Fix build system regressions

jserv · jserv · commit 30ac02575c68 · 2025-11-12T15:56:56.000+08:00
This commit addresses multiple regressions introduced in recent changes:
1. DTB compilation regression
   - DTB dependencies moved outside CC_IS_EMCC conditional
   - Ensures DTB builds for system mode regardless of compiler
   - Fixes mk/wasm.mk structure for cross-platform consistency
2. Makefile syntax error in mk/toolchain.mk
   - Fixed TAB characters before $(warning) on lines 25, 28
   - Changed to spaces for proper control flow
   - This was blocking all Makefile parsing
3. emcc configuration pollution
   - Added 'make distclean' before emcc builds in workflow
   - Prevents ENABLE_SYSTEM=1 from leaking between builds
   - Fixes "build/minimal.dtb does not exist" errors
4. Ubuntu ARM64 apt-get failures
   - Implemented exponential backoff retry mechanism (30s, 60s delays)
   - Added pipefail to preserve apt exit codes through tee
   - Explicit APT_EXIT capture to detect masked failures
   - Added InRelease to failure pattern (modern combined Release+GPG)
   - Ignore non-critical dep11 metadata failures
   - Focus on core package indices (Packages/Sources/Release/InRelease)
5. TSAN cross-compiler compatibility (fixed __has_feature issue)
   - Changed from defined(__has_feature) to defined(__clang__)
   - GCC doesn't support __has_feature, causing preprocessor errors
   - __has_feature only works when __clang__ is defined
   - Ensures __tsan_default_options() works with both GCC and clang
6. TSAN cross-platform compatibility
   - Guarded setarch with ifeq ($(UNAME_S),Linux) in Makefile
   - setarch doesn't exist on macOS, now conditionally applied
   - macOS TSAN builds require SIP disabled for ASLR control
7. Trace functionality regression
   - Reverted .log_level from LOG_INFO back to LOG_TRACE
   - LOG_INFO suppressed rv_log_trace() stream used by -t flag
   - Restores instruction trace output for debugging
diff --git a/.github/workflows/main.yml b/.github/workflows/main.yml
@@ -151,14 +151,15 @@ jobs:
     - name: default build using emcc
       if: success()
       run: |
+            make distclean
             make CC=emcc ENABLE_JIT=0 $PARALLEL
 
     - name: default build for system emulation using emcc
       if: success()
       run: |
             make distclean
             make CC=emcc ENABLE_SYSTEM=1 ENABLE_JIT=0 $PARALLEL
-            make distclean ENABLE_SYSTEM=1
+            make distclean
 
