Add leeway to revocation test #55571
Conversation
ghost
added
the
|
Tagging subscribers to this area: @bartonjs, @vcsjones, @krwq, @GrabYourPitchforks Issue DetailsWindows 7 has been observed timing out the fetch operation slightly early. For example, 15.9 seconds instead of the expected 16 seconds. Since this is OS behavior that we can't control, the test has been changed to allow a small amount of leeway in the expected timeout. Closes #55554
|
vcsjones
changed the title
There was a problem hiding this comment.
Where does "16 seconds" come from? I see TimeSpan.FromSeconds(15) above.
There was a problem hiding this comment.
~The delay * 2 (delay is up on line 37). There are two certificates that we are going to check revocation for in this chain, and each one is expected to take 8 seconds since that delay is a per-certificate timeout. So in general we would expect the whole operation to take at least 16 seconds. But, as this PR is trying to address, Windows 7 appears to be timing out just a bit late.
I see. I will reword the comment since it appears I noted the wrong thing.
There was a problem hiding this comment.
Okay - I think I misunderstood myself what was going on here and reworded the comment and tightened the timeout.
Basically, we're doing revocation checking for two certificates. Our fake server introduces an 8 second delay for each response. So we expect the response to complete in >= 16 seconds. But the issue is that it is completing in 15.99 seconds. I suspect this is because the delays are introduced using Thread.Sleep, which I suspect is imprecise and can complete sooner than 8 seconds (maybe in 7.99 seconds).
3 participants
Windows 7 has been observed timing out the fetch operation slightly early. For example, 15.9 seconds instead of the expected 16 seconds. Since this is OS behavior that we can't control, the test has been changed to allow a small amount of leeway in the expected timeout.
Closes #55554