-
Notifications
You must be signed in to change notification settings - Fork 18k
net/http: apparent deadlocks in TestDisableKeepAliveUpgrade on longtest builders #43073
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Comments
This regression was partially masked in the builders by #42942. |
Hmm, interesting that it only occurs on Windows. Looks like the |
Nothing stands out, the test may be surfacing an unrelated regression. |
Actually I'm going to push up a change that may fix it. |
1. The test now checks the response status code. 2. The transport does not set `Connection: Close` if DisableKeepAlive is set but the request is a HTTP/1.1 protocol upgrade request. Updates golang#43073
1. The test now checks the response status code. 2. The transport does not set `Connection: Close` if DisableKeepAlive is set but the request is a HTTP/1.1 protocol upgrade request. Updates golang#43073
1. The test now checks the response status code. 2. The transport does not set `Connection: Close` if DisableKeepAlive is set but the request is a HTTP/1.1 protocol upgrade request. Updates golang#43073
Opened #43086 |
1. The test now checks the response status code. 2. The transport has been changed to not set `Connection: Close` if DisableKeepAlive is set but the request is a HTTP/1.1 protocol upgrade. Updates golang#43073
1. The test now checks the response status code. 2. The transport has been changed to not set `Connection: Close` if DisableKeepAlive is set and the request is a HTTP/1.1 protocol upgrade. Updates golang#43073
Change https://golang.org/cl/276375 mentions this issue: |
1. The test now checks the response status code. 2. The transport has been changed to not set `Connection: Close` if DisableKeepAlive is set and the request is a HTTP/1.1 protocol upgrade. Updates golang#43073
1. The test now checks the response status code. 2. The transport has been changed to not set `Connection: Close` if DisableKeepAlive is set and the request is a HTTP/1.1 protocol upgrade. Updates golang#43073
1. The test now checks the response status code. 2. The transport has been changed to not set "Connection: Close" if DisableKeepAlive is set and the request is a HTTP/1.1 protocol upgrade. Updates golang#43073
1. The test now checks the response status code. 2. The transport has been changed to not set "Connection: Close" if DisableKeepAlive is set and the request is a HTTP/1.1 protocol upgrade. Updates golang#43073
1. The test now checks the response status code. 2. The transport has been changed to not set "Connection: Close" if DisableKeepAlive is set and the request is a HTTP/1.1 protocol upgrade. Updates golang#43073
1. The test now checks the response status code. 2. The transport has been changed to not set "Connection: Close" if DisableKeepAlive is set and the request is a HTTP/1.1 protocol upgrade. Updates #43073 Change-Id: I9977a18b33b8747ef847a8d11bb7b4f2d8053b8c GitHub-Last-Rev: f809ceb GitHub-Pull-Request: #43086 Reviewed-on: https://go-review.googlesource.com/c/go/+/276375 Run-TryBot: Bryan C. Mills <[email protected]> TryBot-Result: Go Bot <[email protected]> Reviewed-by: Damien Neil <[email protected]> Trust: Dmitri Shuralyov <[email protected]>
Thanks for looking into this @nhooyr. There haven't been instances of this failure since CL 276375 was submitted, although given it happened infrequently earlier, that's not a definitive signal. Do you know if anything more needs to be done for this issue, or is it a matter of waiting for some more data before closing? I'll add okay-after-beta1 since I'm not seeing something here that needs to block beta 1. Please comment if I missed something. |
I didn't notice anything else that could cause it so yes I'd agree that it is a matter of waiting for some more data. |
Damn, that's unfortunate. Hope I have some time this weekend to investigate but if not, I think we should just revert my changes & the test and I'll try and reproduce myself first. |
Change https://golang.org/cl/285596 mentions this issue: |
2020-12-07T21:01:46-7ad6596/windows-amd64-longtest
2020-12-04T08:49:16-b67b7dd/windows-amd64-longtest
2020-12-02T16:43:58-10240b9/windows-amd64-longtest
Marking as release-blocker for Go 1.16: this is a new test, merged Dec. 1 in CL 213277 for #36381.
CC @bradfitz @nhooyr @neild @fraenkel @golang/release
The text was updated successfully, but these errors were encountered: