I think you used vndr directly to revendor. We currently had to wrap it in a script to ignore a local copy of golang's archive package. Can you try running hack/vendor.sh instead (or revert the changes in vendor/archive if that's faster)

[2021-07-09T00:26:33.174Z] The result of vndr differs
[2021-07-09T00:26:33.174Z] 
[2021-07-09T00:26:33.174Z] ?? vendor/archive/tar/example_test.go
[2021-07-09T00:26:33.174Z] ?? vendor/archive/tar/reader_test.go
[2021-07-09T00:26:33.174Z] ?? vendor/archive/tar/strconv_test.go
[2021-07-09T00:26:33.174Z] ?? vendor/archive/tar/tar_test.go
[2021-07-09T00:26:33.174Z] ?? vendor/archive/tar/testdata/
[2021-07-09T00:26:33.174Z] ?? vendor/archive/tar/writer_test.go

We should update the vendor-check to update the Please vendor your package with github.com/LK4D4/vndr message accordingly; in

should be something like:

Please vendor your package with hack/vendor.sh

Otherwise LGTM.

Assuming we want this to be included in a docker 20.10 release (which currently uses the moby branch on hcsshim, I also opened microsoft/hcsshim#1066, which back ports the fix mentioned to that branch

also prepared #42613 for 20.10

@thaJeztah I did not realize there was a separate vendor script, I'll try that :)

Yes, it's a bit unfortunate that we (again) had to fork the archive package, due to a security issue in the archive package (when using under specific conditions); we should try to get rid of it (more details in #42402)

Yes, it's a bit unfortunate that we (again) had to fork the archive package, due to a security issue in the archive package (when using under specific conditions); we should try to get rid of it (more details in #42402)

I've created a new PR to fix the vendor error message and also allow the vendor script to work on Windows: #42634

Test TestNetworkDBIslands in the CI seems to be failing with lots of errors like dial tcp 172.17.0.2:10033: connect: connection refused. This seems unrelated to the changes here.

Yes, CI is super flaky since libnetwork was moved into this repository (it was flaky in that repository, but becomes more apparent here). Some PR's in flight to try to improve the situation (#42625, #42594)

Failure on arm64 is a known flaky test: #42582

=== FAIL: libnetwork/drivers/bridge TestCreateParallel (0.28s)
time="2021-07-13T20:05:30Z" level=warning msg="bridge store not initialized. kv object docker/network/v1.0/bridge/net31/ is not added to the store"
time="2021-07-13T20:05:30Z" level=warning msg="bridge store not initialized. kv object docker/network/v1.0/bridge/net3/ is not added to the store"
    bridge_test.go:1133: Success should be 1 instead: 2

everything else is green

Thanks @thaJeztah! :)