mt-daapd-devel Mailing List for Multi-Threaded DAAP Daemon

Hi Ron et al,

One issue I'd like to fix is that identical MP3s at different paths
are result in multiple database entries. This really bugs my gf since
music that shows up in both of our distinct mp3 collections shows up
on the Roku twice. I'm just starting to look at the code so I wanted
to mention my plans here in case there's a better way :)

My initial thought was to do a query in db.c db_add() for identical
length, name, (other metadata? suggestions?) and abort returning
success if there's a hit. Sound right?

Thanks -- Regards -- Andy

Hi there,

I have mt-daapd running on my edge box, in this case the ubuntu server.
   I am running a firewall on it.  But even with the firewall Off I am
unable to see the box getting advertised.  Therefore itunes on my other
machines are not able to see the server.  Could somebody suggest ways to
troubleshoot this?

Cheers,

Noah

Hi there,

I am having some strange issues with mt-daapd and was wondering if there 
is an appropriate mail list out there to contact?

Cheers,

Noah

I think there's a problem with the fedora init script in the contrib 
directory, it has a lower priority than avahi-daemon on which it appears 
to depend, so it fails to start up correctly on boot.  The avahi-daemon 
init script has priority 96 in fedora 9.  Everything is fine if you 
start it manually after boot (ie once avahi-daemon is running).

I'm no expert but I think its simply a matter of upping the priority of 
the init script as below.

Cheers,

Chris Simmons.

Index: contrib/init.d/mt-daapd-fedora.templ
===================================================================
--- contrib/init.d/mt-daapd-fedora.templ        (revision 1737)
+++ contrib/init.d/mt-daapd-fedora.templ        (working copy)
@@ -1,6 +1,6 @@
 #!/bin/sh
 #
-# chkconfig: 2345 85 15
+# chkconfig: 2345 99 15
 # description: mt-daapd is a multi-threaded DAAP server for iTunes
 # processname: mt-daapd
 # pidfile: /var/run/mt-daapd

While maintaining the OpenBSD port of mt-daapd, I was made aware of
some 64-bit issues in the code.  I have some patches (attached) that
should take care of the problem.

Arnaud

> I use the project mt-daapd and it's very useful!
> However, I wanted to add the following lines to my httpd.conf file:
>     RewriteEngine On
>

Okay, I'm slow, but this is merged now and will be around next release.

  -- Ron

> Attached is a patch which modifies src/plugins/Makefile.am to use
> pkglibdir instead of pkgdatadir for the plugins location.  I also
> modified contrib/Makefile.am and contrib/mt-daapd.conf.templ to use
> sysconf for the mt-daapd.playlist and libdir for plugins.

Nicely done, sir.  That's been on my radar for a long time, but I've  
been to lazy to do it.  I'll apply.

Thanks.  Let me know if there is anything else that would be helpful  
from an upstream perspective.  Also, if you do get specfiles and  
whatnot done, I wouldn't mind seeing those as well so I can add FC  
packages to nightly builds.

  -- Ron

On 7/1/07, Ian Burrell <ian...@gm...> wrote:
>
> I am working on a patch for src/plugins/Makefile.am that changes
> pkgdatadir to pkglibdir.
>

Attached is a patch which modifies src/plugins/Makefile.am to use
pkglibdir instead of pkgdatadir for the plugins location.  I also
modified contrib/Makefile.am and contrib/mt-daapd.conf.templ to use
sysconf for the mt-daapd.playlist and libdir for plugins.

 - Ian

I am trying to package the mt-daapd nightly snapshots for Fedora.  I
ran into a snag with the location of the plugins directory.

The plugins directory really should be /usr/lib/mt-daapd/plugins
instead of /usr/shared/mt-daapd/plugins.  The reason is that the
plugins contain shared libraries which are architecture specific.
/usr/share is supposed to contain arch-independent files.  For x86_64,
the plugins should go in /usr/lib64/mt-daapd/plugins to keep them
separate from the i386 ones.

