Audacious

Well, I went and did it, getting it to work perfectly w/MP3s with essentially 1 line of code in gio.cc (but added a cpl. more to pretty it up! The reason I used "stdin://-.wav" instead of /dev/stdin (though it gets converted to that in gio.cc) is that Audacious needs (I think) to know the file type being streamed in w/o trying to "rewind" stdin, and this seems easiest, as I know what I'm piping in. Anyway, here goes:

diff gio.cc.org gio.cc
37c37
< static const char * const gio_schemes[] = {"ftp", "sftp", "smb"};
---

static const char * const gio_schemes[] = {"ftp", "sftp", "smb", "stdin"};

110a111

#ifdef _WINDOWS

111a113,116

#endif
#ifndef _WINDOWS
m_file = (strstr(filename, "://-.")) ? g_file_new_for_path ("/dev/stdin") : g_file_new_for_uri (filename); /* JWT:ADDED TO HANDLE INPUT FROM stdin */
#endif

257a263,264

if (strstr(m_filename, "://-.")) return 0; /* JWT: CAN'T SEEK ON STDIN!! */

HOWEVA!!-- The other two (ogg and wav) that I wanted required changes to the input plugins themselves to only "open" the file once and never close it, etc. I had hoped to NOT modify the plugins this time around, but could not figure out how (NOTE, I'm completely wet behind the ears w/C++ pgming!). I've also added some useful (at least to me) command-line options:

audacious36hacked --help

...
-c, --clear Clear Playlist // open up w/an empty playlist
-D, --delete DELETE all playlists!
-n, --new New Instance // force opening a new instance
-P, --pause-mute Pause mutes instead of pausing
-z, --no-equalizer Force Equalizer Off // useful when streaming talkshow, etc. streams

Also added a simple "-" stdin file to pipe in a list of files to play, useful for linuxy-stuff like:

$>find ~/Music -name "*dancin*" -print | audacious36hacked -

I'm currently testing everything out. Any suggestions on how/where to send my patches for consideration?

Thanks,

Jim

You can attach patches to this ticket. But it may be that some of the changes for this one use case have undesirable side effects for general use.

Thanks, I'll look at these as I have time. Some initial comments though:

1. I would rather not confuse "pause" and "mute". It might be useful to add a "mute" command, but aren't most systems nowadays set up with a mute button on the keyboard already?

2. Speaking without having looked at the patches yet: running a new instance reliably involves more than might be apparent at first glance. The main issue is to keep the two instances from fighting over the config files, one overwriting the other's changes.

I'm working on some changes to the VFS layer that might bring "audacious /dev/stdin" closer to a reality. I'd like to avoid changing individual input plugins if at all possible. Just because from past experience: any additional burden/restriction placed on input plugins means that some obscure plugin won't be updated correctly, and feature X, Y, or Z will be broken as soon as that plugin is used.

1) The reason I replace "Pause" with "Mute" in this option vs. a separate "Mute" option is that I use the Winamp skin interface generally and don't know how to add another button to the skins for "mute", but wanted to use the existing "Pause" button. The reason for wanting to "Mute" is that simply muting via the keyboard mutes the sound card itself, whereas my primary goal for this option is this option is to be able to mute just THIS instance of Audacious whilst still having the sound card unmuted so I can temporarily listen to something else (ie. while the "muted" stream is playing commercials, etc.)

2) Good point. This may require further work/testing. (My use of the new instance option was to be able to stream 2 different stations, switching now and then between them, "muting" one while listening to the other, while both continue stream away.

3) Awesome! I totally agree ! It would be ideal if we don't have to modify ANY of the input plugins, but could handle the stdin file-handle within the core. For example, I spent several hours trying to get ".wav" to work thru it's regular plugin ffaudio to no avail due to my inability to get the VFS layer to play nice w/stdin and modified sndfile to grab it. I basically see the main issue to making stdin "input plugin-independent" is the need to avoid closing, re-opening, and attempting to seek on stdin and the likely need to determine the input format solely by extension (provided in my model for handling stdin) vs "by content" which, for some formats, seems to cause opening and reading parts of the stream then rewinding it before reopening it to play. I assume that the "neon" plugin somehow avoids this, since it has to deal w/un-seekable streams?

On a slightly different subject, could you consider adding a [Close] button to the "Playlist Manager" screen? I use the "Afterstep" window manager and it considers it a "transient" window, thus no "window decorations", therefore no easy way to "close" that window! If you want, I can file a separate "bug" for this issue.

Thanks,

Jim

Try Git master and see how it works for you with stdin (use "audacious -"). I tested MP3 and Ogg Vorbis files; both seemed to play okay (though the MP3 tag gets ignored). Since a few releases ago, we've already had some buffering magic that allows the input plugins to think we're rewinding the file without actually doing so; I just massaged the code a bit so that we can continue to use that same buffered data when we actually start playing the file.

