Subject: Re: [Bluez-devel] bccmd patch, find connection handle by bdaddr
From: Marcel Holtmann <marcel@holtmann.org>
To: bluez-devel@lists.sourceforge.net
In-Reply-To: <20050809144239.29D538AD@arbetsmyra.dyndns.org>
References: <20050808115620.EEB243BF@arbetsmyra.dyndns.org>
	 <1123522342.7751.49.camel@pegasus>
	 <20050809144239.29D538AD@arbetsmyra.dyndns.org>
Content-Type: text/plain
Message-Id: <1124206111.7466.7.camel@pegasus>
Mime-Version: 1.0
Sender: bluez-devel-admin@lists.sourceforge.net
Errors-To: bluez-devel-admin@lists.sourceforge.net
Reply-To: bluez-devel@lists.sourceforge.net
Date: Tue, 16 Aug 2005 17:28:30 +0200

Hi Ronny,

> > and it includes a free(NULL) bug.
> 
> That's actually ok. According to both K&R and the GNU manuals it is 
> fully legal. The C-lib checks for NULL and in that case does nothing. I 
> can add an extra check though if you want the double safetey. B.T.W I 
> saw your most resent commit on  libs/src/bluetooth.c  contains a 
> bt_malloc(). Is that to be used instead from now on?

this is actually a style issue. Do the exit/failure path in the function
right and this doesn't happen.

The bt_malloc() is only for internal stuff to make sure that the library
is linked to the same malloc() and free() function that are used in the
tools. For bccmd this makes no difference and so use malloc() and free()
for now.

> > The handle is uint32_t 
> 
> Really? Strange. The kernel #includes define the handle as an uint16_t 
> only (struct hci_conn). And according to the v2.0 spec only the least 
> 12-bits is actually used. I then used a signed 32 to be able to do a 
> return of both those unsigned 12/16 bits and a -1 error indicator. 
> Perhaps I've missed something, where does the remaining data come from?

My mistake. Don't know was on my mind when I wrote that. Maybe I was
already on holiday. Sorry for the confusion.

> > and you should keep the endian conversion in mind.
> 
> The handle returned from ioctl() is in host byte order already since 
> the kernel makes a conversion. What to do next with it one can discuss. 
> Especially when it comes to be used with "complex" lengthed Varid's 
> it's easy to make mistakes due to quite odd BCCMD protocol byte 
> ordering (where 32-bits are neither of big/little-endian). I was 
> thinking it's probably easiest to leave the handle as-is for all of 
> those many:
> 	buf[0] = handle & 0xff;
> 	buf[1] = handle >> 8;
> which complex BCCMD require. I'm open to suggestions though.

Look at hcitool.c and see how we convert it. Then we have to correct it
for the BCCMD protocol. I know that this is odd, but that's the only
sane way. Otherwise parts of the code gets reused and the endian thing
will be mixed up.

Regards

Marcel


-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
_______________________________________________
Bluez-devel mailing list
Bluez-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bluez-devel