I am working on a patch for src/plugins/Makefile.am that changes
pkgdatadir to pkglibdir.

 - Ian

On May 13, 2007, at 2:11 PM, St=E9phane Moreau wrote:

> <rewriteurl.patch>


That looks reasonable.  I'll probably merge that for next nightly.  =20
Thanks for the patch.

  -- Ron

Hi all.

    I don't know if this is fixed already, but I found that if the
gdbm database is corrupted (mt-daapd-0.2.4), gdbm calls exit(), which
terminates mt-daapd.  I wrote a small fix for it, which will catch the
exit() before exiting, and will remove the corrupted database
automatically.  That way, when mt-daapd is restarted, it will rebuild
the database cleanly.

Here is the patch:
--- a/src/main.c        Wed Dec 06 23:08:59 2006 -0500
+++ b/src/main.c        Fri Jan 05 14:02:47 2007 -0500
@@ -128,6 +128,28 @@ static int start_signal_handler(pthread_
 static int start_signal_handler(pthread_t *handler_tid);
 static void daap_handler(WS_CONNINFO *pwsc);
 static int daap_auth(char *username, char *password);
+static void daap_on_exit(int exit_val, void *args);
+
+/**
+ * Check the type of exit.  If we have a failure, remove
+ * the database, since it could be corrupted.  GDBM exits witout telling
+ * the caller
+ */
+static void
+daap_on_exit(int exit_val, void *args)
+{
+    char db_path[PATH_MAX + 1];
+
+    if (0 != exit_val)
+    {
+        /*
+         * Remove database file
+         */
+        snprintf(db_path,sizeof(db_path),"%s/%s",config.dbdir,"songs.gdb");
+        remove(db_path);
+    }
+
+}

 /**
  * Handles authentication for the daap server.  This isn't the
@@ -815,6 +837,11 @@ int main(int argc, char *argv[]) {
        config.pid=0;
     }

+    /*
+     * Initialize on_exit function
+     */
+    (void)on_exit(daap_on_exit, NULL);
+
     /* DWB: shouldn't this be done after dropping privs? */
     if(db_open(config.dbdir, reload))
        DPRINTF(E_FATAL,L_MAIN|L_DB,"Error in db_open: %s\n",strerror(errno));

FC6 uses the avahi-compat-howl packages rather than howl.  A few minor
changes to the RPM spec file seems to have fixed the build/install
dependency issues.  See the "diff -c" output below.

Do with it what you please.  Thx.

~Chris



*** mt-daapd.spec.original      2006-11-09 17:41:16.000000000 -0600
--- mt-daapd.spec       2006-11-09 17:54:01.000000000 -0600
***************
*** 7,14 ****
  URL: http://sourceforge.net/project/showfiles.php?group_id=98211
  Source0: %{name}-%{version}.tar.gz
  BuildRoot: %{_tmppath}/%{name}-%{version}-%{release}-buildroot
! Requires: libid3tag gdbm howl
! BuildRequires: libid3tag-devel gdbm-devel howl-devel

  %description
  A multi-threaded implementation of Apple's DAAP server, mt-daapd
--- 7,14 ----
  URL: http://sourceforge.net/project/showfiles.php?group_id=98211
  Source0: %{name}-%{version}.tar.gz
  BuildRoot: %{_tmppath}/%{name}-%{version}-%{release}-buildroot
! Requires: libid3tag gdbm avahi-compat-howl
! BuildRequires: libid3tag-devel gdbm-devel avahi-compat-howl

  %description
  A multi-threaded implementation of Apple's DAAP server, mt-daapd
***************
*** 19,25 ****
  %setup -q

  %build
! ./configure --prefix=$RPM_BUILD_ROOT/usr --enable-howl
--with-howl-includes=/usr/include/howl-0.9.8

  make RPM_OPT_FLAGS="$RPM_OPT_FLAGS"

--- 19,25 ----
  %setup -q

  %build
! ./configure --prefix=$RPM_BUILD_ROOT/usr --enable-howl
--with-howl-includes=/usr/include/avahi-compat-howl

  make RPM_OPT_FLAGS="$RPM_OPT_FLAGS"

> 1) Why is mt-daapd not responding to the INT and HUP signals sent  
> from svc?

