init:

example/c-sum/README.md

@@ -0,0 +1,49 @@
+# c-sum — cross-platform C example
+
+Minimal example: a C program that reads `N` then `N` integers and prints their
+sum. Tested with `judge` across **gcc / clang / MSVC** on **Linux / Windows**,
+with optional **valgrind** and **sanitizer** runs.
+
+## Files
+
+- `solution.c` — student-facing solution (could be what the student submits)
+- `sum.jdg` — judge test suite
+- `.gitea/workflows/judge.yml` — Gitea CI matrix
+
+## Run locally
+
+```sh
+# Linux / macOS
+CC=gcc   judge sum.jdg .
+CC=clang judge sum.jdg .
+
+# With valgrind
+judge sum.jdg . --wrapper="valgrind --error-exitcode=99 --leak-check=full -q"
+
+# With ASan+UBSan build
+CC=clang CFLAGS="-O1 -g -fsanitize=address,undefined" judge sum.jdg .
+```
+
+On Windows (inside an MSVC dev cmd shell), `build_windows` kicks in and
+produces `solution.exe`, which the runner auto-detects.
+
+## Notes about the `.jdg`
+
+- `normalize_crlf = true` — Windows `printf` emits `\r\n`; we strip `\r` before
+  matching so the same expected outputs work on both platforms.
+- `trim_trailing_ws = true` — forgives trailing spaces a student's output might
+  pick up (rare but annoying to debug).
+- `binary = "solution"` — the runner tries `solution` first, then
+  `solution.exe` on Windows automatically.
+- Per-group `scoring = all_or_none` on `stress` gives weight only if every
+  stress test passes.
+
+## Adapting the Gitea workflow
+
+- `runs-on: ${{ matrix.toolchain.system }}-Runner` assumes you have
+  self-hosted Gitea runners labelled `Linux-Runner` / `Windows-Runner` (same
+  naming as your existing `fixed_floating` pipeline).
+- `secrets.VAR_TOKEN` is only needed when pulling a student repo from a
+  private org.
+- The summary job shells `grep` over the JSON; swap to `jq` if available on
+  your runners.