net: TestBuffers_WriteTo failures

[gopherbot]

#!watchflakes
post <- pkg == "net" && test == "TestBuffers_WriteTo"

Issue created automatically to collect these failures.

Example (log):

--- FAIL: TestBuffers_WriteTo (38.65s)
    --- FAIL: TestBuffers_WriteTo/WriteTo/0 (25.92s)
        writev_test.go:120: dial tcp 127.0.0.1:50922: connect: operation timed out

— watchflakes

[gopherbot]

Found new dashboard test flakes for:

#!watchflakes
post <- pkg == "net" && test == "TestBuffers_WriteTo"

2023-04-25 14:24 darwin-amd64-13 go@14f833f1 net.TestBuffers_WriteTo (log)

--- FAIL: TestBuffers_WriteTo (38.65s)
    --- FAIL: TestBuffers_WriteTo/WriteTo/0 (25.92s)
        writev_test.go:120: dial tcp 127.0.0.1:50922: connect: operation timed out

— watchflakes

[bcmills]

(attn @ianlancetaylor @neild)

[ianlancetaylor]

In both cases the timeout is being reported in the call to withTCPConnPair. So this may not have anything to do with TestBuffers_WriteTo as such.

[ianlancetaylor]

This appears to have happened exactly once so far. It seems clearly impossible based on the code, and on the fact that other tests are using withTCPConnPair successfully. Let's see if it happens again.

[gopherbot]

Found new dashboard test flakes for:

#!watchflakes
post <- pkg == "net" && test == "TestBuffers_WriteTo"

2024-02-23 08:34 android-amd64-emu go@ad377e90 net.TestBuffers_WriteTo (log)

--- FAIL: TestBuffers_WriteTo (0.02s)
    --- FAIL: TestBuffers_WriteTo/WriteTo/1025 (0.00s)
        writev_test.go:118: accept tcp4 127.0.0.1:46781: accept4: bad file descriptor
        writev_test.go:118: dial tcp 127.0.0.1:46781: connect: connection reset by peer

— watchflakes

[bcmills]

That most recent failure was almost certainly #65857.