I do not know.  It sounds like a pthreads issue though.  The signal  
strategy used is to block all signals on all threads except for a  
signal handling thread.  That thread runs with signals set to SIG_IGN  
and sits in a sigwait loop to be notified of signals.

I had understood that to be "best practice" in handling signals in  
multi-threaded applications, but apparently not all pthreads  
libraries deliver signals to the unblocked thread.  I believe that to  
be a pthread impelmentation bug.

> 2) Why was INT chosen as the shutdown signal rather than TERM?

Just because I was looking to cleanly shutdown on a ctrl-c when  
running in the foreground, and I hadn't bothered to add sigterm  
handling until recently.  It does handle sigterm in current nightlies.

   -- Ron

All,

I am attempting to manage mt-daapd 0.2.4 stable on Yellow Dog Linux 4.1 
from daemontools (http://cr.yp.to/daemontools.html).  I am able to 
successfully start mt-daapd with the following run script:

#!/bin/sh
exec 2>&1
mt-daapd -f

Sending an INT signal to the service via the following command:

# svc -i /service/mt-daapd

does not cause the desired effect, the service stopping.  Running the 
above run script with "-d 9" also shows that mt-daapd is either not 
seeing or is ignoring the INT signal.  Similar behavior is observed with 
the HUP signal.

Sending the INT and HUP signals to the mt-daapd via the following command:

# kill [SIGNAL] [PID]

results in the appropriate behaviors, service shutdown and reload.

My questions are below.

1) Why is mt-daapd not responding to the INT and HUP signals sent from svc?
2) Why was INT chosen as the shutdown signal rather than TERM?

Thanks,
Patrick

That's quite OK, Ron... I didn't read it that way!

BTW, hope your house move is going OK?

Phil

On Mon, 2005-09-26 at 20:29 -0500, Ron Pedde wrote:
> >> In any event, working + ugly is probably better than not-working +
> >> not-quite-as-ugly, so working it is.  Thanks for the heads-up on
> >> this, Phil -- I used much of your stuff.
> 
> Okay, I just read this with fresh eyes, and it looks like I'm saying:
> 
> "Your code is ugly, Phil".
> 
> That's not what I meant to say!  What I meant to say is that I used  
> much of your code, and the architecture is such that the queries  
> required are ugly.  The ugly is a factor of poor planning on my part,  
> and necessity on your part.
> 
> Wasn't implying anything tacky.
> 
> Sorry!
> 
>   -- Ron
> 
> 
> 
> 
> -------------------------------------------------------
> SF.Net email is sponsored by:
> Tame your development challenges with Apache's Geronimo App Server. Download
> it for free - -and be entered to win a 42" plasma tv or your very own
> Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php
> _______________________________________________
> mt-daapd-devel mailing list
> mt-...@li...
> https://lists.sourceforge.net/lists/listinfo/mt-daapd-devel
> 
-- 
Now! Also available at phe...@gm...

>> In any event, working + ugly is probably better than not-working +
>> not-quite-as-ugly, so working it is.  Thanks for the heads-up on
>> this, Phil -- I used much of your stuff.

Okay, I just read this with fresh eyes, and it looks like I'm saying:

"Your code is ugly, Phil".

That's not what I meant to say!  What I meant to say is that I used  
much of your code, and the architecture is such that the queries  
required are ugly.  The ugly is a factor of poor planning on my part,  
and necessity on your part.

Wasn't implying anything tacky.

Sorry!

  -- Ron

Thanks, Ron

I'll check it out and give it a try later.

