fix: gcov Windows本地化 + runner合并 + 决策覆盖率修复

- gcov从WSL 15.2改为Windows MinGW 9.2.0，消除.gcno格式不匹配
- coverage.py去除leaf_stats复位bug（决策覆盖率恢复）
- runner.py合并V3编译+SOURCE分组执行+出力保存
- __init__.py gcov条件/outdir解析修复

cobol_testgen/__init__.py

@@ -288,6 +288,9 @@ def main():
     if '--gcov' in args:
         gcov_mode = True
         args.remove('--gcov')
+        if not _HAVE_RUNNER:
+            logger.warning("--gcov: runner.py not found. Compile/run will be skipped. "
+                           "Use --gcov without runner only generates test data + static coverage.")
     i = 0
     while i < len(args):
         if args[i] == '--temp-dir':
@@ -309,7 +312,7 @@ def main():
     outdir = None
     for a in args:
         p = Path(a)
-        if p.is_dir():
+        if p.is_dir() or (not p.suffix and p.parent.exists()):
             outdir = p
         elif p.suffix.upper() in ('.CBL', '.COB', '.CPY'):
             cobol_files.append(p)
@@ -553,7 +556,7 @@ def main():
                            data_fields=fields_dict, select_info=select_info)
 
         gcov_data = None
-        if gcov_mode and proc_div and _HAVE_GCOV:
+        if gcov_mode and proc_div and _HAVE_GCOV and _HAVE_RUNNER:
             _temp = temp_dir or str(outdir / '.gcov_cache')
             source_dir = str(filepath.parent)
             expected_records: list[dict] = [{}] * len(records)

cobol_testgen/core.py

@@ -1402,15 +1402,22 @@ def propagate_assignments(rec, assignments, fields, file_sec=None):
         if pi.get('type') == 'numeric':
             digits = pi.get('digits', 0)
             decimal = pi.get('decimal', 0)
-            total = digits + decimal
             s = str(val)
             neg = s.startswith('-')
             if neg:
                 s = s[1:]
+            # Handle explicit decimal point (e.g. "008." or "12.34")
+            if '.' in s:
+                parts = s.split('.')
+                int_part = parts[0]
+                dec_part = parts[1] if len(parts) > 1 else ''
+            else:
+                total = digits + decimal
                 s = s.zfill(total)
                 int_part = s[:digits] if digits else '0'
-            dec_part = s[digits:] if decimal > 0 else '0'
-            result = float(int(int_part or '0') + int(dec_part or '0') / (10 ** decimal))
+                dec_part = s[digits:] if decimal > 0 else ''
+            dec_val = int(dec_part.ljust(decimal, '0')[:decimal] or '0') if decimal > 0 else 0
+            result = float(int(int_part or '0') + dec_val / (10 ** decimal))
             return -result if neg else result
         try:
             return float(val)

cobol_testgen/coverage.py

@@ -1199,9 +1199,7 @@ def run_coverage(branch_tree, branch_paths_with_assigns, fields,
 
     if gcov_data:
         mark_from_gcov(decision_points, gcov_data, branch_tree)
-        for ls in leaf_stats:
-            ls.covered_true = False
-            ls.covered_false = False
+        # leaf_stats 保留静态分析结果（gcov 无 -b 时不提供叶条件级别的分支数据）
 
     _source_note = ''
     if gcov_data:

cobol_testgen/gcov.py

@@ -18,7 +18,7 @@ def parse_cbl_gcov(gcov_path: str) -> dict[int, int]:
       -:     17: 源码行   → 不可执行（注释/声明行，跳过）
     """
     counts = {}
-    with open(gcov_path, encoding='utf-8') as f:
+    with open(gcov_path, encoding='utf-8', errors='replace') as f:
         for line in f:
             m = re.match(r'^\s*(#####|\d+\*?|-):\s*(\d+):', line)
             if not m:
@@ -35,7 +35,10 @@ def parse_cbl_gcov(gcov_path: str) -> dict[int, int]:
 
 
 def run_gcov(program_name: str, work_dir: str) -> dict[int, int]:
-    """在 work_dir 中通过 WSL 执行 gcov 并解析 COBOL 行计数。
+    """在 work_dir 中执行 gcov 并解析 COBOL 行计数。
+
+    使用 Windows 本地的 gcov（与 GnuCOBOL 内置 MinGW 同版本）。
+    不要通过 WSL 调用 gcov，否则 .gcno/.gcda 版本不匹配。
 
