…labels

- cbm_progress_sink_fn() parses msg= tag from structured log lines
- Maps pipeline.discover, pipeline.route, pass.start, pass.timing (9 passes),
  pipeline.done, parallel.extract.progress to human-readable stderr output
- parallel.extract.progress uses \r for in-place terminal updates
- Unknown tags pass through to previous sink (MCP UI routing preserved)
- cbm_progress_sink_init/fini save and restore previous sink

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…cancel

- Scan argv for --progress before tool dispatch; strip it and shift args
- Add g_cli_pipeline global and cli_sigint_handler (calls cbm_pipeline_cancel)
- When --progress: call cbm_progress_sink_init(stderr) and register SIGINT handler
- For index_repository + --progress: bypass cbm_mcp_handle_tool, call
  cbm_pipeline_new/cbm_pipeline_run directly, set g_cli_pipeline before run
- Assemble JSON result (project/status/nodes/edges) via snprintf, print to stdout
- After run, call cbm_progress_sink_fini(); all progress output goes to stderr

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

- CLI_SRCS now includes src/cli/progress_sink.c alongside src/cli/cli.c
- Build verified clean: build/c/codebase-memory-mcp produced with no warnings
- All 2042 tests pass with no regressions

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

- Add test_cli_progress_stderr_labels: injects pipeline.discover log event,
  asserts progress sink writes "Discovering" to target FILE*
- Add test_cli_progress_stdout_json: injects pass.start + pipeline.done events,
  asserts "[1/9]" phase label and "Done:" appear; confirms output is not JSON
- Include <cli/progress_sink.h> and <foundation/log.h> headers
- Register both tests in SUITE(cli) under group G

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Normalize alignment whitespace and line-continuation style per project
clang-format configuration.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

When a custom log sink is registered via cbm_log_set_sink(), suppress the
default fprintf(stderr, ...) output in both cbm_log() and cbm_log_int().
The sink is now the sole output handler rather than an additive listener.

Also pre-scan for --progress in main() before cbm_mem_init() so the sink is
installed before mem.init fires, keeping stderr completely clean.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

- volatile on g_cli_pipeline so signal handler always observes the pointer
- volatile on s_needs_newline to prevent stale-read between worker/main threads
- Fix incorrect PIPE_BUF thread-safety comment (correct reason: per-FILE* locking)
- Add comment documenting --progress silent-ignore for non-index_repository tools
- Rename test_cli_progress_stdout_json → test_cli_progress_phase_labels
- Add test_cli_progress_parallel_extract: exercises \r path + pass.timing flush
- Add test_cli_progress_unknown_tag: verifies unknown events are silently dropped

2046 tests pass.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…utput

Two display bugs found during manual testing against a large repo:

1. Phases 6 and 7 were swapped: HTTP links fires before git history in the
   actual pipeline execution order. Swap their phase numbers in the sink.

2. "Done: 0 nodes" was shown because cbm_gbuf_dump_to_sqlite() frees
   node_by_qn before pipeline.done is logged, making cbm_gbuf_node_count()
   return 0. Fix: capture node/edge counts from the gbuf.dump event (which
   fires with the real counts before the hash table is freed) and use them
   for the Done: display line.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Replace the pre-scan cbm_progress_sink_init(stderr) in main() with a
temporary log level raise to WARN around cbm_mem_init(). This suppresses
the mem.init log line without installing the sink twice — run_cli()
remains the sole owner of the progress sink lifecycle.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

- Add static helper verify_project_indexed() before handle_get_graph_schema
- Replace inline guard in handle_get_architecture with helper call
- Apply guard to handle_search_graph, handle_get_graph_schema,
  handle_trace_call_path, handle_get_code_snippet, handle_query_graph
- All five query handlers now return {"error":"project not indexed —
  run index_repository first"} instead of silently returning empty results

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

- Add cbm_store_open_path_query() that opens with SQLITE_OPEN_READWRITE
  only (no SQLITE_OPEN_CREATE); returns NULL when file is absent
- Declare cbm_store_open_path_query() in store.h
- Change resolve_store() in mcp.c to call cbm_store_open_path_query
  so querying a nonexistent project never creates a ghost .db file
- Indexing path (cbm_store_open_path) retains SQLITE_OPEN_CREATE

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Asserts that query handlers return a guard error for unknown projects
and that no ghost .db file is created in the cache directory.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

- Update stale comment in verify_project_indexed: resolve_store now uses
  cbm_store_open_path_query (no SQLITE_OPEN_CREATE), so store is NULL for
  missing files; the helper catches the empty-but-present .db case
- Guard state updates in resolve_store behind successful open check: only
  set owns_store=true and update current_project when store is non-NULL,
  preventing misleading state when an unknown project is queried
- Expand smoke_guard.sh to test all 5 guarded handlers (search_graph,
  query_graph, get_graph_schema, trace_call_path, get_code_snippet) instead
  of search_graph only; each checks both the guard error and no-ghost-file
  invariant

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

- Fix handle_get_code_snippet inline store-NULL check to return the same
  JSON error format as REQUIRE_STORE and the other inline checks:
  {"error":"no project loaded"} instead of a plain string
- Fix smoke_guard.sh query_graph invocation: pass "query" parameter
  (not "cypher") to match what handle_query_graph actually reads; the
  wrong key caused the handler to early-return before reaching the guard
- Remove extra blank line in handle_get_graph_schema guard block

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

yyjson_mut_obj_add_str stores the string pointer without copying it.
The ADR read path allocated buf, added it to the yyjson doc, then
immediately freed buf — leaving the doc with a dangling pointer. When
yy_doc_to_str serialized the doc, it read freed memory, producing
garbage bytes. cbm_jsonrpc_format_response then called yyjson_read on
the corrupted JSON, which failed silently, so no "result" field was
emitted and the MCP client hung waiting for a valid response.

Fix: hoist adr_buf to function scope, initialized to NULL, and free it
after yy_doc_to_str has serialized the document.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>