[PATCH] JDK-8228580 DnsClient TCP socket timeout

[Pavel Rappo]

> On 22 Aug 2019, at 08:20, Milan Mimica <milan.mimica at gmail.com> wrote:
> 
> Hi
> 
> Indeed I have forgotten. Re-attached, with aesthetic fixes to the test.

Milan, it seems that there is more work than we might've imagined initially. So as to proceed efficiently I would recommend you create a proper RFR [1] based on your most recent patch. If this is for some reason not possible, ping me and I will help you.

At first sight, the code change seems okay. I think you did a really good job mimicking the existing (UDP) code, so much so that it looks like the new code was there all the time. Sure there are some issues like a possible numeric overflow, use of currentTimeMillis instead of monotonic nanoTime, not checking timeout for negative values, etc., but they are present in the UDP code too. If not yet robust, it is at least consistent with what there is today. We might address those issues later, filing separate bugs, but I don't think we are there yet.

The attached test is, however, a bit different. I can't tell exactly what kind of issues [2] you've had with it, but there are some visible sources of potential instabilities. Create the RFR and let's discuss them.

Thanks,
-Pavel



-------------------------------------------------------------------------------
[1] https://openjdk.java.net/guide/codeReview.html
[2] "main/timeout=20 does not work for me"