fix: 阶段二循环上界填空题面与答案大小写口径不一致

services/demo_experience.py

@@ -300,7 +300,7 @@ def _ensure_preset(assignment):
             'step_id': 2,
             'part_name': '主程序',
             'type': 'fill',
-            'question': '补全循环条件，确保从第三项计算到第 N 项。',
+            'question': '补全循环条件，确保从第三项计算到第 N 项（程序读入的变量名为小写 n）。',
             'context_before': 'for (int i = 2; i <',
             'context_after': '; ++i) { ... }',
             'blank_hint': '输入循环上界',

tests/test_demo_stage2_loop_bound_prompt.py

@@ -0,0 +1,51 @@
+"""Regression for the stage-2 loop-bound fill-in prompt.
+
+During a real guided-learning run the fill-in question asks the student to
+"计算到第 N 项" (uppercase N), but the accepted answer is the lowercase
+identifier ``n`` because the reference program declares ``int n;``. The
+question context only shows ``for (int i = 2; i < __; ++i)`` -- the variable
+declaration is not visible -- so a student following the question's own wording
+types ``N`` and is marked wrong for a "case error" created by the content
+itself.
+
+Fix contract: the question prompt must state explicitly that the program's
+variable is lowercase ``n``, while ``correct_answer`` stays ``n`` (the value
+that actually compiles).
+
+Stdlib only: this test reads the seeder source text. It does not import Flask,
+open a database/Redis connection, or start the application.
+"""
+import re
+from pathlib import Path
+
+SOURCE = (
+    Path(__file__).resolve().parents[1]
+    / "services"
+    / "demo_experience.py"
+)
+
+
+def _loop_bound_step_block():
+    """Return the source text of the step_id 2 quiz-step dict."""
+    text = SOURCE.read_text(encoding="utf-8")
+    start = text.index("'step_id': 2")
+    end = text.index("'step_id': 3", start)
+    return text[start:end]
+
+
+def test_loop_bound_question_names_lowercase_variable():
+    block = _loop_bound_step_block()
+    question_match = re.search(r"'question':\s*'([^']*)'", block)
+    assert question_match is not None
+    question = question_match.group(1)
+    # The prompt keeps the math symbol N but must tell the student what the
+    # program variable is called, so the natural answer matches correct_answer.
+    assert "第 N 项" in question
+    assert "小写 n" in question
+
+
+def test_loop_bound_correct_answer_stays_lowercase_n():
+    # n must remain the accepted answer: the reference program reads into a
+    # variable declared as `int n;`, so only n keeps the generated code valid.
+    block = _loop_bound_step_block()
+    assert re.search(r"'correct_answer':\s*'n'", block) is not None