     Args:
         program_name: 程序名（不含扩展名），如 "ALLCMDS"
@@ -44,21 +47,26 @@ def run_gcov(program_name: str, work_dir: str) -> dict[int, int]:
     Returns:
         {COBOL行号: 执行次数} 字典。失败时返回空 dict。
     """
-    wsl_work = _wsl_path(work_dir)
-    cmd = ['wsl', 'sh', '-c', f'cd {wsl_work} && gcov {program_name}.c']
+    try:
         result = subprocess.run(
-        cmd,
+            ['gcov', f'{program_name}.cbl'],
             capture_output=True, text=True,
             encoding='utf-8', errors='replace',
             timeout=30,
+            cwd=work_dir,
         )
+    except FileNotFoundError:
+        logger.warning("gcov 命令未找到（未安装 MinGW gcov 或不在 PATH 中）")
+        return {}
     if result.returncode != 0:
         logger.warning(f"gcov 失败 (exit={result.returncode}): {result.stderr.strip()}")
         return {}
 
     cbl_gcov = Path(work_dir) / f'{program_name}.cbl.gcov'
     if not cbl_gcov.exists():
-        logger.warning(f"gcov 输出不存在: {cbl_gcov}")
+        cbl_gcov = Path(work_dir) / f'{program_name}.gcov'
+    if not cbl_gcov.exists():
+        logger.warning(f"gcov 输出不存在 (tried .cbl.gcov / .gcov)")
         return {}
 
     gcov_data = parse_cbl_gcov(str(cbl_gcov))

