Delete fuzzer (#198)

.github/workflows/linux.yml

@@ -19,13 +19,3 @@ jobs:
     - run: ./runtests.sh --verbose './jou ${{ matrix.opt-level }} --verbose %s'
     - run: ./runtests.sh --verbose --valgrind './jou ${{ matrix.opt-level }} %s'
     # valgrind+verbose isn't meaningful: test script would ignore valgrind output
-
-  fuzzer:
-    runs-on: ubuntu-latest
-    strategy:
-      matrix:
-        llvm-version: [11, 13]
-    steps:
-    - uses: actions/checkout@v3
-    - run: sudo apt install -y llvm-${{ matrix.llvm-version }}-dev clang-${{ matrix.llvm-version }} make valgrind
-    - run: LLVM_CONFIG=llvm-config-${{ matrix.llvm-version }} ./fuzzer.sh

.github/workflows/windows.yml

@@ -131,19 +131,3 @@ jobs:
         shell: bash
       - run: cd "test dir" && ./runtests.sh --verbose
         shell: bash
-
-  fuzzer:
-    needs: build
-    runs-on: windows-latest
-    steps:
-      - uses: actions/checkout@v3
-        with:
-          path: repo
-      - uses: actions/download-artifact@v3
-        with:
-          name: windows-zip
-      - run: unzip jou.zip
-      - run: mv jou/* repo/tests repo/fuzzer.sh .
-        shell: bash
-      - run: ./fuzzer.sh
-        shell: bash

CONTRIBUTING.md

@@ -140,14 +140,6 @@ This doesn't do anything with tests that are supposed to fail with an error, for
 There are also a few other ways to run the tests.
 You can look at `.github/workflows/` to see how the CI runs tests.
 
-Sometimes the fuzzer discovers a bug that hasn't been caught with tests.
-It mostly finds bugs in the tokenizer,
-because the fuzzer works by feeding random bytes to the compiler.
-
-```
-$ ./fuzzer.sh
-```
-
 
 ## Windows Release Builds