PDA

View Full Version : Slimserver suddenly not reading Ogg Vorbis tags



Steve Bernard, Jr
2005-03-17, 19:16
Hey folks,

I'm running svn revision 2552, and when I just wiped my db file to
rescan it, I found that it wasn't picking up the tags for any of my
Ogg Vorbis files, which comprise nearly all of my collection. I ran
the server with --d_info and got a lot of:

2005-03-17 21:04:46.0012 Using format "/ARTIST/ALBUM/TRACKNUM. TITLE"
= //([^\/]+)/([^\/]+)/(\d+)\. ([^\/]+)/...

and messages like it instead of it reading my Vorbis tags. This only
started happening today after updating the server. I checked on my
Ogg::Vorbis::Header installation to make sure it didn't disappear (Dan
helped me out by suggesting I build my own last week) and it seems to
be fine.

I happened to have a copy of the 3-10-05 nightly in my /home so I ran
that, and the tags are picked up correctly, so I think this is a bug
introduced today. Is anyone else seeing it?

-Steve

Dan Sully
2005-03-17, 20:01
* Steve Bernard, Jr shaped the electrons to say...

>I'm running svn revision 2552, and when I just wiped my db file to
>rescan it, I found that it wasn't picking up the tags for any of my
>Ogg Vorbis files, which comprise nearly all of my collection. I ran
>the server with --d_info and got a lot of:
>
>and messages like it instead of it reading my Vorbis tags. This only
>started happening today after updating the server. I checked on my
>Ogg::Vorbis::Header installation to make sure it didn't disappear (Dan
>helped me out by suggesting I build my own last week) and it seems to
>be fine.

Steve - I checked in a patch today to lazily load the format parsers.

Could you get me more of the log around this scan with --d_info?

Thanks.

-D
--
<iNoah> you know, most free operating systems come preinstalled with their own high horse.

Dan Sully
2005-03-17, 20:04
* Steve Bernard, Jr shaped the electrons to say...

>I happened to have a copy of the 3-10-05 nightly in my /home so I ran
>that, and the tags are picked up correctly, so I think this is a bug
>introduced today. Is anyone else seeing it?

Steve - this seems to happen when using the compiled Ogg header parser, not
the pure perl one. I'm looking into it.

-D
--
On second thought, let's not go to Camelot. It is a silly place.

Dan Sully
2005-03-17, 23:44
* Steve Bernard, Jr shaped the electrons to say...

>and messages like it instead of it reading my Vorbis tags. This only
>started happening today after updating the server. I checked on my
>Ogg::Vorbis::Header installation to make sure it didn't disappear (Dan
>helped me out by suggesting I build my own last week) and it seems to be fine.

Steve - so I took a look at the problem, and it looked nasty, some Inline/C
can't load issue with the Ogg::Vorbis::Header module.

So - I decided to see what was slow about the Ogg::Vorbis::Header::PurePerl
module, which is originally why I told you to try the compiled version. Upon
inspection of the code - I came to the conclusion that the author was doing
some.. unoptimal things. So I fixed them. :)

I've checked in these changes, and made the PurePerl module the default - as
now it's just as fast as the compiled version. Benchmarks for reading a file 1024 times:

Ogg::Vorbis::Header::PurePerl before:

Total Elapsed Time = 74.92390 Seconds
User+System Time = 74.77390 Seconds
Exclusive Times
%Time ExclSec CumulS #Calls sec/call Csec/c Name
42.1 31.52 31.528 416768 0.0001 0.0001 Ogg::Vorbis::Header::PurePerl::_findPage
38.7 29.00 76.614 1024 0.0283 0.0748 Ogg::Vorbis::Header::PurePerl::_calculateTrackLeng th
21.9 16.41 16.417 433152 0.0000 0.0000 Ogg::Vorbis::Header::PurePerl::_decodeInt

---------------------------------------------------------

Ogg::Vorbis::Header::PurePerl after:

Total Elapsed Time = 4.055065 Seconds
User+System Time = 4.105065 Seconds
Exclusive Times
%Time ExclSec CumulS #Calls sec/call Csec/c Name
75.6 3.104 3.111 1024 0.0030 0.0030 Ogg::Vorbis::Header::PurePerl::_calculateTrackLeng th
9.26 0.380 0.533 1024 0.0004 0.0005 Ogg::Vorbis::Header::PurePerl::_loadComments
6.70 0.275 0.275 18432 0.0000 0.0000 Ogg::Vorbis::Header::PurePerl::_decodeInt
3.46 0.142 0.256 1024 0.0001 0.0003 Ogg::Vorbis::Header::PurePerl::_loadInfo
1.39 0.057 0.057 1024 0.0001 0.0001 Ogg::Vorbis::Header::PurePerl::_checkHeader

---------------------------------------------------------

Ogg::Vorbis::Header compiled Inline/C module:

Total Elapsed Time = 3.938179 Seconds
User+System Time = 3.928179 Seconds
Exclusive Times
%Time ExclSec CumulS #Calls sec/call Csec/c Name
36.5 1.437 1.437 1024 0.0014 0.0014 Ogg::Vorbis::Header::_load_comments
34.8 1.367 1.367 1024 0.0013 0.0013 Ogg::Vorbis::Header::_load_info
24.6 0.967 0.967 1024 0.0009 0.0009 Ogg::Vorbis::Header::_new

-D
--
"A good messenger expects to get shot." --Larry Wall

Steve Bernard, Jr
2005-03-18, 07:11
On Thu, 17 Mar 2005 22:44:29 -0800, Dan Sully <dan (AT) slimdevices (DOT) com> wrote:
> * Steve Bernard, Jr shaped the electrons to say...
>
> >and messages like it instead of it reading my Vorbis tags. This only
> >started happening today after updating the server. I checked on my
> >Ogg::Vorbis::Header installation to make sure it didn't disappear (Dan
> >helped me out by suggesting I build my own last week) and it seems to be fine.
>
> Steve - so I took a look at the problem, and it looked nasty, some Inline/C
> can't load issue with the Ogg::Vorbis::Header module.
>
> So - I decided to see what was slow about the Ogg::Vorbis::Header::PurePerl
> module, which is originally why I told you to try the compiled version. Upon
> inspection of the code - I came to the conclusion that the author was doing
> some.. unoptimal things. So I fixed them. :)

Hey guys on the list, remember when Sean posted here all excited that
they'd hired Dan Sully and I was all like "Who?" No? You weren't in
my house at the time that I read that email? Well, anywho, I get it
now.

Okay, I tried this out when I got up this morning, and at first time
it rebuilt I wound up with like 300 tracks listed with No Album, No
Artist, and No Genre. vorbiscomment -l was showing correct tags, so I
rebuilt again with --d_info to get the output, but this time it didn't
do it, so I'll give it a miss.

I notice that my artist count has dropped by like four. But if I read
my checkins correctly these last couple of days the case sensitivity
of tags has been lessened, yes? I'm sure I had a couple instances
where one album was by 'Windy and Carl' and the other was 'Windy And
Carl' or whatever. Or am I wrong and I better go comb through to see
what's missing?

Thanks much,

-Steve

Dan Sully
2005-03-18, 09:38
* Steve Bernard, Jr shaped the electrons to say...

>I notice that my artist count has dropped by like four. But if I read
>my checkins correctly these last couple of days the case sensitivity
>of tags has been lessened, yes? I'm sure I had a couple instances
>where one album was by 'Windy and Carl' and the other was 'Windy And
>Carl' or whatever. Or am I wrong and I better go comb through to see
>what's missing?

No - you are correct - the case sensetivity is gone. This is back to the 5.4 behavior.

-D
--
( ( ( [ ] ) ) )
In Stereo Where
Available