All the best

Phil

On Sun, 2005-09-25 at 21:18 -0500, Ron Pedde wrote:
> > I think you'll get there before I do, and as I say, they way I've
> > currently got it is s-l-o-w!
> 
> This is in cvs.  I really don't like the query, but then, I really  
> don't like the lack of database abstraction.  Maybe once the database  
> stuff gets abstracted properly it won't be so ugly.
> 
> In any event, working + ugly is probably better than not-working +  
> not-quite-as-ugly, so working it is.  Thanks for the heads-up on  
> this, Phil -- I used much of your stuff.
> 
>    -- Ron
> 
> 
> 
> 
> -------------------------------------------------------
> SF.Net email is sponsored by:
> Tame your development challenges with Apache's Geronimo App Server. Download
> it for free - -and be entered to win a 42" plasma tv or your very own
> Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php
> _______________________________________________
> mt-daapd-devel mailing list
> mt-...@li...
> https://lists.sourceforge.net/lists/listinfo/mt-daapd-devel
> 
-- 
Now! Also available at phe...@gm...

> I think you'll get there before I do, and as I say, they way I've
> currently got it is s-l-o-w!

This is in cvs.  I really don't like the query, but then, I really  
don't like the lack of database abstraction.  Maybe once the database  
stuff gets abstracted properly it won't be so ugly.

In any event, working + ugly is probably better than not-working +  
not-quite-as-ugly, so working it is.  Thanks for the heads-up on  
this, Phil -- I used much of your stuff.

   -- Ron

On Wed, 2005-09-21 at 21:36 -0500, Ron Pedde wrote:
> Okay, I see what's up there.
> 
> Yeah, seems like it does need help in the playlistitems.  Maybe we  
> need an id, playlistid and songid.  id being an autoincrement field,  
> and playlist_id being what used to be called id.
> 
> then we could sort by id on that table.
> 
> > So, I decided that the only way I could reliably get back what I  
> > wanted
> > was to add that sort order field?
> 
> Yeah, I'm buying that, but I think I'd like to see it as an  
> autoincrement in playlistitems.
> 
> If i don't hear from you before that, I'll code it up tomorrow night.
> 
>   -- Ron
> 

I bow to your superior database-fu!

<grin>


I think you'll get there before I do, and as I say, they way I've
currently got it is s-l-o-w!

All the best

Phil

-- 
Now! Also available at phe...@gm...

> Hi, Ron, thanks for your email and sorry for the delay in replying.

s'okay, I'm slow myself.  :)

> sqlite> select id,title from songs where (id in (select songid from
> playlistitems where id=7));
> 938|Makin' Whoopee!.flac
> 939|Get Me to the Church on Time.flac
> 940|Street of Dreams.flac
> 941|You Make Me Feel So Young.flac
> 942|Where Or When.flac
> 943|The September of My Years.flac
> 944|Come Fly with Me.flac
> 945|One for My Baby and One for the Road.flac
> 946|All of Me.flac
> 947|One O'Clock Jump.flac
> 948|Don't Worry 'Bout Me.flac
> 949|Sinatra Closing Monologue.flac
> 950|It was a Verry Good Year.flac
> 951|My Kind of Town (Reprise).flac
> 952|Fly Me to the Moon.flac
> 953|Frank Sinatra Monologue.flac
> 954|I've Got a Crush on You.flac
> 955|Angel Eyes.flac
> 956|My Kind of Town.flac
> 957|I've Got You Under My Skin.flac
> 958|The Shadow of Your Smile.flac

Okay, I see what's up there.

Yeah, seems like it does need help in the playlistitems.  Maybe we  
need an id, playlistid and songid.  id being an autoincrement field,  
and playlist_id being what used to be called id.

then we could sort by id on that table.

> So, I decided that the only way I could reliably get back what I  
> wanted
> was to add that sort order field?

Yeah, I'm buying that, but I think I'd like to see it as an  
autoincrement in playlistitems.

If i don't hear from you before that, I'll code it up tomorrow night.

