Re: [Qemu-devel] [PATCH v4 4/4] hw/input/adb.c: implement QKeyCode support

[Programmingkid]

On Mar 13, 2016, at 11:40 AM, Peter Maydell wrote:

> On 12 March 2016 at 05:40, Programmingkid <address@hidden> wrote:
>> 
>> On Mar 11, 2016, at 10:30 PM, Peter Maydell wrote:
>> 
>>> 
>>>> +    }
>>>> +    keycode = s->data[s->rptr];
>>>> +    if (++s->rptr == sizeof(s->data)) {
>>>> +        s->rptr = 0;
>>>>    }
>>>> +    s->count--;
>>>> +
>>>> +    obuf[0] = keycode;
>>> 
>>> You are still trying to put a two byte keycode (ADB_KEY_POWER)
>>> into this one-byte array slot. I don't know what the right way to
>>> send a two-byte keycode is but this is obviously not it, as
>>> I said before.
>>> 
>>>> +    /* NOTE: could put a second keycode if needed */
>>>> +    obuf[1] = 0xff;
>>>> +    olen = 2;
>>>> +
>>>>    return olen;
>>>> }
>> 
>> Is this ok?
>> 
>>    /* The power key is the only two byte value key, so it is a special case. 
>> */
>>    if (keycode == (ADB_KEY_POWER & 0x00ff)) {
>>        obuf[0] = ADB_KEY_POWER & 0x00ff;
>>        obuf[1] = ADB_KEY_POWER & 0xff00 >> 8;
>>        olen = 2;
>>    } else {
>>        obuf[0] = keycode;
>>        /* NOTE: could put a second keycode if needed */
>>        obuf[1] = 0xff;
>>        olen = 2;
>>    }
>> 
>> The keycode value comes from an 8 bit array so holding the
>> full value of the power key is not possible.
> 
> Ah, I hadn't noticed that -- that is not a good approach.
> You should either:
> (a) deal with the fact that ADB_KEY values may be 16 bits
> all the way through (including having that array be uint16_t
> rather than uint8_t)

I did try this but a bunch of errors showed up. To see all the errors just make 
this change in adb.c:

typedef struct KBDState {
    /*< private >*/
    ADBDevice parent_obj;
    /*< public >*/

    uint16_t data[128];      <----- was uint8_t
    int rptr, wptr, count;
} KBDState;

The errors:

/include/migration/vmstate.h:248:48: error: invalid operands to binary - (have 
'uint8_t (*)[256]' and 'uint16_t (*)[128]')
 #define type_check_array(t1,t2,n) ((t1(*)[n])0 - (t2*)0)

/include/migration/vmstate.h:261:6: note: in expansion of macro 
'type_check_array'
      type_check_array(_type, typeof_field(_state, _field), _num))

So I'm not sure now changing the type to 16 bit is the best thing to do. It 
would require a lot more changes to other files.