MOD-15749 Short-form CLUSTERSET: reject on non-cluster (standalone) shard (RED-207632)

[gabsow]

Summary

Fixes a NULL-pointer SIGSEGV in the LibMR event-loop thread (timeseries-el) when a short-form CLUSTERSET reaches a non-cluster (standalone) shard — crash reported in RED-207632 (test_asm_happyflow_flag_toggle, ASM nightly).

Root cause

SetClusterDataShortForm builds the topology entirely from the cluster API. It calls InitClusterData → SetMyId, which does:

const char* myId = RedisModule_GetMyClusterID();   // NULL when cluster mode is off
...
memcpy(cluster->myId + zerosPadding, myId, myIdLen); // memcpy(dst, NULL, 40) -> SIGSEGV

On a standalone shard RedisModule_GetMyClusterID() returns NULL, so this is memcpy(dst, NULL, REDISMODULE_NODE_ID_LEN). It runs in the event-loop thread because MR_ClusterSet schedules MR_ClusterSetFromCommand via MR_EventLoopAddTask. This matches the crash exactly (faulting thread timeseries-el, RSI=0, RDX=0x28 = 40 = REDISMODULE_NODE_ID_LEN).

The long form is unaffected — it takes myId from the command arguments (argv[CLUSTERSET_MYID_LONG_FORM_INDEX]), never from GetMyClusterID().

Fix

Guard the short-form path with the same REDISMODULE_CTX_FLAGS_CLUSTER check MR_RefreshClusterData already uses: if the shard is not in cluster mode, log a warning and return REDISMODULE_ERR (the same failure contract as the sibling "slot-ranges API missing" rejection right above it) instead of crashing. The shard stays unconfigured until cluster mode is enabled or a long-form CLUSTERSET / REFRESHCLUSTER arrives.

Verification

• make -C src/ compiles cleanly with the change.
• Path analysis confirms only the short-form + standalone case is affected; long-form and clustered short-form paths are unchanged.

---

Note

Low Risk
Small defensive guard on an error path already used elsewhere; long-form and clustered short-form behavior unchanged.

Overview
Fixes a SIGSEGV in the LibMR event-loop thread when short-form CLUSTERSET is handled on a standalone (non-cluster) Redis instance.

Short-form topology setup calls SetMyId, which uses RedisModule_GetMyClusterID(); that API returns NULL without cluster mode, leading to a NULL memcpy and crash. SetClusterDataShortForm now checks REDISMODULE_CTX_FLAGS_CLUSTER (same pattern as MR_RefreshClusterData): if cluster mode is off, it logs a warning and returns REDISMODULE_ERR, leaving the shard unconfigured until cluster mode is on or a long-form CLUSTERSET / REFRESHCLUSTER succeeds.

Long-form CLUSTERSET and short-form on clustered shards are unchanged.

^{Reviewed by Cursor Bugbot for commit 3020b2e. Bugbot is set up for automated code reviews on this repo. Configure here.}

[gabsow]

Superseded by #103, which combines this with the sibling short-form CLUSTERSET fail-safe into a single PR (one review + one pin bump). Closing in favour of #103.