cobol_testgen/runner.py

@@ -0,0 +1,504 @@
+"""非DB COBOL程序的编译·执行·验证层（被 __init__.py 的 --gcov/--run 调用）
+
+V3 编译方式（Windows 本地 cobc）+ SOURCE 分组执行+字段对比+出力保存。
+"""
+
+import logging
+import os
+import re
+import shutil
+import subprocess
+from dataclasses import dataclass, field
+from pathlib import Path
+
+from . import file_io
+
+logger = logging.getLogger(__name__)
+
+
+# ── 数据模型 ──
+
+
+@dataclass
+class GroupInfo:
+    """一组执行用例（SOURCE 版）"""
+    name: str
+    records: list[dict]
+    expected_outputs: list[dict]
+    expected_returncode: int = 0
+    fd_field_dicts: dict = field(default_factory=dict)
+    open_dir: dict = field(default_factory=dict)
+    select_info: dict = field(default_factory=dict)
+    overlap_mask: list[bool] = field(default_factory=list)
+    multi_write_fds: set = field(default_factory=set)
+
+
+@dataclass
+class CompareDetail:
+    """单字段对比结果（兼容 __init__.py 的 DetailItem 接口）"""
+    field: str
+    expected: str
+    actual: str
+    match: bool
+
+
+@dataclass
+class GroupResult:
+    """单组执行+对比结果"""
+    name: str
+    returncode: int
+    passed: bool
+    details: list[CompareDetail] = field(default_factory=list)
+    error_message: str = ''
+
+
+# ── 路径解析（V3）──
+
+
+def _resolve_sub_dir(source_dir: str) -> str:
+    p = Path(source_dir).resolve()
+    for d in [p.parent / 'sub', p / 'sub']:
+        if d.is_dir():
+            return str(d)
+    return str(p.parent / 'sub')
+
+
+def _resolve_cpy_dir(source_dir: str) -> str:
+    p = Path(source_dir).resolve()
+    for d in [p.parent / 'cpy', p / 'cpy']:
+        if d.is_dir():
+            return str(d)
+    return str(p.parent / 'cpy')
+
+
+def _input_assign_names(select_info: dict, open_dir: dict,
+                        fd_fields: dict) -> list[str]:
+    names = []
+    seen = set()
+    for fd_name in fd_fields:
+        direction = (open_dir or {}).get(fd_name, '')
+        if direction not in ('INPUT', 'I-O'):
+            continue
+        assign = select_info.get(fd_name, {}).get('assign', '')
+        if assign and assign not in seen:
+            seen.add(assign)
+            names.append(assign)
+    return names
+
+
+def _output_assign_names(select_info: dict, open_dir: dict,
+                         fd_fields: dict) -> list[str]:
+    names = []
+    seen = set()
+    for fd_name in fd_fields:
+        direction = (open_dir or {}).get(fd_name, '')
+        if direction not in ('OUTPUT', 'I-O'):
+            continue
+        assign = select_info.get(fd_name, {}).get('assign', '')
+        if assign and assign not in seen:
+            seen.add(assign)
+            names.append(assign)
+    return names
+
+
+# ── SUB 模块编译（V3）──
+
+
+def compile_sub_modules(sub_dir: str, work_dir: str,
+                        cpy_dir: str | None = None) -> list[str]:
+    sub_path = Path(sub_dir)
+    work_path = Path(work_dir)
+    work_path.mkdir(parents=True, exist_ok=True)
+    o_files = []
+
+    if not sub_path.is_dir():
+        logger.warning(f"  SUB目录不存在: {sub_dir}")
+        return o_files
+
+    for cbl in sorted(sub_path.glob('*.cbl')):
+        o_path = work_path / f'{cbl.stem}.o'
+        gcno_path = work_path / f'{cbl.stem}.gcno'
+        if o_path.exists() and gcno_path.exists():
+            o_files.append(str(o_path))
+            continue
+
+        cmd = ['cobc', '-c', '-g', '--coverage', '-o', str(o_path)]
+        if cpy_dir:
+            cmd.extend(['-I', cpy_dir])
+        cmd.append(str(cbl))
+
+        logger.info(f"  SUB: {cbl.name}")
+        orig = os.getcwd()
+        try:
+            os.chdir(str(work_path))
+            r = subprocess.run(cmd, capture_output=True, text=True, timeout=120)
+        finally:
+            os.chdir(orig)
+        if r.returncode != 0:
+            logger.warning(f"  SUB编译失败 {cbl.name}: {r.stderr.strip()[:200]}")
+            continue
+        o_files.append(str(o_path))
+
+    return o_files
+
+
+def compile_program(program_name: str, source_dir: str, work_dir: str,
+                    sub_objects: list[str],
