See cockroachdb#59194 and cockroachdb#58489 for more details.

In cockroachdb#58489 we observed a scary lack of synchronization around how we set
the ReplicaState for a given replica, and how we mark a replica as
"initialized". What this meant is that it was possible for the entry in
Store.mu.replicas to be both "initialized" and have an empty
ReplicaState. This is now more likely to bite us given the migrations
infrastructure attempts to purge outdated replicas at start up time
(when replicas are being initialized, and we're iterating through extan
replicas in the Store.mu.replicas map).

We believed this was addressed as part of cockroachdb#58378, but that appears not
to be the case. Lets re-introduce this stop-gap while we investigate.

Release note: None

60429: kv: (re-)introduce a stopgap for lack of ReplicaState synchronization r=irfansharif a=irfansharif

See #59194 and #58489 for more details.

In #58489 we observed a scary lack of synchronization around how we set
the ReplicaState for a given replica, and how we mark a replica as
"initialized". What this meant is that it was possible for the entry in
Store.mu.replicas to be both "initialized" and have an empty
ReplicaState. This is now more likely to bite us given the migrations
infrastructure attempts to purge outdated replicas at start up time
(when replicas are being initialized, and we're iterating through extan
replicas in the Store.mu.replicas map).

We believed this was addressed as part of #58378, but that appears not
to be the case. Lets re-introduce this stop-gap while we investigate.

Release note: None

Co-authored-by: irfan sharif <irfanmahmoudsharif@gmail.com>

60281: sql/pgwire: send placeholder BackendKeyData r=asubiotto,spaskob a=rafiss

fixes #13191 

Some tools expect this message to be returned at connection time and
will not connect without it. CockroachDB does not support pgwire
cancellation, but we can still send a placeholder value here, and
continue ignoring cancellation requests like we already do.

Added a small test to make sure nothing broke.

Release note (sql change): When a connection is established, CockroachDB
will now return a placeholder BackendKeyData message in the response.
This is for compatibility with some tools, but using the BackendKeyData
to cancel a query will still have no effect, just the same as before.



60429: kv: (re-)introduce a stopgap for lack of ReplicaState synchronization r=irfansharif a=irfansharif

See #59194 and #58489 for more details.

In #58489 we observed a scary lack of synchronization around how we set
the ReplicaState for a given replica, and how we mark a replica as
"initialized". What this meant is that it was possible for the entry in
Store.mu.replicas to be both "initialized" and have an empty
ReplicaState. This is now more likely to bite us given the migrations
infrastructure attempts to purge outdated replicas at start up time
(when replicas are being initialized, and we're iterating through extan
replicas in the Store.mu.replicas map).

We believed this was addressed as part of #58378, but that appears not
to be the case. Lets re-introduce this stop-gap while we investigate.

Release note: None

60441: bazel: quash unnecessary dependency on `pkg/util/uuid` from protos r=rickystewart a=rickystewart

This dependency can be replaced with a few `# keep` deps in a few choice
proto targets, which is what we should have done the whole time anyway.
This fixes build failures elsewhere in tree -- for example,
`pkg/util/uuid:uuid_test`, which doesn't play nicely with `rules_go` in
the presence of this dependency.

Fixes #59778.

Release note: None

Co-authored-by: Rafi Shamim <rafi@cockroachlabs.com>
Co-authored-by: irfan sharif <irfanmahmoudsharif@gmail.com>
Co-authored-by: Ricky Stewart <ricky@cockroachlabs.com>