Repository navigation
Audit and DB disagree on what a function is: 48,318 vs 45,550 on the crucible (slicer pseudo-units like CREATE_Statement in the audit only) #4110
Description
Problem
The audit and the scan DB disagree about which units are functions.
On the language-crucible scan, the audit's per-file 5. Function Analysis lists 48,318 functions; the DB's function_data has 45,550. Synthetic slices don't explain the gap: both sides carry __global_context__.
The difference is concentrated in 555 files, mostly shell (252), lua (101), livecode (98), sqlite (72), db2_sql, sql, fortran and ruby. The audit includes slicer pseudo-units that the DB filters out, e.g.:
| File | Audit | DB | Extra in audit |
|---|---|---|---|
sqlite/mediawiki_sqlite_tables/tables-generated.sql |
198 | 0 | CREATE_Statement ×198 |
sqlite/yii2_sqlite_schema/tests_data_sqlite.sql |
130 | 0 | INSERT_Statement ×69, CREATE_Statement ×31, DROP_Statement ×30 |
db2_sql/db2unit/06-AssertsNoMessage.sql |
109 | 0 | Declarative_Block ×108 |
matlab/eeglab/eeglab.m |
74 | 13 | Anonymous_Block ×60 |
301 files with functions in the audit have none in the DB.
Why it matters
- The golden master tracks the audit's population, while archetypes, the function mix and the DB use the other. Function counts in reports and in the DB don't agree.
- Per-function statistics in the audit (impact, documented %, etc.) may be averaged over units that aren't functions (the functions_found counts the slicer's synthetic __global_context__ buckets, inflating every per-function descriptor for 5 languages #2691 class of bug).
To decide
Which population is intended? Trace where record_keeper filters (calls_only, is_synthetic_slice, or a name-based filter) versus audit_recorder. Then make both outputs use one shared predicate, or label pseudo-units explicitly in the audit. Expect a golden re-bless.
Activity
fix(#4110): one shared definition of "a function" for counting and li…