  -- Ron

Hi, Ron, thanks for your email and sorry for the delay in replying.

I went back to mt-daapd-cvs-20050815 and made some investigations -
here's what I found:



A/	Order files appear on disk:

Note that if sorted by i-node they are in the order I want (but that's
only coz they were ripped in that order)

[root@nas Sinatra At The Sands with Count Basie]# ls -li *.flac
2187 -rw-rw-r--  1 pep media 17432240 Sep 15 17:57 All of Me.flac
2248 -rw-rw-r--  1 pep media 16396694 Sep 15 18:03 Angel Eyes.flac
2120 -rw-rw-r--  1 pep media 23577527 Sep 15 17:51 Come Fly with Me.flac
2212 -rw-rw-r--  1 pep media 21715206 Sep 15 18:00 Don't Worry 'Bout
Me.flac
2179 -rw-rw-r--  1 pep media 16928503 Sep 15 17:54 Fly Me to the
Moon.flac
2185 -rw-rw-r--  1 pep media 64613693 Sep 15 17:56 Frank Sinatra
Monologue.flac
2205 -rw-rw-r--  1 pep media 15777183 Sep 15 17:58 Get Me to the Church
on Time.flac
2206 -rw-rw-r--  1 pep media 20649677 Sep 15 17:59 It was a Verry Good
Year.flac
2126 -rw-rw-r--  1 pep media 13932121 Sep 15 17:51 I've Got a Crush on
You.flac
2135 -rw-rw-r--  1 pep media 22409519 Sep 15 17:52 I've Got You Under My
Skin.flac
2213 -rw-rw-r--  1 pep media 25232139 Sep 15 18:01 Makin' Whoopee!.flac
2249 -rw-rw-r--  1 pep media 19779074 Sep 15 18:04 My Kind of Town.flac
2254 -rw-rw-r--  1 pep media  6297065 Sep 15 18:04 My Kind of Town
(Reprise).flac
2178 -rw-rw-r--  1 pep media 21921667 Sep 15 17:53 One for My Baby and
One for the Road.flac
2180 -rw-rw-r--  1 pep media  6685813 Sep 15 17:54 One O'Clock Jump.flac
2250 -rw-rw-r--  1 pep media 13719884 Sep 15 18:04 Sinatra Closing
Monologue.flac
2157 -rw-rw-r--  1 pep media 14226127 Sep 15 17:53 Street of Dreams.flac
2204 -rw-rw-r--  1 pep media 15298366 Sep 15 17:57 The September of My
Years.flac
2136 -rw-rw-r--  1 pep media 13462837 Sep 15 17:52 The Shadow of Your
Smile.flac
2214 -rw-rw-r--  1 pep media 17463873 Sep 15 18:02 Where Or When.flac
2186 -rw-rw-r--  1 pep media 23090035 Sep 15 17:56 You Make Me Feel So
Young.flac

B/	Order files appear in playlist on disk (which is the order I want in
iTunes etc):

[root@nas Sinatra At The Sands with Count Basie]# cat "Sinatra At The
Sands with Count Basie.m3u"
Come Fly with Me.flac
I've Got a Crush on You.flac
I've Got You Under My Skin.flac
The Shadow of Your Smile.flac
Street of Dreams.flac
One for My Baby and One for the Road.flac
Fly Me to the Moon.flac
One O'Clock Jump.flac
Frank Sinatra Monologue.flac
You Make Me Feel So Young.flac
All of Me.flac
The September of My Years.flac
Get Me to the Church on Time.flac
It was a Verry Good Year.flac
Don't Worry 'Bout Me.flac
Makin' Whoopee!.flac
Where Or When.flac
Angel Eyes.flac
My Kind of Town.flac
Sinatra Closing Monologue.flac
My Kind of Town (Reprise).flac

C/	Here's what happens in the database:

1/	Find the playlist

sqlite> select * from playlists where title like "%Sinatra%";
7|Sinatra At The Sands with Count Basie|2|21||
1126966724|/exports/media/music/Frank Sinatra/Sinatra At The Sands with
Count Basie/Sinatra At The Sands with Count Basie.m3u|0

2/	Look at the linking table 

sqlite> select * from playlistitems where id=7;
7|944
7|954
7|957
7|958
7|940
7|945
7|952
7|947
7|953
7|941
7|946
7|943
7|939
7|950
7|948
7|938
7|942
7|955
7|956
7|949
7|951

Which does seem to return things in the order I want, as I guess the
loop that adds items to this table goes through in order of the lines in
the m3u file.


