Re: [g-a-devel]GNOME Speech

[Marc Mulcahy <marc mulcahy sun com>]

Hey Michael,

Sorry it's been long to get this stuff checked in.  It's in the 
gnome-speech module on the gnome_speech0-1 branch

the only driver with code in at the moment is the new IBM Viavoice 
driver.  I'll try to get the bare-bones Festival driver in soon.

The problem code is in gnome-speech/synthesisdriver.c.  Have a look at 
impl_getSupportedParameters.  the concept is:

1, The driver implementation keeps a list of ParameterPriv structs which 
contain information about the supported parameters.
2. getSupportedParameters iterates through this list and dynamically 
creates GNOME_Speech_Parameter structs as you'll see in 
impl_getSupportedParameters.  thus, to the question in your previous mail, 
the CORBA_string_dup call is duping a dynamically allocated (gchar *) 
member of a ParameterPriv struct.

If you happen to have the Viavoice RPM installed, you'll notice that the 
driver doesn't crash on the call to getSupportedParameters, but that the 
client gets a parameter with an empty string as the name but all other 
fields (min, max, current, enumerated) are fine.

The Viavoice driver currently implements one parameter with a name of 
"baseline pitch".  If you change the name to simply "pitch" in 
drivers/viavoice/viavoicesynthesidriver.c:viavoice_synthesis_driver_init () 
you'll notice that the string comes through to the client OK.

I've turned on ORBIT2_DEBUG with both traces and giop, and stuff seems to 
be coming across the wire fine.

BTW: I've re-written the test (test-speech) as a boring but interactive 
console app which includes a parameter test (option 1).

Regards,

Marc

At 09:48 AM 9/11/2002 +0100, Michael Meeks wrote:
> Hi Marc,
>
> On Tue, 2002-09-10 at 15:59, Marc Mulcahy wrote:
> > Thanks Michael, will do on the ORB trace and let you know.  The other
> > interesting tidbit is that strings 10 characcters or smaller are 
> marshalled
> > properly, but over 10 characters come out on the client side as empty
> > ("").
>
>         ;-) Well - I can assure you that there are no hard-coded '10's in the
> ORB ;-) It might be down to some memory pool difference for longer
> strings or something; it sounds extremely wierd; you're sure you're
> doing:
>
>         foo->str = CORBA_string_dup ("this is more than ten"); ?
>
>   Strange-- I don't have any arrays of 10 characters char[10],
> > everything's dynamic.
>
>         Yep; same here :-)
>
> > I'm the only one who has written code in GNOME speech, except I think Bill
> > may have committed some namespace changes, so maybe just removing and then
> > re-adding wouldn't be bad.  It's not like there is much history here.
>
>         Fine; the concern is to preserve the history data at all costs; if
> there is none - it's not so much a problem.
>
> > Do you have a problem if I just go ahead and do it this way for 0.1, and
> > then we can do cvs surgery for any future renaming (which hopefully there
> > won't be much of)?  Or maybe we have to do the surgery anyway.
>
>         Sounds fine - I'd just go ahead; the rule is an advisory only - there
> is no way of making such rules stick ;-) be aware that most people do
> this for fun, so you have to make the guidelines severe to stop them
> charging right ahead.
>
> > Thanks for the help.
>
>         No problem, when it's checked in I can help with the string problem,
> let me know.
>
>         Regards,
>
>                 Michael.
>
> --
>  mmeeks gnu org  <><, Pseudo Engineer, itinerant idiot
>
> _______________________________________________
> Gnome-accessibility-devel mailing list
> Gnome-accessibility-devel gnome org
> http://mail.gnome.org/mailman/listinfo/gnome-accessibility-devel