+                    cpy_dir: str | None = None) -> str:
+    work_path = Path(work_dir)
+    work_path.mkdir(parents=True, exist_ok=True)
+
+    src = Path(source_dir) / f'{program_name}.cbl'
+    if not src.exists():
+        raise FileNotFoundError(f"源文件不存在: {src}")
+
+    exe = work_path / f'{program_name}.exe'
+
+    cmd = ['cobc', '-x', '-g', '--coverage', '-o', str(exe)]
+    if cpy_dir:
+        cmd.extend(['-I', cpy_dir])
+    cmd.append(str(src))
+    cmd.extend(sub_objects)
+
+    logger.info(f"  LINK: {program_name}.exe")
+    orig = os.getcwd()
+    try:
+        os.chdir(str(work_path))
+        r = subprocess.run(cmd, capture_output=True, text=True, timeout=120)
+    finally:
+        os.chdir(orig)
+    if r.returncode != 0:
+        raise RuntimeError(f"编译失败 {program_name}: {r.stderr.strip()[:500]}")
+    return str(exe)
+
+
+# ── 目录管理（SOURCE）──
+
+
+def _clean_gcda(temp_dir: str):
+    for f in Path(temp_dir).glob('*.gcda'):
+        try:
+            f.unlink()
+        except OSError:
+            pass
+
+
+# ── 工具函数（SOURCE）──
+
+
+def _build_fd_field_dicts(fd_fields: dict, fields_dict: list) -> dict:
+    name_map = {f['name']: f for f in fields_dict}
+    result = {}
+    for fd_name, names in fd_fields.items():
+        result[fd_name] = []
+        for n in names:
+            if n in name_map:
+                pi = name_map[n].get('pic_info', {})
+                if pi.get('type') == 'unknown':
+                    continue
+                result[fd_name].append(name_map[n])
+    return result
+
+
+def _name_to_field(fields_dict: list[dict]) -> dict:
+    return {f['name']: f for f in fields_dict
+            if not f.get('is_88') and not f.get('is_filler')}
+
+
+# ── 读取出力（SOURCE）──
+
+
+def read_outputs(fd_field_dicts: dict, open_dir: dict,
+                 select_info: dict, temp_dir: str) -> dict[str, list[dict]]:
+    result = {}
+    for fd_name, fds in fd_field_dicts.items():
+        direction = open_dir.get(fd_name, '')
+        if direction not in ('OUTPUT', 'I-O'):
+            continue
+        sel = select_info.get(fd_name, {})
+        if isinstance(sel, dict):
+            assign = sel.get('assign', fd_name)
+            org = sel.get('organization', 'SEQUENTIAL')
+            rec_mode = sel.get('recording_mode', 'F')
+        else:
+            assign = sel
+            org = 'SEQUENTIAL'
+            rec_mode = 'F'
+        outpath = os.path.join(temp_dir, assign)
+        if not os.path.exists(outpath):
+            logger.warning(f"  出力文件不存在: {outpath}")
+            continue
+        line_seq = (org == 'LINE SEQUENTIAL')
+        try:
+            records = file_io.read_output_file(
+                outpath, fds, line_sequential=line_seq, recording_mode=rec_mode
+            )
+            result[fd_name] = records
+        except Exception as e:
+            logger.warning(f"  读取出力文件失败 {outpath}: {e}")
+            continue
+    return result
+
+
+# ── 对比（SOURCE）──
+
+
+def compare_outputs(actual: list[dict], expected: list[dict],
+                    fd_fields: list[dict],
+                    subset_match: bool = False) -> tuple[bool, list[CompareDetail]]:
+    all_pass = True
+    details = []
+
+    if subset_match:
+        for exp in expected:
+            for fd in fd_fields:
+                fname = fd['name']
+                expected_val = exp.get(fname, '')
+                if not expected_val:
+                    continue
+                found = any(
+                    act.get(fname, '') == expected_val
+                    for act in actual
+                )
+                if not found:
+                    all_pass = False
+                    details.append(CompareDetail(
+                        field=fname,
+                        expected=expected_val,
+                        actual='(not found in actual)',
+                        match=False,
+                    ))
+        return all_pass, details
+
+    if len(actual) != len(expected):
