Fixes for Valkey 8.0.5

[vitarb]

This backports all remaining PRs from Valkey 8.0 (view)

• ✔️ PR Only mark the client reprocessing flag when unblocked on keys #2109
• ✔️ PR Fix memory corruption in sharded pubsub unsubscribe #2137
• ✔️ PR Free module context even if there was no content written in auxsave2 #2132
• ✔️ PR CLIENT UNBLOCK should't be able to unpause paused clients #2117
• ✔️ PR Detect SSL_new() returning NULL in outgoing connections #2140
• ✔️ PR Correctly cast the extension lengths #2144
• ✔️ PR Incorporate Redis CVE for CVE-2025-27151 #2146
• ✔️ PR Fix replica can't finish failover when config epoch is outdated #2178
• ✔️ PR Fix cluster myself CLUSTER SLOTS/NODES wrong port after updating port/tls-port #2186
• ✔️ PR Add packet-drop to fix the new flaky failover test #2196 (is a test dependency of Fix replica can't finish failover when config epoch is outdated #2178)
• ✔️ PR Add a filter option to drop all cluster packets #1252 (is a test dependency of Fix replica can't finish failover when config epoch is outdated #2178)
• ✔️ PR Minor log fixes when failover auth denied due to slot epoch #1341 (is a test dependency of Fix replica can't finish failover when config epoch is outdated #2178)
• ✔️ PR Trigger the election as soon as possible when doing a forced manual failover #1067
• ✔️ PR Make manual failover reset the on-going election to promote failover #1274 (is a test dependency of Fix replica can't finish failover when config epoch is outdated #2178, also we can backport it)
• ✔️ Avoid timing issue in diskless-load-swapdb test #1077
  • Fixes [exception]: Executing test client: CLUSTERDOWN The cluster is down.
• ✔️ Test coverage for ECHO for reply schema validation #1549
  • Fixes reply-schema-validator job
• ✔️ Try to stabilize aof test #2399
• ✔️ Disable clang-format in 8.0 to simplify backporting
• ✔️ update build-debian-old to use Bullseye instead of EOL buster #2345
• ✔️ Fix defrag timeouts in 8.0
• ✔️ Ensure empty error tables in scripts don't crash Valkey #2229
• ✔️ Fix client tracking memory overhead calculation #2360
• ✔️ Converge divergent shard-id persisted in nodes.conf to primary's shard id #2174
• ✔️ Fix assumptions that pthread functions set errno #2526
• ✔️ Fix pre-size hashtables per slot when reading RDB files #2466

[hpatro]

Let's update the version.h file and update release notes under 00-RELEASENOTES
#2173

[hpatro]

@enjoy-binbin Seems like the replica is able to get the epoch update, possibly flaky?

*** [err]: Replica can update the config epoch when trigger the failover - automatic in tests/unit/cluster/failover2.tcl
Expected '5' not equal to '5' (context: type eval line 29 cmd {assert_not_equal $R0_config_epoch [dict get [cluster_get_node_by_id 3 $R0_nodeid] config_epoch]} proc ::test)
*** [err]: Replica can update the config epoch when trigger the failover - manual in tests/unit/cluster/failover2.tcl
Expected '5' not equal to '5' (context: type eval line 29 cmd {assert_not_equal $R0_config_epoch [dict get [cluster_get_node_by_id 3 $R0_nodeid] config_epoch]} proc ::test)

https://github.com/valkey-io/valkey/actions/runs/15642106732/job/44071643820?pr=2210

[enjoy-binbin]

yeah, that is flaky i guess, it is odd since drop all the packets, i will take a look this day.

[zuiderkwast]

One PR missing? I see 10 commits but 11 PRs mentioned and 11 in To be backported.

[enjoy-binbin]

One PR missing? I see 10 commits but 11 PRs mentioned and 11 in To be backported.

PR #1539 is an empty commit i saw. (That was already in 8.0.3)

[codecov]

Codecov Report

❌ Patch coverage is 83.33333% with 21 lines in your changes missing coverage. Please review.
✅ Project coverage is 70.78%. Comparing base (f77cbfb) to head (85a2b9d).

Files with missing lines	Patch %	Lines
src/kvstore.c	69.23%	4 Missing ⚠️
src/io_threads.c	0.00%	3 Missing ⚠️
src/rdb.c	40.00%	3 Missing ⚠️
src/bio.c	66.66%	2 Missing ⚠️
src/config.c	50.00%	2 Missing ⚠️
src/valkey-check-aof.c	33.33%	2 Missing ⚠️
src/blocked.c	87.50%	1 Missing ⚠️
src/call_reply.c	0.00%	1 Missing ⚠️
src/cluster_legacy.c	97.14%	1 Missing ⚠️
src/module.c	0.00%	1 Missing ⚠️
... and 1 more

Additional details and impacted files

@@            Coverage Diff             @@
##              8.0    #2210      +/-   ##
==========================================
+ Coverage   70.73%   70.78%   +0.05%     
==========================================
  Files         114      114              
  Lines       62991    63062      +71     
==========================================
+ Hits        44554    44641      +87     
+ Misses      18437    18421      -16     