3/	Here's the query that returns the playlist

sqlite> select id,title from songs where (id in (select songid from
playlistitems where id=7));
938|Makin' Whoopee!.flac
939|Get Me to the Church on Time.flac
940|Street of Dreams.flac
941|You Make Me Feel So Young.flac
942|Where Or When.flac
943|The September of My Years.flac
944|Come Fly with Me.flac
945|One for My Baby and One for the Road.flac
946|All of Me.flac
947|One O'Clock Jump.flac
948|Don't Worry 'Bout Me.flac
949|Sinatra Closing Monologue.flac
950|It was a Verry Good Year.flac
951|My Kind of Town (Reprise).flac
952|Fly Me to the Moon.flac
953|Frank Sinatra Monologue.flac
954|I've Got a Crush on You.flac
955|Angel Eyes.flac
956|My Kind of Town.flac
957|I've Got You Under My Skin.flac
958|The Shadow of Your Smile.flac

And note that I don't get them back in the order they are mentioned in
the playlist

And an "order by id" clause won't get me what I need either as you can
see

So, I decided that the only way I could reliably get back what I wanted
was to add that sort order field?

What do you think?


Regards

Phil



On Sat, 2005-09-17 at 16:58 -0500, Ron Pedde wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> > I do hope I haven't missed anything really obvious and so wasted a few
> > hours here, but I couldn't see anything in the code to make it play a
> > *dumb* m3u playlist in the same order as the songs were listed in the
> > original playlist file (IYSWIM!)
> 
> Only in that one assumes that ids are monitonically increasing and  
> that a select with no order by implicitly orders by id.
> 
> That appeared to be the case, but if you noticed otherwise, I'd be  
> interested to find out why.
> 
> Might just be easier to do an "order by id" in the playlist query.
> 
> > The other thing about it is, having very little sql fu myself, the
> > replacement SQL query in dbs-sqlite.c around line 783 sucks  
> > mightily as
> > it's horribly slow. I'm sure there must be a neater way to make it do
> > the right thing there?
> 
> Yes!  The thing with the whole sql deal is that it's terribly  
> broken.  The sql part of it needs to be abstracted, for one, and the  
> all the hard-coded sql queries have to be replaced with functions to  
> build a db-specific query.  I'm looking at something similar to the  
> adodb interface.
> 
> Why I didn't look at sql abstraction libraries before I tried to  
> refactor the db code, I don't know.
> 
> That's probably the ugliest, stupidest part of the whole code,  
> though, and really needs to be fixed.
> 
>    -- Ron
> 
> 
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.2.4 (Darwin)
> 
> iD8DBQFDLJF2Dl/VbZJe82cRArq/AKC6La5Zuec5Jq/cn+Hc6BZzlVQOJACg75LH
> e4ePeKof1ChBBT9BsINDYw4=
> =1+sV
> -----END PGP SIGNATURE-----
> 
> 
> -------------------------------------------------------
> SF.Net email is sponsored by:
> Tame your development challenges with Apache's Geronimo App Server. Download
> it for free - -and be entered to win a 42" plasma tv or your very own
> Sony(tm)PSP.  Click here to play: http://sourceforge.net/geronimo.php
> _______________________________________________
> mt-daapd-devel mailing list
> mt-...@li...
> https://lists.sourceforge.net/lists/listinfo/mt-daapd-devel
> 
-- 
Now! Also available at phe...@gm...