+        logger.warning(f"  compare_outputs: actual={len(actual)} vs expected={len(expected)}")
+    for i, (act, exp) in enumerate(zip(actual, expected)):
+        for fd in fd_fields:
+            fname = fd['name']
+            actual_val = act.get(fname, '')
+            expected_val = exp.get(fname, '')
+            match = (actual_val == expected_val)
+            if not match:
+                all_pass = False
+            details.append(CompareDetail(
+                field=f'{fname}[{i}]',
+                expected=expected_val,
+                actual=actual_val,
+                match=match,
+            ))
+
+    if len(actual) != len(expected):
+        all_pass = False
+        details.append(CompareDetail(
+            field='record_count',
+            expected=str(len(expected)),
+            actual=str(len(actual)),
+            match=False,
+        ))
+
+    return all_pass, details
+
+
+# ── 单组执行（SOURCE 改 - Native）──
+
+
+def run_group(group: GroupInfo, exe_path: str, temp_dir: str) -> GroupResult:
+    logger.info(f"  执行组: {group.name} ({len(group.records)} 条记录)")
+
+    exe = Path(exe_path).resolve()
+    work_dir = Path(temp_dir).resolve()
+
+    orig = os.getcwd()
+    try:
+        os.chdir(str(work_dir))
+        result = subprocess.run(
+            [str(exe)], capture_output=True, text=True,
+            encoding='utf-8', errors='replace', timeout=60,
+        )
+    except subprocess.TimeoutExpired:
+        return GroupResult(name=group.name, returncode=-1, passed=False,
+                           error_message='timeout')
+    finally:
+        os.chdir(orig)
+
+    rc = result.returncode
+    all_pass = (rc == group.expected_returncode)
+    all_details = []
+
+    if rc != group.expected_returncode:
+        all_details.append(CompareDetail(
+            field='returncode',
+            expected=str(group.expected_returncode),
+            actual=str(rc),
+            match=False,
+        ))
+
+    return GroupResult(
+        name=group.name,
+        returncode=rc,
+        passed=all_pass,
+        details=all_details,
+    )
+
+
+# ── 主编排 ──
+
+
+def run_all(program_name: str, outdir: str, temp_dir: str,
+            fields_dict: list[dict], fd_fields: dict,
+            select_info: dict, open_dir: dict,
+            term_types: list[str], records: list[dict],
+            expected_records: list[dict] | None = None,
+            source_dir: str | None = None,
+            path_infos: list | None = None,
+            multi_write_fds: set | None = None
+            ) -> list[GroupResult]:
+    """完整编排：编译 → 准备目录 → 逐组执行 → 出力保存。"""
+    source_dir = source_dir or str(Path(outdir).parent)
+    work_dir = Path(temp_dir)
+    work_dir.mkdir(parents=True, exist_ok=True)
+    expected = expected_records or []
+    path_infos = path_infos or []
+    multi_write_fds = multi_write_fds or set()
+
+    fd_field_dicts = _build_fd_field_dicts(fd_fields, fields_dict)
+
+    # ── 1. SUB 编译（V3）──
+    sub_dir = _resolve_sub_dir(source_dir)
+    cpy_dir = _resolve_cpy_dir(source_dir)
+    sub_o = compile_sub_modules(sub_dir, str(work_dir), cpy_dir)
+
+    # ── 2. 主程序编译（V3）──
+    exe_path = compile_program(
+        program_name, source_dir, str(work_dir), sub_o, cpy_dir
+    )
+
+    # ── 3. 入力ファイル配置（V3）──
+    input_dir = Path(outdir) / 'input'
+    assign_names = _input_assign_names(select_info, open_dir, fd_fields)
+    if input_dir.is_dir():
+        for assign in assign_names:
+            src = input_dir / assign
+            if src.exists():
+                shutil.copy2(str(src), str(work_dir / assign))
+                logger.info(f"  INPUT: {assign} ({src.stat().st_size} bytes)")
+
+    # ── 4. 清理旧 gcda + 执行 ──
+    _clean_gcda(str(work_dir))
+
+    def _is_output_fd(fd_name: str) -> bool:
+        dir_val = open_dir.get(fd_name, '')
+        return dir_val in ('OUTPUT', 'I-O')
+
+    # output_input_files 只写入非 abend 记录，同步过滤 expected
+    filtered_expected = []
+    for i, rec in enumerate(expected):
+        term = term_types[i] if i < len(term_types) else 'normal'