Files with missing lines	Coverage Δ
src/acl.c	88.82% <100.00%> (ø)
src/adlist.c	75.38% <100.00%> (+0.38%)	⬆️
src/db.c	88.67% <100.00%> (ø)
src/debug.c	53.62% <ø> (ø)
src/eval.c	57.15% <100.00%> (ø)
src/evict.c	98.87% <100.00%> (+0.37%)	⬆️
src/functions.c	95.72% <100.00%> (+0.02%)	⬆️
src/listpack.c	91.63% <100.00%> (+1.43%)	⬆️
src/pubsub.c	97.04% <100.00%> (ø)
src/replication.c	87.81% <100.00%> (+0.25%)	⬆️
... and 18 more

... and 9 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[vitarb]

One PR missing? I see 10 commits but 11 PRs mentioned and 11 in To be backported.

Was already merged, but still on the to be backported list, so my merge script skipped it and mentioned as merged. No-op.

[zuiderkwast]

test-sanitizer-undefined (clang) reports https://github.com/valkey-io/valkey/actions/runs/15717322590/job/44290197737?pr=2210#step:6:278 Not sure which PR it is from...

[err]: Sanitizer error: adlist.c:63:25: runtime error: call to function engineLibraryFree through pointer to incorrect function type 'void (*)(void *)'
/home/runner/work/valkey/valkey/src/functions.c:147: note: engineLibraryFree defined here
SUMMARY: UndefinedBehaviorSanitizer: undefined-behavior adlist.c:63:25

Are the other failures just flaky tests?

[vitarb]

test-sanitizer-undefined (clang) reports https://github.com/valkey-io/valkey/actions/runs/15717322590/job/44290197737?pr=2210#step:6:278 Not sure which PR it is from...

[err]: Sanitizer error: adlist.c:63:25: runtime error: call to function engineLibraryFree through pointer to incorrect function type 'void (*)(void *)'
/home/runner/work/valkey/valkey/src/functions.c:147: note: engineLibraryFree defined here
SUMMARY: UndefinedBehaviorSanitizer: undefined-behavior adlist.c:63:25

Are the other failures just flaky tests?

TIMEOUTs look like flakes. I've added retries on CLUSTERDOWN error, that should address another cause.
For the sanitizer error, I've fixed the one that triggered before, but got another similar error with a function pointer

[err]: Sanitizer error: rax.c:1206:50: runtime error: call to function listRelease through pointer to incorrect function type 'void (*)(void *)'

[zuiderkwast]

@vitarb For the last commits you added to fix sanitizer error and uninitialized variable, are there no PRs in unstable for these?

If we add custom commits in the 8.0 branch, it can become harder to backport more commits to 8.0 later, I guess. WDYT?

IMO we could ignore clang-format errors for the same reason. The formatting and code standard is useful for the unstable branch, but for old release branches like 8.0 it's more important to avoid changes.

[zuiderkwast]

TIMEOUTs look like flakes. I've added retries on CLUSTERDOWN error, that should address another cause. For the sanitizer error, I've fixed the one that triggered before, but got another similar error with a function pointer

The errors we see here should be fixed in unstable so there should be some more PRs we can cherry-pick to fix these? If not, I don't understand why we get the errors here and not in unstable.

[hpatro]

TIMEOUTs look like flakes. I've added retries on CLUSTERDOWN error, that should address another cause. For the sanitizer error, I've fixed the one that triggered before, but got another similar error with a function pointer

The errors we see here should be fixed in unstable so there should be some more PRs we can cherry-pick to fix these? If not, I don't understand why we get the errors here and not in unstable.

Incorrect merge conflict resolution is a potential issue for backport.

[zuiderkwast]

Yes, you're right, merge conflict can introduce many problems.

[hpatro]

This is a trick we used to do internally, we push out the first commit with the conflict and then the next commit with the resolved one. Helped reviewers understand what was accepted. Maybe we can go with the same approach here. However, that's feasible if we do independent PR per backport PR.

[vitarb]

TIMEOUTs look like flakes. I've added retries on CLUSTERDOWN error, that should address another cause. For the sanitizer error, I've fixed the one that triggered before, but got another similar error with a function pointer

The errors we see here should be fixed in unstable so there should be some more PRs we can cherry-pick to fix these? If not, I don't understand why we get the errors here and not in unstable.

Incorrect merge conflict resolution is a potential issue for backport.

Do you see anything incorrect or just making an assumption?

TIMEOUTs look like flakes. I've added retries on CLUSTERDOWN error, that should address another cause. For the sanitizer error, I've fixed the one that triggered before, but got another similar error with a function pointer

The errors we see here should be fixed in unstable so there should be some more PRs we can cherry-pick to fix these? If not, I don't understand why we get the errors here and not in unstable.

