<AWT Dev> [8] Review request for 7117595: ArrayIndexOutOfBoundsException in Win32GraphicsEnvironment if display is removed

[Anthony Petrov]

On 09/23/13 17:38, sergey malenkov wrote:
>> I suggest to check both bounds for the default screen index. Something
>> like the following should do: max(0, min(index, screens.length - 1)).
> Seems we should fix other environments too.

+1.

>> Also, I assume that getScreenDevices() can't return null or an empty
>> array, can it? For added safety, we might want to check this too, and
>> return null in this case.
> The getScreenDevices method can't return null, but it could be an empty
> array. There are many places, where the getDefaultScreenDevice method is
> used. Our code does not expect the null value. We could throw
> HeadlessException instead. What do you think?

That indeed seems logical. Moreover, the HeadlessException doesn't have 
a specification yet (see 8016313). However, I think that AWT doesn't 
consider itself headless even if the devices array is empty. Headless is 
a special mode, after all. I'd go with returning null for now - other 
code will catch NPEs, but that's OK. Later we may decide whether 
consider this case headless or not.

--
best regards,
Anthony

>
>> --
>> best regards,
>> Anthony
>>
>> On 09/23/13 16:47, sergey malenkov wrote:
>>> Hello,
>>>
>>> Could you please review the following fix:
>>> fix:http://cr.openjdk.java.net/~malenkov/7117595.8.0/
>>> bug:https://bugs.openjdk.java.net/browse/JDK-7117595
>>>
>>> Thanks,
>>> SAM
>