[backport] Backport sweep for 8.0

[valkeyrie-ops]

Backport sweep for 8.0

Automated cherry-picks from PRs marked "To be backported".

Applied

Source PR	Title	Detail
#3459	Fix missing signalModifiedKey calls for stream commands	conflicts resolved by Claude Code
#4613	Move build-debian-old CI job to Debian Bookworm
#4676	Fix return types of three module API declarations to match definitions	conflicts resolved by Claude Code

AI resolution details are posted as comments on this PR when available.

---

Generated by valkey-ci-agent using Claude Code.

[valkeyrie-ops]

AI backport resolution: source PR #3459

Fix missing signalModifiedKey calls for stream commands

2 AI-edited files

Claude Summary

All conflicts are resolved and no markers remain. The files stay listed as "UU" in git status only because I did not run git add, per the constraints.

What I did:

src/t_stream.c (+28/-3): The PR adds signalModifiedKey() calls to stream commands that mutate consumer-group/PEL metadata. All nine conflicts were caused by two unrelated drifts on the newer branch: an o->ptr → objectGetVal(o) accessor refactor, and reply-ordering changes (upstream moved addReply after the notify calls, plus a new XDELEX/XACKDEL propagation block in xdelCommand). Neither exists on 8.0 — I verified objectGetVal and XDELEX appear nowhere on this branch — so I kept 8.0's ->ptr accessors and reply ordering, and merged in only the PR's additions:

• xgroupCommand: moved the MKSTREAM signalModifiedKey from stream creation to successful group creation, and added signals to SETID, DESTROY, CREATECONSUMER, and DELCONSUMER.
• xsetidCommand and xackCommand: added the missing signal calls.
• xclaimCommand / xautoclaimCommand: kept the incoming modified flag tracking (set on consumer creation, NACK removal, claims, and last-id propagation) with a single signalModifiedKey per command. The int modified = 0; declarations and the xreadCommand modified_stream changes had already auto-merged cleanly.
• xdelCommand: kept HEAD's version unchanged — the PR's edit there only reordered the xdel event for XDELEX DELREF, a command 8.0 doesn't have.

tests/unit/type/stream.tcl (no net change): The conflict was a ~900-line block of XACKDEL/XDELEX tests that exist only on the newer branch, with an empty HEAD side. The PR's actual change to this file was deleting one of those XDELEX tests, which isn't on 8.0, so I dropped the incoming block entirely and the file matches HEAD.

tests/unit/type/stream-cgroups.tcl (auto-merged, +95): Three new WATCH-dirtying tests. I verified the helpers they use (valkey_deferring_client, wait_for_blocked_clients_count) already exis
... summary truncated.

AI-edited files requiring review

• src/t_stream.c — view diff
• tests/unit/type/stream.tcl — view diff

Full backport commit diff: commit 2fcb42aafaed.

Please review these AI resolutions for correctness before merging.

[coderabbitai]

Important

Review skipped

Bot user detected.

To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Advanced

Run ID: ac84642a-7571-4f48-9dfa-07d60472be8d

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[valkeyrie-ops]

AI backport resolution: source PR #4676

Fix return types of three module API declarations to match definitions

1 AI-edited file

Claude Summary

The conflict in src/valkeymodule.h is resolved and the tree is clean of conflict markers. Here's a summary of what I found and did:

The conflict: The source PR changes three function-pointer declarations in the module API table from void to int return type (ValkeyModule_FreeModuleUser, ValkeyModule_ACLAddLogEntry, ValkeyModule_ACLAddLogEntryByUserName) so they match their int-returning implementations in module.c. The cherry-pick conflicted because the upstream diff's context includes a ValkeyModule_ACLCheckPermissions declaration that sits between the ACL check functions and the ACL log functions on unstable — that API doesn't exist on the 8.0 branch.

The resolution: I kept the incoming side's two int (*ValkeyModule_ACLAddLogEntry...) declarations (the actual PR change) and dropped the ACLCheckPermissions declaration, since it's not part of this PR and doesn't exist anywhere on 8.0 (verified by grep against HEAD). The ValkeyModule_FreeModuleUser void→int change at valkeymodule.h:1536 had already auto-merged cleanly, as had the doc-comment addition in module.c.

Verification:

• No conflict markers remain in src/valkeymodule.h.
• The implementations on 8.0 already return int (VM_FreeModuleUser at module.c:9596, VM_ACLAddLogEntry at module.c:9796, VM_ACLAddLogEntryByUserName at module.c:9810), so the new declarations match.
• The resolved diff is 12 insertions / 10 deletions across the same two files — identical in size to the upstream PR.
• Only the two files from the cherry-pick were touched; nothing was staged or committed.

AI-edited files requiring review

• src/valkeymodule.h — view diff

Full backport commit diff: commit 0149f76d1047.

Please review these AI resolutions for correctness before merging.

[valkeyrie-ops]

Automatic follow-up for the current backport head.

I did not push a fix: The failure is not deterministic: the identical test at the identical commit passed in ~29 sibling jobs of the same CI run (typically in ~20s), and the other failing jobs in this run failed on entirely different cluster tests (slot-ownership.tcl, replica-migration), the classic scatter pattern of flaky cluster-convergence tests. No default-branch candidate commit touches manual-takeover.tcl or this convergence path (FAILOVER_AUTH_NACK is a later unrelated feature, confirmed absent from this checkout; the takeover path uses no election votes). The only 'fixes' available would be inflating timeouts or weakening the recovery assertion, which the rules forbid; therefore no safe, attributable fix exists.

Looked at the failure from this run.

Other checks also failed in that run; re-invoke with the same command to address the next one:

• test-almalinux9-tls-module
• test-sanitizer-address (clang)
• test-sanitizer-undefined (gcc)

[valkeyrie-ops]

Automatic follow-up for the current backport head.

I did not push a fix: This is a timing-dependent race exposing a real product bug: FAILOVER_AUTH_ACKs carry no election epoch, so a single voter's ACKs from two consecutive elections are double-counted (src/cluster_legacy.c:3628-3638). The upstream mitigation direction (FAILOVER_AUTH_NACK epoch echo, 5c24e5d) depends on the NACK feature, which is confirmed absent from this 8.0 branch (no FAILOVER_AUTH_NACK anywhere in src/), so it does not port cleanly, and no self-contained test-scaffolding fix exists that would not weaken what the test verifies.

Looked at the failure from this run.

Other checks also failed in that run; re-invoke with the same command to address the next one:

• test-ubuntu-jemalloc-fortify (Cluster should eventually be up again in tests/unit/cluster/manual-takeover.tcl)

[valkeyrie-ops]

Automatic follow-up for the current backport head.

I did not push a fix: This is a timing-dependent/flaky failure: whether keys get evicted during init_test depends on how quickly the server releases the prior test's client buffer memory, and the helper has no wait/synchronization. It is not a product bug and not deterministic scaffolding breakage. None of the discovered default-branch commits touch tests/unit/maxmemory.tcl or this eviction path, so there is no upstream fix to port, and authoring a fix would mean designing new flakiness mitigation (e.g. waiting for used_memory to settle), which falls outside a safe mechanical fix per the rules.

Looked at the failure from this run.