OpenNab 0.02

NerdUno a écrit:

I'll look around in the code, but you can probably point to the problem in a few seconds. I'm not that smart. :-)

I cannot test it right now but I think replacing:
$this->fileName = 'burrows/'.$serialNumber;
at line 45 in file /vl/includes/burrow.php
by
$this->fileName = 'burrows/'.strtolower($serialNumber);
will do the job.

oaz a écrit:

$this->fileName = 'burrows/'.strtolower($serialNumber);

Yep! That did the job! Woohoo!

Unfortunately the mp3 does indeed get cut off at about 1 minute. Anyone have any idea why? It doesn't seem to matter how big the file is, I tried max quality mp3 (320) and low quality (64), but both seem to stop after about 1 minute. So I'm guessing this is somehow hardcoded into the system?

The burrow.php fix works for internal messages, but we still do not get messages played from Violet where our MAC address was entered in ALL UPPER CASE.

NerdUno a écrit:

The burrow.php fix works for internal messages, but we still do not get messages played from Violet where our MAC address was entered in ALL UPPER CASE.

Odd. Works for me.

Interesting. Perhaps you registered your bunny with lower case but were trying the demo with UPPER CASE. I've noticed that changing burrow.php with either strtolower() or strtoupper() fixes the problem for playing local MP3s. But it does not resolve the Violet transmissions. Looking in the log, even after setting strtoupper(MACaddr), the log still shows MAC entries in lower case so it's obviously getting set independently in other places. That would explain why the Violet transmissions still are not working if set to an all UPPER CASE MAC address.

NerdUno a écrit:

Looking in the log, even after setting strtoupper(MACaddr), the log still shows MAC entries in lower case so it's obviously getting set independently in other places. That would explain why the Violet transmissions still are not working if set to an all UPPER CASE MAC address.

The entries in the log come directly from the rabbit : it just means that the rabbit always send her own serial number in lower case. It is not related to OpenNab.
By the way, I tried to use Violet's API with sn in upper case and it worked.

That does not solve your problem but I think we should look in other directions to understand why you does not receive any sound from Violet when using OpenNab.

tout es okay pour moi, maintenant faut trouver comment
faire passer le record de la reconnaissance vocal, et replonger dans nos trames pour le stream

Did some more checking. Still no luck. It's not a firewall problem. I put the bunny in the DMZ and had the same results. Does anyone recall registering their rabbit with ALL UPPER CASE MAC address at Violet? That's all I can think of that could be different.

NerdUno a écrit:

Did some more checking. Still no luck. It's not a firewall problem. I put the bunny in the DMZ and had the same results. Does anyone recall registering their rabbit with ALL UPPER CASE MAC address at Violet? That's all I can think of that could be different.

I do not remember how I entered the mac address on the violet site
but I can tell, for sure, that if you have Violet sound when directly connecting to Violet and no Violet sound when connecting through OpenNab then it cannot be a mac address issue.
The reason is that OpenNab do not parse the mac address for forwarding sound orders and requests between the rabbit and the Violet server : it just uses the plain URI sent by the rabbit.

Now we've got the following options :
(1) you have no Violet sound whether you go through OpenNab or not : this is a Violet issue. Contact them.
(2) you have no Violet sound only when connectiong through OpenNab. This is an OpenNab issue. We must find out if the issue is in
(a) upstream ping: is the ping from Nabaztag forwarded to Violet ?
(b) downstream order: is the Violet reply containing the play order forwarded to the rabbit ?
(c) upstream mp3 request : is the rabbit mp3 request forwarded to Violet ?
(d) downstream mp3 : is the mp3 data from Violet forwarded to the rabbit ?

When you send a message to your rabbit using Violet api or web interface, you should get logs with quite long type 0a blocks [0a,xxx] yyyy.... with xxx > 300
If you do not have these logs then issue is in (a) or (b)
If you have these logs then issue is in (c) or (d)

These logs should be followed by calls to /broad/<whatever>
If you do not have these logs then issue is in (c)
If you have these logs then issue is in (d)

Just an idea : do you have a correct .htaccess in /broad. I mean a htaccess forwarding 404 to /vl/media.php
If you do not have this .htaccess, requests for OpenNab mp3 are ok because they are found on the OpenNab web server but requests for Violet mp3 will fail (you will see that in the Apache log) because /vl/media.php which performs (c) and (d) is not called.

clayde a écrit:

tout es okay pour moi, maintenant faut trouver comment
faire passer le record de la reconnaissance vocal, et replonger dans nos trames pour le stream

héhé
J'ai comme l'impression que c'est toujours des blocks 0a avec le même encodage
Il y a juste un truc qui me chifonne : 2 appels consécutifs à l'api urlList avec les mêmes paramètres diffèrent très légèrement (entre le 9ème et le 13ème octet).

