Re: [tuxdroid-user] Daemon to USB connection

[David B. <da...@ja...>]

On Sat, 02 Jun 2007 10:01:06 +0200, neimad <ror...@gm...> wrote=
:

> [...] I always put the comment on the line before the statement, so
> I would write your example above as follows:
>
>     /* wait 200ms for the pull-up to rise before next standalone
>      * behavior otherwise position switches signals are wrong */
>     event_timer =3D 2;
>

Thanks, that's what I wanted to know, though I'm so used to it that I'll=
  =

have a hard time ... ;-)
What about defines, enums, etc., you'd put the comment/description of  =

define_1 on the right or on the line before?
enum my_enum {
     define_1,
     define_2,
}

>> There's the same for function calls, should we do
>>      void my_long_function_call(parameter_1, parameter_2,  parameter_=
3,
>>                                 parameter_4)
>> or
>>      void my_long_function_call(parameter_1, parameter_2,  parameter_=
3,
>>              parameter_4)
>
> Both styles are ok and I have used both. Depends on the coding style
> conventions. (For the second one, it is advised to indent the second
> part with two indentation levels.)
>
> I prefer the first one though, as I find it a bit cleaner (and it is
> how Emacs indents C by default ;-)).

vi defaults to the second, I think we'll never agree on this ;-)

> [...]
> Hmm, ok then. If you have pointers on embedded-specific programming
> practices, I'd be very interested (if only to avoid committing code
> considered bad from an embedded pov).

Looking into this again (to find you some pointers), I'm not sure my  =

previous statement was right, I'll have to do some tests again. This was=
  =

discussed on the avrlibc mailing list some time ago. But what I'm sure n=
ow  =

is that a swicth can be very efficient if all cases are incremental as i=
t  =

can use a lookup table, otherwise it generates a if-then-else sequence  =

which takes much more space. In i2c.c (tuxcore) I created the lookup tab=
le  =

manually as the compiler never wanted to do it, and I saved more than 10=
0  =

bytes doing so.
I'll let you know if I find something interesting.

> In commit r335:
>
>      typedef enum
>      {
>          USB_CONNECTION_CONNECT =3D 1,
>          USB_CONNECTION_DISCONNECT =3D 2,
>          USB_CONNECTION_ID_LOOKUP =3D 3,
>          USB_CONNECTION_CHANGE_ID =3D 4,
>          USB_CONNECTION_WAKEUP =3D 5,
>          USB_CONNECTION_WIRELESS_CHANNEL =3D 6,
>      } USB_ID_PARAM1;
>
> The type is badly named: "USB_ID_PARAM1" is about its *use* whereas
> its name should really be about what it *is*.

Well, it *is* the first parameter of the USB ID command which I later  =

renamed USB connection command and probably forgot about it here.

>
> Thus I would rather have the following:
>
>      typedef enum
>      {
>          USB_CONNECTION_CONNECT =3D 1,
>          USB_CONNECTION_DISCONNECT =3D 2,
>          USB_CONNECTION_ID_LOOKUP =3D 3,
>          USB_CONNECTION_CHANGE_ID =3D 4,
>          USB_CONNECTION_WAKEUP =3D 5,
>          USB_CONNECTION_WIRELESS_CHANNEL =3D 6,
>      } usb_connection_t;
>
> The type is in lower case suffixed instead of all-uppercase, it is
> suffixed with _t (a widespread convention) and it is consistent with
> its values: usb_connection_t / USB_CONNECTION_xxx

Yep, my fault, I was probably asleep at the time I wrote that. I'll fix =
 =

it, write it 100 times and go to sleep without eating ;-)

Thanks,
David