Support for Patch

[VaughnVernon]

Hi Cedric!

This is a surprisingly simple fix compared to what Netty did.

May we request your help in patching our socket channel reader for Linux? We can pay a stipend for a solution that works across various Linux editions. So far we know we have problems on Heroku Ubuntu and CentOS, and it may also be a problem on AWS CentOS (only partially confirmed).

vlingo/xoom-wire#28

Thanks in advance.

[cedric780]

Hello,

I have a full time job and already spent one working week to narrow down this bug, look for workarounds, and file an issue to Oracle. So I am not available for support.

Oracle created an official issue since my first commit, I updated the README to
add the link and add some more details.

The issue seems to be fixed on JDK 11, the demonstrator does not trigger it. The best solution would then be to upgrade the JDK.

[VaughnVernon]

Hi @cedric780

Thanks for your reply and your awesome work! No problem not having time to help.

We are a toolset and thus baselined on JDK 8 as lowest common denominator. It looks like it supports JDK 8. True?

Do you think that this fix is straightforward enough that we can support it in our JDK 8 based toolset by just patching it in a runtime when Linux is detected?

Thanks again!

[cedric780]

I don't master enough the internals of the JDK to give a clear answer.
I did some quick tests on my own machine and without getting other issues but these were not extensive ones.

My boss preferred to downgrade the client Web widget to a version whose behavior does not trigger the bug .

I think you should do some extensive tests and maybe remove some log console outputs before applying the patch.

[VaughnVernon]

@cedric780 Would you mind explaining further?

My boss preferred to downgrade the client Web widget to a version whose behavior does not trigger the bug .

Is this a widget that is usable by a Java toolset such as ours to solve this problem without patching code?