This is my first time doing this, I'm seeking advice. Do we want to have a clean pass of all builds or is it okay to ignore some failures given we mostly know what they are? Should I revert custom commits and just keep backports?

[hpatro]

The errors we see here should be fixed in unstable so there should be some more PRs we can cherry-pick to fix these? If not, I don't understand why we get the errors here and not in unstable.

Incorrect merge conflict resolution is a potential issue for backport.

Do you see anything incorrect or just making an assumption?

Assumption hence mentioned potential. Very difficult to spot them without seeing the conflict(s).

[vitarb]

I'll try to fix these test failures as much as possible to get a clean pass — then it should be easier to decide whether to land or forego those changes.

---

🔍 Current Summary of Failures

🧪 Platform-Specific Test Failures

• 14_test-ubuntu-32bit.txt
• 35_reply-schemas-validator.txt
• 39_test-fedorarawhide-jemalloc.txt

2025-06-20T21:50:49.0029560Z [TIMEOUT]: clients state report follows.
2025-06-20T21:50:49.0030367Z sock55a50c439cb0 => (IN PROGRESS) Active defrag big list: standalone
2025-06-20T21:50:49.0179048Z Killing still running Valkey server

---

🧼 Valgrind Tests

• 15_test-valgrind-test.txt
• 25_test-valgrind-no-malloc-usable-size-test.txt

2025-06-21T00:01:33.7857660Z *** [err]: Multiple primary nodes are down, rank them based on the failed primary in tests/unit/cluster/failover2.tcl
2025-06-21T00:01:33.7858132Z No failover detected

---

🍏 macOS Test

• 34_test-macos-latest.txt

2025-06-20T22:25:04.4074660Z *** [err]: Half-finish importing in tests/unit/cluster/half-migrated-slot.tcl
2025-06-20T22:25:04.4078200Z Expected 'xyz' to be equal to ''
  (context: type source line 88 file /Users/runner/work/valkey/valkey/tests/unit/cluster/half-migrated-slot.tcl 
   cmd {assert_equal "xyz" [$cluster get aga]} proc ::test)

---

🧯 Sanitizer Builds

• 36_test-sanitizer-undefined (gcc).txt
  Likely no-op / cancelled build

2025-06-20T21:18:40.1048075Z ##[error]The operation was canceled.

• 37_test-sanitizer-undefined (clang).txt

2025-06-20T21:18:28.5072653Z SUMMARY: UndefinedBehaviorSanitizer: undefined-behavior rax.c:1206:50 
2025-06-20T21:18:28.5124428Z Logged sanitizer errors (pid 3787):
2025-06-20T21:18:28.5126608Z rax.c:1206:50: runtime error: call to function listRelease through pointer to incorrect function type 'void (*)(void *)'
2025-06-20T21:18:28.5128986Z /home/runner/work/valkey/valkey/src/adlist.c:75:10: note: listRelease defined here

---

Will continue narrowing these down.

[vitarb]

Backported couple more commits that seemed relevant for the failures that we are facing. Let's see if that helps.

[zuiderkwast]

Tests look pretty green now. A few flaky tests failed occationally. I'm going to ignore them and release as is.

Failed jobs

test-ubuntu-tls:

*** [err]: evict clients only until below limit in tests/unit/client-eviction.tcl
Expected 4 == [expr 10 / 2] (context: type eval line 53 cmd {assert {$connected_clients == [expr $client_count / 2]}} proc ::test)

test-almalinux9-tls-module:

*** [err]: Replica can update the config epoch when trigger the failover - automatic in tests/unit/cluster/failover2.tcl
6428
Expected 'STILL 3' to match 'BUMPED *' (context: type eval line 18 cmd {assert_match {BUMPED *} $res} proc ::test) 

macos-latest: (uses TCL 8.5, while all other platforms have 8.6 or later?)

bad argument "": should be "nonewline"
    while executing
"read $fd $bytes"
    (procedure "read_from_test_client" line 3)
    invoked from within
"read_from_test_client sock6"
Executing test client: error flushing "sock7": broken pipe.
error flushing "sock7": broken pipe
    while executing
"flush $fd"
    (procedure "send_data_packet" line 5)
    invoked from within
"send_data_packet $::test_server_fd server-killing $pid"
    (procedure "kill_server" line 53)
    invoked from within
"kill_server $srv"
    (procedure "start_server" line 269)
    invoked from within
"start_server {tags {"obuf-limits external:skip logreqres:skip"}} {
    test {CONFIG SET client-output-buffer-limit} {
        set oldval [lindex [r co..."
    (file "tests/unit/obuf-limits.tcl" line 1)
    invoked from within
"source $path"
    (procedure "execute_test_file" line 4)
    invoked from within
"execute_test_file $data"
    (procedure "test_client_main" line 10)
    invoked from within
"test_client_main $::test_server_port "
[TIMEOUT]: clients state report follows.
sock6 => (IN PROGRESS) Client output buffer hard limit is enforced
Killing still running Valkey server 10476

code-coverage:

(some crash...)