, the other ones have only one artist tag.

I was wondering if I can just remove the old mediatomb, clone the master (your fork) and start mediatomb.

Do you use the same tables as before?

Thanks a lot for this fork!!!!

I have some scripts that push data into mediatomb/gerbera via the web interface AJAX API, using my PHP library Services_MediaTomb. After switching to Gerbera, they fail and gerbera crashes because an assertion fails.

The following URLs are called:

http://dojo:49152/content/interface?req_type=auth&action=get_sid&sid=null
http://dojo:49152/content/interface?req_type=auth&action=get_token&sid=866cc65df0f5ef9876cdbac783bcb828
http://dojo:49152/content/interface?sid=866cc65df0f5ef9876cdbac783bcb828&req_type=auth&action=login&username=cweiske&password=b37098fdcdedca1257c81248bcdc9d1f
http://dojo:49152/content/interface?sid=866cc65df0f5ef9876cdbac783bcb828&req_type=containers&parent_id=0&select_it=0
http://dojo:49152/content/interface?sid=866cc65df0f5ef9876cdbac783bcb828&req_type=containers&parent_id=39338&select_it=0
http://dojo:49152/content/interface?sid=866cc65df0f5ef9876cdbac783bcb828&req_type=add_object&parent_id=39339&title=item+%230&location=%2Ftmp%2Fmediatombtest%2Ffile.txt&class=object.item&obj_type=item&description=Just+a+test&mime-type=text%2Fplain

After the last URL call, gerbera crashes:

Apr 16 21:30:25 dojo gerbera[29527]: gerbera: /usr/local/src/gerbera/src/storage/sql_storage.cc:225: zmm::Ref SQLStorage::checkRefID(zmm::Ref): Zusicherung »0« nicht erfüllt.

It's german but the message says "Assertion »0« not fulfilled".

The text file that gets added actually exists and has a couple of bytes content.

Link to it as shared lib instead.

This is an example of a mkv file with meta-data.

metadata.mkv.zip

This is the matroska documentation for tagging files.
And this is the cover art matroska documentation.

The goal is to enhance mediatomb metadata handler to use these mkv meta-data.

Thanks a lot for resurrecting this amazing project!!!

I've a media server running the old mediatomb and I was planning to use the master branch.

Is the old database compatible or should I just drop the old one and recreate it?

Sorry for asking a question here, but I didn't find a more suitable place.

I'm trying to compile gerbera on a Debian 8 system which ships CMake 3.0.2.
gerbera says it requires 3.2. Is that correct? Could that be lowered to 3.0 to make it work on Debian?

Since the introduction of folder images (which I absolutely love!), opening folders with many subfolders is too slow. It takes around 10 seconds to get the list of artists here (about 130). This is too long.

When I enable inotify on a particular directory, mediatomb crashes. Will add more details to this issue once I've narrowed down what's causing the crash.

first -TERM:
INFO: MediaTomb shutting down. Please wait...

I can wait all day, no exit

second -TERM:
INFO: Mediatomb still shutting down, signal again to kill.

same thing

third -TERM:
ERROR: Clean shutdown failed, killing MediaTomb!

Along those same lines, -HUP now does nothing

Is this related to the timer/nullptr changes?

if i would like to add media through symlinked directory, that directory is not visible through web gui.

The freedesktop folder specification says that configuration files should be put into ~/.config/ (or $XDG_CONFIG_HOME if it is set).

Gerbera creates the config.xml file in ~/.gerbera/config.xml, which clutters the home directory with another folder.

I suggest ~/.config/gerbera.xml as default location.

See http://upnp.org/specs/av/UPnP-av-ContentDirectory-v1-Service.pdf

2.7.5. Search

This action allows the caller to search the content directory for objects that match some search criteria.
The search criteria are specified as a query string operating on properties with comparison and logical
operators.

AlbumArt is only given for the tracks themselves and there is also no artist information for albums.

Is AlbumArtist handled?

In order to get DLNA to work where discovery fails [broadcast issues, igmpproxy fail with VLAN, etc], many DLNA clients allow specifying the discovery URL directly. For mediatomb it is:
http://host:port/description.xml
[minidlna uses rootDesc.xml]

This file doesn't get served if "ui enabled=no".

That seems to be a mistake; disabling the web GUI shouldn't disable root device publishing.

The logic here isn't obvious, as it's an implicit interaction with pupnp, or I'd volunteer to fix it straight away.

Cheers

I try to make openSUSE RPM

   55s]   CXX      libmediatomb_a-cds_resource_manager.o
[   55s]   CXX      libmediatomb_a-config_manager.o
[   55s] ../src/config_manager.cc:37:23: fatal error: uuid/uuid.h: No such file or directory
[   55s] compilation terminated.
[   55s] make[2]: *** [Makefile:1460: libmediatomb_a-config_manager.o] Error 1
[   55s] make[2]: *** Waiting for unfinished jobs....
[   56s] make[2]: Leaving directory '/home/abuild/rpmbuild/BUILD/mediatomb-mediatomb-0.12.2_pre20160522/build'
[   56s] make[1]: *** [Makefile:535: all-recursive] Error 1
[   56s] make[1]: Leaving directory '/home/abuild/rpmbuild/BUILD/mediatomb-mediatomb-0.12.2_pre20160522'
[   56s] make: *** [Makefile:466: all] Error 2
[   56s] error: Bad exit status from /var/tmp/rpm-tmp.DtReCV (%build)

How to fix this?

• I had to build and install pupnp with configure --enable-ipv6
• for some reason target specific cxx version settings in the cmakelists.txt do not work so I had to change the global flags set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall -std=gnu++14")
• src/tools.cc raised a problem so changed to log_debug("Looking for '%s'\n", ip.c_str());

I wanted to have a bit of an open discussion about what the expectations/directions are for the code base going forward, specifically for this fork. The existing code base contains a number of constructs that are now provided by most modern C++ compilers.

Is it safe to modernize the code base? If so, should it target C++11? C++14?

Many of the utilities in the zmm and hash directories have more robust cousins in C++11 or later. For example: string, regex, atomic, array, stack, queue, hash, shared pointer. Is it worth transitioning to the standard C++ equivalents over time?

One of the reasons I'm bringing this up is that I've noticed various issues. For example, some song titles are garbled when viewed on the web. Is that due to MediaTomb's String class? Or is it somewhere else in the mp3/flac tags, taglib, internal MT code, iconv, JS scripts, html output chain? Who knows. There's lots of custom string handling code around, and I suspect some of it may be responsible. I'd feel better about it if std::u16string or one of its friends was used to handle user strings.

I've also seen error logging from the hash table code about "maximal number of iterations exceeded". I haven't debugged it, but I'd feel better about it if a standard C++ container was used instead.

That by itself is not enough justification to make massive sweeping changes, but perhaps over time new code can start using std::string, std::stack, etc... instead of the zmm::Foo versions.

The config.xml file allows the user to specify directories to be auto-scanned (import - scripting - autoscan). When the application starts up with no pre-existing database, a new database is created, but the directories listed in the config file are not scanned to populate the database.