coverage

Author: eldiablolives

Makefile

@@ -245,6 +245,7 @@ test-validation: test-zem-diag-ret-trunc-ld32s64
 test-validation: test-zem-coverage-smoke
 test-validation: test-zem-coverage-blackholes
 test-validation: test-zem-rep-scan
+test-validation: test-zem-rep-render-from-report
 test-validation: test-zem-strip-uncovered-ret
 test-validation: test-zem-strip-uncovered-delete
 
@@ -294,6 +295,9 @@ test-zem-coverage-blackholes: zas zem
 test-zem-rep-scan: zem
 	sh test/zem_rep_scan.sh
 
+test-zem-rep-render-from-report:
+	sh test/zem_rep_render_from_report.sh
+
 test-zem-caps: zem
 	sh test/zem_caps.sh
 

src/zem/DEBLOAT.md

@@ -137,14 +137,24 @@ The coverage report is line-delimited JSON (JSONL). Record keys:
 
 - `k == "zem_rep_cov"` (optional repetition/coverage enrichment)
   - `v`: schema version (currently `1`)
+  - `module_hash`: from coverage summary (string)
   - `total_instr`: from coverage summary
   - `covered_instr`: from coverage summary
+  - `total_labels`: number of labels seen in coverage
   - `blackhole_labels`: number of labels with uncovered instructions
 
+- `k == "zem_rep_blackhole"` (optional top blackhole labels; emitted when `--rep-max-report N` is non-zero)
+  - `v`: schema version (currently `1`)
+  - `label`: label name (string)
+  - `uncovered_instr`: uncovered instruction records under this label
+  - `covered_instr`: covered instruction records under this label
+  - `total_instr`: total instruction records under this label
+  - `first_pc`: first IR record index where this label appears
+
 Notes:
 
 - This is a C implementation intended to eventually replace the Python repetition scan for CI/one-binary workflows.
-- The detailed “top repeated n-grams” report (`zem_rep_ngram`) is not emitted yet.
+- The detailed “top repeated n-grams” report (`k == "zem_rep_ngram"`) is emitted when `--rep-max-report N` is non-zero.
 
 ### Requirements
 

src/zem/main.c

@@ -106,6 +106,7 @@ static void print_help(FILE *out) {
       "  --rep-n N          N-gram length (e.g. 8)\n",
       "  --rep-mode MODE    MODE: exact | shape\n",
       "  --rep-out PATH     Write repetition report JSONL to PATH (or '-' for stdout)\n",
+      "  --rep-max-report N Emit up to N zem_rep_ngram records (top repeated n-grams)\n",
       "  --rep-coverage-jsonl PATH Optional coverage JSONL to enrich bloat score\n",
       "  --rep-diag         Print one-line bloat_diag summary to stdout\n",
       "  --                Stop option parsing; remaining args become guest argv\n",
@@ -224,6 +225,7 @@ int main(int argc, char **argv) {
   const char *rep_mode = "shape";
   const char *rep_out = NULL;
   const char *rep_cov = NULL;
+  int rep_max_report = 0;
 
   zem_proc_t proc;
   memset(&proc, 0, sizeof(proc));
@@ -321,6 +323,16 @@ int main(int argc, char **argv) {
       rep_out = argv[++i];
       continue;
     }
+    if (strcmp(argv[i], "--rep-max-report") == 0) {
+      if (i + 1 >= argc) {
+        return zem_failf("--rep-max-report requires a number");
+      }
+      char *end = NULL;
+      long v = strtol(argv[++i], &end, 10);
+      if (!end || end == argv[i]) return zem_failf("bad --rep-max-report value");
+      rep_max_report = (int)v;
+      continue;
+    }
     if (strcmp(argv[i], "--rep-coverage-jsonl") == 0) {
       if (i + 1 >= argc) {
         return zem_failf("--rep-coverage-jsonl requires a path");
@@ -621,7 +633,7 @@ int main(int argc, char **argv) {
       return zem_failf("cannot read program IR from stdin while --debug-script - uses stdin");
     }
     return zem_rep_scan_program(inputs, ninputs, rep_n, rep_mode, rep_cov, rep_out,
-                                rep_diag);
+                                rep_max_report, rep_diag);
   }
 
   if (!dbg.debug_events_only) {