cachedb_memcached, cachedb_cassandra: fix NULL deref in connection init

[dondetir]

Summary

Fix NULL pointer dereferences in cachedb_memcached and cachedb_cassandra connection initialization, following the same pattern fixed in cachedb_redis (8fb569c).

Details

• Type: Bug fix
• Motivation: Commit 8fb569c fixed a NULL dereference in cachedb_redis where redisConnect() could return NULL on allocation failure but the return was used without a NULL check. A systematic audit of all other cachedb backends revealed the same pattern in two more modules:

1. cachedb_memcached — memcached_create(NULL) can return NULL on allocation failure (documented behavior). No NULL check exists before passing the pointer to memcached_server_push(), causing a crash.

2. cachedb_cassandra — cass_cluster_new() can return NULL on allocation failure. A NULL check exists, but it comes after cass_cluster_set_credentials() already dereferences the pointer (when credentials are configured), so the check is too late.

• Affected scenarios: Both bugs trigger on memory allocation failure during connection initialization. While OOM during init is rare, the redis fix addressed the identical case and was merged.
• Generic problem — not specific to any particular UAC or scenario.

Solution

• memcached: Added NULL check immediately after memcached_create(), following the existing error pattern in the function (pkg_free(con); return 0).
• cassandra: Moved the existing NULL check to immediately after cass_cluster_new(), before any use of the pointer. No new code — just reordered.
• Remaining concerns: None. Both fixes are minimal and follow established patterns.

Compatibility

• No backward compatibility issues
• No migration needed
• No SIP interoperability impact
• These are purely defensive NULL checks on error paths that previously would crash

Closing Issues

N/A — no open issue for these. Found via code audit following 8fb569c.