Code Review Request : 7011591 JDWP socket transport should restart interrupted system calls (EINTR)

[David Buck]

Hi Alan!

After looking at the code, I think that not-honoring the timeout would 
be a bug. I will fix this and resubmit my review request.

(I also have other changes to make based on Dmitry Samersoff's comment, 
so this may take a bit longer before the new version is submitted.

Thanks for the catch!

Cheers,
-Buck

On 08/05/11 04:40, Alan Bateman wrote:
> Daniel D. Daugherty wrote:
>> :
>>
>>> I looked at the webrev and it mostly looks okay to me. One issue is
>>> that dbgsysPoll probably should adjust the timeout when interrupted.
>>
>> Adjust it how? (Just curious)
> If the poll is interrupted close to the timeout then it will restart the
> poll with the original timeout so it will be longer than it should. For
> the debugger transport I don't think this is a big issue as the timeout
> is only used in the initial handshake (if I remember correctly). It may
> be more trouble that its worth to fix this. As to how then gettimeofday
> could be used so that we get the elapsed time of the interrupted poll
> and that could be used to adjust the timeout. There are a couple of
> example in other places in the JDK.
>
> -Alan