Investige peek failure on Windows

[Thomasdezeeuw]

Surfaced in #1137. Only log available:

test peek ... test peek has been running for over 60 seconds

[Thomasdezeeuw]

@dtacalau do you have time to look into this?

[dtacalau]

@Thomasdezeeuw , of cpurse, please assign this issue to me.

[dtacalau]

@Thomasdezeeuw can't reproduce this issue on my machine anymore, peek test is always successful. I don't know what has changed. I even rebooted my machine couple of times and retested but w/o success.

Here's how I've tried to reproduce it:

• fetch your branch Thomasdezeeuw/win-tcpstream, enable peek test, run test -> does not reproduce;
• cherry-pick the commit from your branch to latest master -> does not reproduce;
• cherry-pick the commit to an older master where it reproduced two days ago -> does not reproduce;

Any idea what else I can try? Maybe get access to CI machine and debug it there?

[dtacalau]

Interesting read https://support.microsoft.com/en-us/help/192599/info-avoid-data-peeking-in-winsock

[dtacalau]

@Thomasdezeeuw, since I cant reproduce this localy, next stepwould be for me to add couple of println!, run the test again on CI, and see where it blocks. Could we change the settings and have the CI run tests with " --nocapture don't capture stdout/stderr of each task, allow printing directly"?

[Thomasdezeeuw]

I think you can add -- --nocapture to the end of

mio/ci/azure-test-stable.yml

Line 29 in ff1b9f2

- script: cargo ${{ parameters.cmd }}

, then it will work on the CI.

[dtacalau]

@Thomasdezeeuw ok, will try that, thanks!

[dtacalau]

@Thomasdezeeuw
Doesn't seem to reproduce on CI anymore, have a look at PR #1156. I've tried with "--nocapture" enabled and without it, in both cases peek test works fine.

Could you trigger another CI build/test on your branch, to see what happens?

[Thomasdezeeuw]

Haven't been able to reproduce this, seems like a fluke. Closing.