Par contre là pour trouver la table d'encodage ça devrait être beaucoup plus simple puisqu'on peut passer ce que l'on veut comme url...

oaz a écrit:

Just an idea : do you have a correct .htaccess in /broad. I mean a htaccess forwarding 404 to /vl/media.php
If you do not have this .htaccess, requests for OpenNab mp3 are ok because they are found on the OpenNab web server but requests for Violet mp3 will fail (you will see that in the Apache log) because /vl/media.php which performs (c) and (d) is not called.

That was it. THANK YOU. I must have missed that in the French somewhere. Anyway, this was a terrific explanation of how the stuff works which really helps.

By the way, I will have Weather Forecasting systems up and running with OpenNab and Flite and Cepstral by this weekend for everyone to test. Even Flite sounds pretty good, and it's free. Cepstral runs $30 per voice per simultaneous use so for home use it's still dirt cheap.

Thanks again.

oaz a écrit:

Pixel du Rezo a écrit:

Une question sur OpenNab. Je ne sais pas si c'est un problème qui vient de Violet, ou de OpenNab... J'ai essayé la reconnaissance vocale.
Le fichier wav est correctement enregistré, mais le lapin/pin ne me repond jamais, ni ce que j'attend, ni meme une humeur.
ps: j'etais inscrit à la beta, et ca marchait quelques jorus avant que je n'installe OpenNab...

Je dois avouer que je n'ai pas trop eu le temps de me pencher sur la question. D'après ce que tu dis, il y a surement un bug dans le code de OpenNab qui envoie les données audio vers nabaztag.com.
Je regarde ça dès que possible. Je dois d'abord voir si la reconnaissance vocale est activée sur mon compte Nabaztag...

Ca y est. J'ai tiré au clair cette histoire de reco vocale qui passe pas.
La reception du HTTP POST wave chez Violet semble être très sensible aux headers HTTP envoyés.
Bref, impossible de faire un post qui va bien avec cURL mais en revenant au niveau de la socket en envoyant pile poil ce que le lapin nous a envoyé , ça se "passe bien". Entre guillements car j'arrive enfin à avoir une requête http post qui se termine par un code 200 mais au niveau de la reco vocale ça le fait pas complètement : j'en suis à ma 10ème "humeur" consécutive et ce quel que soit le mot que je prononce (meteo, radio, air, trafic, ...)
Pourtant j'ai bien vérifié avec ethereal : ce que j'envoie est exactement ce que le lapin m'a envoyé.
Je vais peut être me tenter le test qui tue : envoyer à la reco vocale un mot prononcé par le text-to-speech

Dans tous les cas, je pense poster un OpenNab 0.03 ce week-end avec les divers correctifs mentionnés dans cette file de discussion.

oaz a écrit:

Je vais peut être me tenter le test qui tue : envoyer à la reco vocale un mot prononcé par le text-to-speech

Ca y est j'ai testé.
J'ai fait prononcer 'météo' au tts.
J'ai récupéré le mp3 correspondant. Un petit coup d'audacity et j'en ai fait un wave mono 16 bits 11kHz identique à un enregistrement dans le lapin.
J'ai simulé l'envoi d'une reconnaissance vocale en utilisant ce fichier.

Le résultat ? Ben j'ai eu la météo !

Alors que quand moi je prononce le même mot, je n'ai droit qu'à une "humeur"

Alors moi je dis :
- Chapeau Violet pour le text-to-speech : claire22s parle un français si parfait que même la reco vocale y voit que du feu.
- Pour la reco vocale, zéro pointé. Je sais que dans le sud on a un peu d'accent mais là quand même...

bcp d'accent tu veux dire , avec moi la reconnaissance vocal c'est réussite à 100%,
j'ai hâte de mettre en place opennab 0.0.3