Thanks John for the quick work! Tell me what I'm doing wrong - I started with a clean untar from the current production v3.6 (which I had originally used to build my hacks), replaced the 9 respective source files with the raw versons in Git.master, but could not get them to compile, tried a few things, and gave up. I didn't see any later source trees, so I assumed, since diff-ing the effected files and that matched with the diffs shown. It said "probe-buffer.h" was missing, I found it and tried again, no joy. I haven't used Git before, so I may be missing something else. The 1st error I got (after finding probe-buffer.h on Git.master) was:

playlist.cc: In function 'void scan_finish(ScanRequest*)':
playlist.cc:559:81: error: 'struct ScanRequest' has no member named 'file'
if (playlist playing_playlist && entry playlist->position && request->file)
^
playlist.cc:560:48: error: 'struct ScanRequest' has no member named 'file'
playback_set_file (std::move (request->file));
^
playlist.cc:560:53: error: 'playback_set_file' was not declared in this scope
playback_set_file (std::move (request->file));
^
Failed to compile playlist.cc (lib)!

I then guessed that you had meant "request->filename", after looking at the struct.  Then got:

playlist.cc: In function 'void scan_finish(ScanRequest*)':
playlist.cc:560:57: error: 'playback_set_file' was not declared in this scope
playback_set_file (std::move (request->filename));
^

At this point I gave up and posted this.

There were more than 9 files updated. You had better use "git clone" rather than trying to update them one by one. See [1] or [2].

1. http://audacious-media-player.org/developers
2. http://redmine.audacious-media-player.org/boards/1/topics/788

Thanks, followed your instructions & got it working as you described. It seems to work fine for ogg and mp3 as you said. wav does not work yet, however. Without an extension on the stdin filehandle, you have it searching thru all the input plugins trying to match by content. It seems that sndfile and vorbis are reading off some of the stream before determining that it's not their format, which then seems to cause ffmpeg to fail when trying wav. After having spent a lot of time with this several years ago and now again recently, it seems to me that we're most likely going to end up having to specify content by extension somehow (ie. "-.ext"), but if you can get it work for at least ogg, mp3, and wav w/o modifying the input plugins, I'm good, but ideally, as far as possible, in a input-format-independent way.

Note: I had also successfully used stdin for double-duty (without extension) for inputting song filenames into the active playlist, ie:

$>find . -name "*dancin*" -print | audacious -

This was done in my patch to main.cc. Either way, I want to thank you for your time and effort so far in working w/me on this feature! I want to continue to assist in anyway I can on this.

Jim

Jim Turner wrote:

Note: I had also successfully used stdin for double-duty (without extension) for inputting song filenames into the active playlist, ie:

$>find . -name "*dancin*" -print | audacious -

Wouldn't audacious `find . -name "*dancin*"` be simpler?

I'll look into the issue with .wav.

Well hot damn, that works!! Thanks! Guess I need to bone up on my shell features too! :D

Jim

.wav from stdin should be working now. It will play in the sndfile plugin, not ffaudio.

Jim Turner wrote:

On a slightly different subject, could you consider adding a [Close] button to the "Playlist Manager" screen? I use the "Afterstep" window manager and it considers it a "transient" window, thus no "window decorations", therefore no easy way to "close" that window! If you want, I can file a separate "bug" for this issue.

There is an easy way to close the window: the Escape key. ;)

1) Separate config files would be fine w/me for these. What did you have in mind, something like an option to specify? ie:

$>audacious -c ~/.config/audacious/myspecialconfigfile

2) Did you change sndfile/plugin.cc to get .wav to work from stdin or were you able to make it work in the core? Reason I'm asking is that I found and tried an ".m4a" file, which fails via stdin, but works otherwise. I don't really care myself if that works or not, but thought you might if you are aiming to do it properly in core. Both methods use ffaudio for input, but stdin appears to fail trying to do illegal seeks.

3) I really DO want that "Pause-does-Mute" option though, to be able to mute commercials, etc. in one stream whlst keeping my sound-card unmuted for other uses! - it's a pretty simple hack to zero out the output to the speakers if the user specifies this option - in summary:

$>diff output.cc output.hacked.cc
32a33
> #include <stdio.h>
65a67,68
> short iamPaused = 0;   /* JWT:  ADDED FOR ALLOWING PAUSE TO CONTINUE READING STREAMS, ONLY PAUSING OUTPUT! */
> 
116,117c119,133
<     cop->pause (s_paused);
<     vis_runner_start_stop (true, s_paused);
---
>     /* JWT:  IF PauseMute SET, PAUSE SHOULD JUST SUSPEND OUTPUT
>              THIS IS SO WE CAN "MUTE" COMMERCIALS FROM RADIO WHILST PLAYING 
>              SOMETHING ELSE!  THIS WAY, WE CAN RESUME AT THE CURRENT POINT vs
>              WHERE WE PAUSED!  THIS HACK WILL KEEP THE VISUALIZATION RUNNING
>              PERMITTING VISUAL CONFIRMATION THAT THE STREAM IS STILL STREAMING,
>              ONLY NO SOUND GOES TO THE SOUNDCARD!
>     */
>     if (aud_get_pausemute_mode ())
>     {
>         iamPaused = s_paused;
>         vis_runner_start_stop (true, false);  /* NOT SURE WHY "s_paused" CAN'T BE USED HERE????! */
>       }
>     else
>     {
>         cop->pause (s_paused);
>         vis_runner_start_stop (true, s_paused);
>     }
230c246
<     const void * out_data = data.begin ();
---
>     void * out_data = data.begin ();
242a259,262
>         /* JWT: ZERO-OUT (SILENCE) STREAM GOING TO OUTPUT WHILST PAUSING IF PAUSE-IS-MUTE OPTION SET: */
>         if (iamPaused)
>             memset (out_data, '\0', out_bytes_held);

Regards,

Jim