In particular, fix the description of ServiceChangeTracker.Update's
return value, and point out that it's different from
EndpointsChangeTracker.EndpointSliceUpdate's (though fortunately, in a
way that doesn't matter for any existing code).
EndpointSliceCache was using the name "endpointInfo" to refer to two
different data types (most egregiously in addEndpoints(), which had a
variable named `endpoint` of type `*endpointInfo` and a variable named
`endpointInfo` of type `Endpoint`).
Continue using "endpointInfo" in places that refer to proxy.Endpoint /
BaseEndpointInfo, since that's consistent with other code, but rename
the local "cache of the Endpoints field of an EndpointSlice" type from
"endpointInfo" to "endpointData". Likewise, rename endpointSliceInfo
to endpointSliceData, for consistency.
Put the ServiceChangeTracker and EndpointsChangeTracker definitions at
the top of the files, and put the ServicePortMap and EndpointsMap
definitions before their methods.
(No code changes.)
Move the ServicePort/BaseServicePortInfo types to serviceport.go.
Move the Endpoint/BaseEndpointInfo types to endpoint.go.
To avoid confusion with the new filenames, rename service.go to
servicechangetracker.go and endpoints.go to endpointschangetracker.go.
(No code changes; this just moves some code from types.go and
services.go to serviceport.go, and some code from types.go and
endpoints.go to endpoint.go.)
(This would become an error rather than a warning once we try to move
this code to another file.)
Also rename an "ok" variable to "exists" since that what it really
means.
ServicePortMap.merge had a giant comment explaining its return value,
but nothing ever used that return value.
ServicePort had an InternalTrafficPolicy() method, but nothing used it
(because it was redundant with InternalPolicyLocal().)
When dealing with unschedulable pods, the intent was to deallocate only claims
which are allocated and use delayed allocation. That if check wasn't handled
correctly, causing also claims with immediate allocation to be considered as
candidates.
Found during code reading, probably has never occurred in practice yet.
ServicePortMap.Update() and EndpointsMap.Update() were just a tiny
wrappers around the corresponding apply() methods, which had no other
callers. So squash them together.
(Also fix the variable naming in ServicePortMap.Update() to match
other methods.)
safeIpset was a wrapper for thread-safely sharing an ipset.IPSet, but
this was unnecessary because ipset.IPSet is just a wrapper around exec
anyway and doesn't need any locking.
Restore `migratedPlugins` entry for in-tree vSphere volumes and mark them
as migrated. Otherwise the CSI volume plugin ignores in-tree vSphere
volumes and they will never get attached / mounted by CSI migration.
The entry in `migratedPlugins` was incorrectly removed during
CSIMigrationvSphere feature gate removal.
