fix(replicated_state): truncate canister log records when log capacity shrinks

[mraszyk]

The "a single record fits the ring buffer" invariant was established only when a record was created: CanisterLog::add_record truncates the content against the log memory store's byte capacity at that moment. Nothing re-established it when the capacity was later lowered, so RingBuffer::append_log hit

debug_assert!(false, "Log record size exceeds ring buffer capacity");
continue;

which aborts every debug-assertions build (cargo test, bazel test, a debug replica) and silently drops the record otherwise.

Both outcomes are reachable from ordinary, unprivileged update_settings traffic that lowers log_memory_limit: LogMemoryStore::resize_impl migrates the already stored records into a newly allocated, smaller ring buffer, where a record created against the old capacity no longer fits.

LogRecord::truncate_to_capacity now re-establishes the invariant at append time, applying the same policy add_record applies at creation time — truncating the record rather than dropping it. This makes the capacity check unconditionally satisfied for both callers of append_log (resize_impl and append_delta_log), so the assertion and its silent-drop fallback are removed.

A zero log_memory_limit deallocates the buffer instead of creating a zero-capacity one, and every live buffer has data_capacity of at least DATA_CAPACITY_MIN, so a truncated record always fits.

The delta log of an in-flight execution cannot carry a stale, larger capacity into append_delta_log: update_settings never runs on a canister with a paused execution (can_execute_subnet_msg returns false for ExecutionTask::PausedExecution), and an aborted execution re-executes from scratch against the new capacity.

The regression test in canister_logging.rs stores a record against a 64 KiB log memory limit and then lowers the limit below the size of that single record; it fails on the assertion above without this fix.

🤖 Generated with Claude Code

[Copilot]

Pull request overview

Fixes canister log migration when reducing ring-buffer capacity by truncating oversized records instead of asserting or dropping them.

Changes:

• Adds capacity-aware log-record truncation.
• Applies truncation during ring-buffer appends.
• Adds an end-to-end resize regression test.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
ring_buffer.rs	Truncates records before appending.
log_record.rs	Implements serialized-record truncation.
canister_logging.rs	Tests shrinking below one record’s size.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[zeropath-ai]

✅ No security or compliance issues detected. Reviewed everything up to e2630d0.

Security Overview

• 🔎 Scanned files: 3 changed file(s)
• 🔗 Scan Link: https://zeropath.com/app/repositories/f5e9d2bb-29a4-4b64-902a-391d65fd8b2d?scanId=60dbcdc7-4395-4623-b671-b9503a196683&codeScanTypes=PrScan&tab=issues

Detected Code Changes

Change Type

Relevant files

Enhancement

► rs/execution_environment/tests/canister_logging.rs     test_canister_resize_down_below_record_size_truncates_logs() ► rs/replicated_state/src/canister_state/system_state/log_memory_store/log_record.rs     Add truncate_to_capacity method and documentation ► rs/replicated_state/src/canister_state/system_state/log_memory_store/ring_buffer.rs     Invoke truncate_to_capacity during log processing to re-establish capacity constraints