+        if term != 'abend':
+            filtered_expected.append(rec)
+    expected = filtered_expected
+
+    group = GroupInfo(
+        name=program_name,
+        records=records,
+        expected_outputs=expected,
+        expected_returncode=0,
+        fd_field_dicts=fd_field_dicts,
+        open_dir=open_dir,
+        select_info=select_info,
+        multi_write_fds=multi_write_fds,
+    )
+
+    r = run_group(group, exe_path, str(work_dir))
+    results = [r]
+
+    # 出力拷贝到 output_dir
+    output_dir = Path(outdir) / 'output'
+    output_dir.mkdir(parents=True, exist_ok=True)
+    for fd_name in fd_field_dicts:
+        if not _is_output_fd(fd_name):
+            continue
+        sel = select_info.get(fd_name, {})
+        assign = sel.get('assign', fd_name) if isinstance(sel, dict) else fd_name
+        src = os.path.join(str(work_dir), assign)
+        dst = output_dir / assign
+        if os.path.exists(src):
+            shutil.copy2(src, str(dst))
+
+    status = '✓' if r.passed else '✗'
+    logger.info(f"  组 '{group.name}': returncode={r.returncode}, "
+                 f"{len(r.details)} fields, {status}")
+    logger.info(f"  EXIT={r.returncode}, 出力パス={work_dir}")
+    return results
+
+
+# ── run_and_compare（被 --run 调用，SOURCE 兼容）──
+
+
+def run_and_compare(program_name: str, outdir: str,
+                    fields_dict: list[dict], fd_fields: dict,
+                    select_info: dict, open_dir: dict,
+                    term_types: list[str], records: list[dict]) -> dict:
+    """旧版接口兼容包装。返回 {normal_pass, normal_count, ...}。"""
+    fd_field_dicts = _build_fd_field_dicts(fd_fields, fields_dict)
+
+    normal_recs = []
+    abend_recs = []
+    for i, term in enumerate(term_types):
+        if term == 'abend' and i < len(records):
+            abend_recs.append(records[i])
+        elif i < len(records):
+            normal_recs.append(records[i])
+
+    result = {
+        'normal_pass': False, 'normal_count': 0,
+        'abend_pass': 0, 'abend_total': len(abend_recs),
+        'output_summary': {},
+    }
+
+    temp_dir = os.path.join(outdir, '.run_cache')
+    source_dir = os.path.join(outdir, '..', 'input')
+
+    work_dir = Path(temp_dir)
+    work_dir.mkdir(parents=True, exist_ok=True)
+    _clean_gcda(temp_dir)
+
+    sub_dir = _resolve_sub_dir(source_dir)
+    cpy_dir = _resolve_cpy_dir(source_dir)
+    sub_o = compile_sub_modules(sub_dir, temp_dir, cpy_dir)
+    exe_path = compile_program(program_name, source_dir, temp_dir, sub_o, cpy_dir)
+
+    if normal_recs:
+        group = GroupInfo(
+            name='normal', records=normal_recs, expected_outputs=[],
+            fd_field_dicts=fd_field_dicts, open_dir=open_dir,
+            select_info=select_info,
+        )
+        r = run_group(group, exe_path, temp_dir)
+        result['normal_pass'] = (r.returncode == 0)
+        result['normal_returncode'] = r.returncode
+        for fd_name in fd_field_dicts:
+            sel = select_info.get(fd_name, {})
+            assign = sel.get('assign', fd_name) if isinstance(sel, dict) else fd_name
+            outpath = os.path.join(temp_dir, assign)
+            fds = fd_field_dicts.get(fd_name, [])
+            org = sel.get('organization', 'SEQUENTIAL') if isinstance(sel, dict) else 'SEQUENTIAL'
+            line_seq = (org == 'LINE SEQUENTIAL')
+            try:
+                recs = file_io.read_output_file(outpath, fds, line_sequential=line_seq)
+                result['output_summary'][fd_name] = len(recs)
+            except Exception:
+                result['output_summary'][fd_name] = -1
+
+    for i, rec in enumerate(abend_recs):
+        group = GroupInfo(
+            name=f'abend_{i+1}', records=[rec], expected_outputs=[],
+            expected_returncode=1,
+            fd_field_dicts=fd_field_dicts, open_dir=open_dir,
+            select_info=select_info,
+        )
+        r = run_group(group, exe_path, temp_dir)
+        if r.returncode != 0:
+            result['abend_pass'] += 1
+
+    return result