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

> I do hope I haven't missed anything really obvious and so wasted a few
> hours here, but I couldn't see anything in the code to make it play a
> *dumb* m3u playlist in the same order as the songs were listed in the
> original playlist file (IYSWIM!)

Only in that one assumes that ids are monitonically increasing and  
that a select with no order by implicitly orders by id.

That appeared to be the case, but if you noticed otherwise, I'd be  
interested to find out why.

Might just be easier to do an "order by id" in the playlist query.

> The other thing about it is, having very little sql fu myself, the
> replacement SQL query in dbs-sqlite.c around line 783 sucks  
> mightily as
> it's horribly slow. I'm sure there must be a neater way to make it do
> the right thing there?

Yes!  The thing with the whole sql deal is that it's terribly  
broken.  The sql part of it needs to be abstracted, for one, and the  
all the hard-coded sql queries have to be replaced with functions to  
build a db-specific query.  I'm looking at something similar to the  
adodb interface.

Why I didn't look at sql abstraction libraries before I tried to  
refactor the db code, I don't know.

That's probably the ugliest, stupidest part of the whole code,  
though, and really needs to be fixed.

   -- Ron


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (Darwin)

iD8DBQFDLJF2Dl/VbZJe82cRArq/AKC6La5Zuec5Jq/cn+Hc6BZzlVQOJACg75LH
e4ePeKof1ChBBT9BsINDYw4=
=1+sV
-----END PGP SIGNATURE-----

Hi, folks

I do hope I haven't missed anything really obvious and so wasted a few
hours here, but I couldn't see anything in the code to make it play a
*dumb* m3u playlist in the same order as the songs were listed in the
original playlist file (IYSWIM!)

In the hope that it'll be useful, here's a patch to do just that. It
only works for m3u playlists and is just bodged round (but in a harmless
way) for other types of playlist.

The other thing about it is, having very little sql fu myself, the
replacement SQL query in dbs-sqlite.c around line 783 sucks mightily as
it's horribly slow. I'm sure there must be a neater way to make it do
the right thing there?

Regards to all

Phil




-- 
Now! Also available at phe...@gm...

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1


There is a new snapshot on the nightlies site 
(http://nightlies.mt-daapd.org) that's actually pretty reasonably 
complete:  query and browse is there, so it works for the soundbridge.  
xml output is working again.  Index support is still broken, but should 
be there shortly.

Some things to be aware of:

1.  the database is now opened after privs are dropped, so the database 
*and the folder containing it* must be writable by the "runas" user.
2.  this requires sqlite 2.8, not sqlite 3.
3.  there are probably lots of broken things, as this is a lot of new 
code.  That said, core stuff seems reasonably stable.
4.  transcoding probably doesn't work
5.  if you used the previous snapshot, delete your database.  this 
version will recreate the database and populate it correctly.  any 
database format changes from here on out will be done in-place 
automatically.

   -- Ron

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (Darwin)

iD8DBQFCPQgkDl/VbZJe82cRAk7YAJwO52PtMhkjo7F/3SHi/tmrSsXO0QCdH5Cb
HVmN9MeZvt4QbK1sEW9V6IM=
=+Xvo
-----END PGP SIGNATURE-----

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

I guess as an update... in cvs there is now query and browse support.  
Index support is still missing, and since wildcard queries are broken, 
a search from a Roku will error.

I have a sqlite package build for NSLU2, and I'm going to work on 
performance with a goal of making the nslu2 as fast against sqlite as 
it is on gdbm.  Fast enough to work, anyway.  :)

Once I get it performing decently, I'll post it as a snapshot.

  -- Ron

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (Darwin)

iD8DBQFCOiFMDl/VbZJe82cRAp+rAJ9Wjicfy0R4ISvsEoH7xRdpKrBbWwCgzCLp
mrAADviNuOBD6cVBTePp7BU=
=/l7Z
-----END PGP SIGNATURE-----