Bump default API version to 1.45 (Moby 26.0/26.1)#3261
Conversation
|
We still need to look at raising the minimum version, but I guess that can wait;
|
|
oh! some failure; let me see if it's related |
|
Looks like it may be, and ISTR @robmry looked at some of that, but now wondering if it needs to be API-version aware (or the test rewritten to take both scenarios into account); |
|
☝️ I think this PR was related to that, but looks like that one wasn't merged; maybe it got included in another PR? |
|
I remember seeing |
|
Wondering if |
Oh; maybe it is removed, and that's the failure? But the test doesn't have a conditional on API version, so removing it would mean it'd fail on older API versions? 🤔 |
nah, the test seems to be |
|
i don't see it when running a |
|
Tried updating the test, but now getting So, looks like that wasn't right 🙈 |
|
Getting closer; only one failure remaining; |
|
what is our expectation here? meaning, are the tests "wrong" or is the engine's behavior slightly out-of-spec from what we were expecting? |
|
Trying to figure that out honestly; thought let me update the test to see what it actually sees, then check with @robmry and @akerouanton if this is the expected result here. 😂 |
|
LOL, and now it's back to; Is there some race condition? Or did I just broke it again? (the "expected" vs "actual" part is also confusing me for sure here 😂 ) |
a94f0ac to
2722873
Compare
|
Unfortunately, I don't actually know anything about version negotiation in the Python code. However, I would imagine it's mostly only ever been exercised in the other direction, e.g. to avoid sending a newly added field/query param to an older daemon. The test logic might also be more for broad test suite compatibility rather than comprehensive matrix coverage. (That is, I'm suggesting it's not unreasonable to be somewhat suspicious of both the library code & tests in this regard.) |
|
Oh! Looks like it's green now! Guess it's time for me to squash commits, and to discuss if these changes are all correct with the networking folks 🥹 |
2722873 to
9f8175e
Compare
| # Aliases no longer include the container's short-id in API v1.45. | ||
| assert (attrs['NetworkSettings']['Networks'][net_name]['Aliases'] | ||
| is None) | ||
| assert (attrs['NetworkSettings']['Networks'][net_name]['DriverOpts'] | ||
| is None) |
There was a problem hiding this comment.
FWIW, I think I tried adding an assertion for DNSNames here, but that didn't work; wondering if something needs to be added to make that field being propagated here, and if that should be checked for.
For reference, here's what inspecting a container created with similar options looks like (Aliases is null, and both the short-ID and custom hello alias are shown in DNSNames);
docker inspect --format='{{json .NetworkSettings.Networks }}' hello | jq .
{
"foo": {
"IPAMConfig": null,
"Links": null,
"Aliases": null,
"MacAddress": "02:42:ac:15:00:02",
"NetworkID": "d527fac86492c97c660830566d00d410281b8b67cc2f9d85fc60229d5d5d76ee",
"EndpointID": "4309540d4d79c6fbd21e883c867657bd5550b54fe55d1a623d9858bbf1846ee5",
"Gateway": "172.21.0.1",
"IPAddress": "172.21.0.2",
"IPPrefixLen": 16,
"IPv6Gateway": "",
"GlobalIPv6Address": "",
"GlobalIPv6PrefixLen": 0,
"DriverOpts": null,
"DNSNames": [
"hello",
"e9e430344c84"
]
}
}- Update API version to the latest maintained release. 0 Adjust tests for API 1.45 Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
9f8175e to
e47e966
Compare
Update API version to the latest maintained release.