Repository navigation
attempt to stabilize CopyToAsync_AllDataCopied_Large for Tls13 - #44741
Conversation
|
Tagging subscribers to this area: @dotnet/ncl DetailsIssue Details
|
|
/azp run runtime-libraries-coreclr outerloop-linux |
|
Azure Pipelines successfully started running 1 pipeline(s). |
|
Why does this make it better? Seems like it shouldn't matter. |
|
I'm still digging into it. But here are some (unverified) thoughts. For TLS 1.2 (and less) when handshake is done the streams are really identical and fully interchangeable. With TLS1.3 part of the handshake piggy-backs on first data chunks - that was done to limit round trips and that is reason why we always hit the "RENEGOTIATION" path on Windows. The test sends unidirectional data and and I'm not sure if we have mechanism to send message unless somebody call encrypt data. Now the interesting part: |
|
Yeah, very weird about the RST. |
|
I think it may be timing after all @geoffkizer . I opened #44764 to track what I see as underlying problem. |
|
/azp run runtime-libraries-coreclr outerloop-linux |
|
Azure Pipelines successfully started running 1 pipeline(s). |
|
/azp run runtime-libraries-coreclr outerloop-linux |
|
Azure Pipelines successfully started running 1 pipeline(s). |
|
/azp run runtime-libraries-coreclr outerloop-linux |
|
Azure Pipelines successfully started running 1 pipeline(s). |
|
test failures are unrelated. |

I'm not 100% what is going on but with this change I did several hundreds runs and I did not see any failure.
The TLS1.3 are not quite symmetric as 1.2.