runners/cobol_runner.py

@@ -1,10 +1,20 @@
+import os
 import subprocess
 from pathlib import Path
 from runners.runner import BuildResult, RunResult
 
 
 class CobolRunner:
+    """COBOL 程序编译·执行器。
+
+    非DB程序（KIN系）使用 ``compile_with_links()`` + ``run_file_based()``。
+    DB程序／旧式调用保持 ``compile()`` + ``run()`` 不变。
+    """
+
+    # ── 旧式（orchestrator.py 互換）──
+
     def compile(self, src: str, dialect="ibm", gcov: bool = False) -> BuildResult:
+        """旧式编译（-std=ibm-strict）。orchestrator.py 用，不修改。"""
         stem = Path(src).stem
         out = str(Path(src).parent / stem)
         cmd = ["cobc", "-x", f"-std={dialect}-strict", "-o", out, src]
@@ -14,8 +24,92 @@ class CobolRunner:
         return BuildResult(success=p.returncode == 0, artifact_path=out, log=p.stdout + p.stderr)
 
     def run(self, binary: str, input_path: str, output_path: str) -> RunResult:
+        """旧式执行（stdin→stdout）。orchestrator.py 用。"""
         with open(input_path, "rb") as f:
             data = f.read()
         p = subprocess.run([binary], input=data, capture_output=True, timeout=30)
         Path(output_path).write_bytes(p.stdout)
         return RunResult(success=p.returncode == 0)
+
+    # ── 新式（非DB KIN系向け）──
+
+    def compile_with_links(self, src: str, work_dir: str,
+                           copybook_dirs: list[str] | None = None,
+                           sub_objects: list[str] | None = None,
+                           gcov: bool = False) -> BuildResult:
+        """编译主程序 + 链接 SUB.o，不使用 -std=ibm-strict。
+
+        Args:
+            src: COBOL 源文件路径。
+            work_dir: 编译工作目录（.gcno 产出位置，必须提前存在）。
+            copybook_dirs: COPYBOOK 搜索路径列表（映射为 -I 参数）。
+            sub_objects: SUB*.o 文件路径列表（静态链接）。
+            gcov: 是否启用 ``--coverage``。
+
+        Returns:
+            BuildResult，artifact_path 指向 .exe 的绝对路径。
+        """
+        work_path = Path(work_dir).resolve()
+        work_path.mkdir(parents=True, exist_ok=True)
+
+        stem = Path(src).stem
+        exe_path = work_path / f'{stem}.exe'
+
+        cmd = ["cobc", "-x", "-g"]
+        if gcov:
+            cmd.append("--coverage")
+        for cpy in (copybook_dirs or []):
+            cmd.extend(["-I", str(cpy)])
+        cmd.extend(["-o", str(exe_path), str(src)])
+        cmd.extend(str(o) for o in (sub_objects or []))
+
+        orig = os.getcwd()
+        try:
+            os.chdir(str(work_path))
+            p = subprocess.run(cmd, capture_output=True, text=True, timeout=120)
+        finally:
+            os.chdir(orig)
+
+        return BuildResult(
+            success=p.returncode == 0,
+            artifact_path=str(exe_path),
+            log=(p.stdout + p.stderr)[:2000],
+        )
+
+    def run_file_based(self, binary: str, run_dir: str,
+                       input_files: dict[str, str] | None = None,
+                       timeout: int = 60) -> RunResult:
+        """基于文件的执行。将输入文件复制到 run_dir 后启动程序。
+
+        Args:
+            binary: 可执行文件路径。
+            run_dir: 执行目录（CWD，输入文件在此，输出文件也在此）。
+            input_files: {assign_name: source_path} 的映射。
+                如 ``{"KIN04R01": "/path/to/input/KIN04R01"}``。
+            timeout: 超时秒数。
+
+        Returns:
+            RunResult。
+        """
+        run_path = Path(run_dir).resolve()
+        run_path.mkdir(parents=True, exist_ok=True)
+
+        # 入力ファイル配置
+        for assign_name, src_path in (input_files or {}).items():
+            src = Path(src_path)
+            if src.exists():
+                dst = run_path / assign_name
+                import shutil
+                shutil.copy2(str(src), str(dst))
+
+        orig = os.getcwd()
+        try:
+            os.chdir(str(run_path))
+            p = subprocess.run([str(binary)], capture_output=True, text=True, timeout=timeout)
+        finally:
+            os.chdir(orig)
+
+        return RunResult(
+            success=p.returncode == 0,
+            log=(p.stdout + p.stderr)[:2000],
+        )