Fix bug where incorrect response is returned [carry 3845]

[thaJeztah]

• depends on cil/command: use dummy client for build-tests #3903
• replaces Fix bug where incorrect response is returned #3845
• fixesIncorrect error message when server is unreachable #3844

When server is unreachable and docker checkpoint (or any command that needs to check the server type) is run, incorrect error was returned.

When checking if the daemon had the right OS, we compared the OSType from the clients ServerInfo(). In situations where the client cannot connect to the daemon, a "stub" Info is used for this, in which we assume the daemon has experimental enabled, and is running the latest API version.

However, we cannot fill in the correct OSType, so this field is empty in this situation.

This patch only compares the OSType if the field is non-empty, otherwise assumes the platform matches.

before this:

docker -H unix:///no/such/socket.sock checkpoint create test test
docker checkpoint create is only supported on a Docker daemon running on linux, but the Docker daemon is running on

with this patch:

docker -H unix:///no/such/socket.sock checkpoint create test test
Cannot connect to the Docker daemon at unix:///no/such/socket.sock. Is the docker daemon running?

- A picture of a cute animal (not mandatory but encouraged)

[thaJeztah]

In cases where we were not able to connect with the daemon, we set Experimental to "true", but as we don't know the daemon's OSType, we can't set that value, so in that case we should ignore these;

cli/cli/command/cli.go

Lines 342 to 343 in 7240f70

	// Default to true if we fail to connect to daemon
	cli.serverInfo=ServerInfo{HasExperimental: true}

We can probably optimise and detect that we didn't have a valid ping, and skip all of these checks, but leaving that for a follow-up.

[thaJeztah]

Looks like I have one test to fix / verify;

#19 64.01 === FAIL: cmd/docker TestBuildkitDisabled (0.00s)
#19 64.01 builder_test.go:137: line 0: expected "" to end with "DEPRECATED: The legacy builder is deprecated and will be removed in a future release."
#19 64.01 builder_test.go:137:

[codecov-commenter]

Codecov Report

Merging #3901 (f33ef47) into master (f33ef47) will not change coverage.
The diff coverage is n/a.

❗ Current head f33ef47 differs from pull request most recent head 20ba591. Consider uploading reports for the commit 20ba591 to get more accurate results

Additional details and impacted files

@@ Coverage Diff @@## master #3901 +/- ##
=======================================
Coverage 59.22% 59.22% =======================================
Files 287 287 Lines 24723 24723 =======================================
Hits 14641 14641 Misses 9197 9197 Partials 885 885 

[thaJeztah]

Hm.. right, so the "unit" test is actually using a docker client, which tries to connect to a daemon (which isn't there);

=== FAIL: cmd/docker TestBuildkitDisabled (0.00s)
#19 76.40 ping err: Cannot connect to the Docker daemon at unix:///var/run/docker.sock. Is the docker daemon running?

Then the check if buildkit must be used is based on the (non-existing) ping result; if that's "windows", it's okay, but it it's non-windows, then it should print an error. As (at this stage) we don't know if the daemon is windows or linux, we probably shouldn't produce an error (which is what this PR does), but the test expects that error to happen.

[thaJeztah]

Thx!

[thaJeztah]

Whoops, and I just realised I introduced a regression here; this was deliberately put inline to allow it to be executed conditionally (see b397391 / #2424)

Opening a PR to fix that 😅