     - name: Build with various optimization levels
       if: success()
@@ -317,41 +318,56 @@ jobs:
         githubToken: ${{ github.token }}
         # No 'sudo' is available
         install: |
-          set +e  # Disable errexit for entire install block
-          # Configure apt to tolerate mirror sync failures
-          cat > /etc/apt/apt.conf.d/99-ci-reliability << 'APTCONF'
-          APT::Update::Error-Mode "any";
-          Acquire::IndexTargets::deb::DEP-11::DefaultEnabled "false";
-          Acquire::IndexTargets::deb::DEP-11-icons::DefaultEnabled "false";
-          Acquire::IndexTargets::deb::DEP-11-icons-hidpi::DefaultEnabled "false";
-          Acquire::Retries "3";
-          Acquire::Check-Valid-Until "false";
-          APTCONF
-          # Update with error tolerance - may fail during mirror sync
-          apt update -qq 2>&1
-          UPDATE_EXIT=$?
-          if [ $UPDATE_EXIT -ne 0 ]; then
-            echo "WARNING: apt update exited with code $UPDATE_EXIT (mirror sync likely in progress)"
-            echo "Continuing with installation using cached/partial indexes..."
-          fi
-          # Install packages - will use whatever indexes are available
-          apt install -yqq make git curl wget clang libsdl2-dev libsdl2-mixer-dev lsb-release software-properties-common gnupg bc || {
-            echo "WARNING: Some packages may have failed to install, retrying critical ones..."
-            apt install -yqq --no-download make git curl wget clang || true
-          }
-          which wget || echo "WARNING: wget not found after installation"
-          set -e  # Re-enable errexit
-          exit 0  # Force success exit code
+          # Retry apt update with exponential backoff for mirror sync issues
+          # Note: dep11 (AppStream metadata) failures are non-critical for build tools
+          set -o pipefail
+          for i in 1 2 3; do
+            if apt update -qq --allow-releaseinfo-change 2>&1 | tee /tmp/apt-update.log; then
+              APT_EXIT=0
+            else
+              APT_EXIT=$?
+            fi
+            # Check for critical failures (package indices), ignore dep11 metadata
+            # Include InRelease which is the combined Release+Release.gpg file
+            if [ $APT_EXIT -eq 0 ] && ! grep -E "Failed to fetch.*/(Packages|Sources|Release|InRelease)" /tmp/apt-update.log; then
+              echo "apt update succeeded (core package lists available)"
+              break
+            fi
+            if [ $i -lt 3 ]; then
+              delay=$((i * 30))
+              echo "apt update attempt $i: errors detected (exit=$APT_EXIT), waiting ${delay}s..."
+              sleep $delay
+            else
+              echo "Warning: Proceeding after 3 attempts - some package lists may be incomplete"
+            fi
+          done
+          # Install packages - exit 0 even if dep11 metadata is incomplete
+          apt install -yqq make git curl wget clang libsdl2-dev libsdl2-mixer-dev lsb-release software-properties-common gnupg bc 2>&1 | tee /tmp/apt-install.log || true
+          # Verify critical packages were installed
+          for pkg in make git curl clang bc; do
+            if ! command -v $pkg >/dev/null 2>&1; then
+              echo "ERROR: Critical package $pkg failed to install!"
+              cat /tmp/apt-install.log
+              exit 1
+            fi
+          done
+          echo "All critical build tools installed successfully"
         # FIXME: gcc build fails on Aarch64/Linux hosts
         env: |
           CC: clang-18
         # Append custom commands here
         run: |
           # Verify and install wget if needed (workaround for install step issues)
           if ! command -v wget > /dev/null 2>&1; then
-            # Config file should already exist from install step, but apt may still fail
-            apt update -qq 2>&1 || true
-            apt install -yqq wget || { echo "wget install failed, trying without update..."; apt install -yqq wget --no-download || true; }
+            echo "wget not found, attempting to install..."
+            apt update -qq --allow-releaseinfo-change 2>&1 | tee /tmp/apt-update-wget.log || true
+            apt install -yqq wget 2>&1 | tee /tmp/wget-install.log || true
+            if ! command -v wget > /dev/null 2>&1; then
+              echo "ERROR: wget installation failed!"
+              cat /tmp/wget-install.log
+              exit 1
+            fi
+            echo "wget installed successfully"
           fi
           git config --global --add safe.directory ${{ github.workspace }}
           git config --global --add safe.directory ${{ github.workspace }}/src/softfloat
@@ -475,14 +491,15 @@ jobs:
      - name: default build using emcc
        if: success()
        run: |
+             make distclean
              make CC=emcc ENABLE_JIT=0 $PARALLEL
 
      - name: default build for system emulation using emcc
        if: success()
        run: |
              make distclean
              make CC=emcc ENABLE_SYSTEM=1 ENABLE_JIT=0 $PARALLEL
-             make distclean ENABLE_SYSTEM=1
+             make distclean
 
      - name: check + tests
        if: success()
@@ -539,14 +556,14 @@ jobs:
                fi
              done
 
-    - name: JIT debug test
-      env:
-        CC: ${{ steps.install_cc.outputs.cc }}
-      run: |
-            # Run JIT tests with debug mode to catch register allocation and cache coherency issues
-            make distclean && make ENABLE_JIT=1 ENABLE_JIT_DEBUG=1 check $PARALLEL
-            make distclean && make ENABLE_EXT_C=0 ENABLE_JIT=1 ENABLE_JIT_DEBUG=1 check $PARALLEL
-      if: ${{ always() }}
+     - name: JIT debug test
+       env:
+         CC: ${{ steps.install_cc.outputs.cc }}
+       run: |
+             # Run JIT tests with debug mode to catch register allocation and cache coherency issues
+             make distclean && make ENABLE_JIT=1 ENABLE_JIT_DEBUG=1 check $PARALLEL
+             make distclean && make ENABLE_EXT_C=0 ENABLE_JIT=1 ENABLE_JIT_DEBUG=1 check $PARALLEL
+       if: ${{ always() }}
 
      - name: undefined behavior test
        if: (success() || failure()) && steps.install_cc.outputs.cc == 'clang'  # gcc on macOS/arm64 does not support sanitizers
diff --git a/Makefile b/Makefile
@@ -106,10 +106,15 @@ override ENABLE_SDL := 0       # SDL (uninstrumented system lib) creates threads
 override ENABLE_LTO := 0       # LTO interferes with TSAN instrumentation
 CFLAGS += -DTSAN_ENABLED       # Signal code to use TSAN-compatible allocations
 # Disable ASLR for TSAN tests to prevent allocations in TSAN shadow memory
+# Note: setarch is Linux-only; macOS requires different approach (SIP disable)
+ifeq ($(UNAME_S),Linux)
 BIN_WRAPPER = setarch $(shell uname -m) -R
 else
 BIN_WRAPPER =
 endif
+else
+BIN_WRAPPER =
+endif
 
 # Enable link-time optimization (LTO)
 ENABLE_LTO ?= 1
@@ -392,7 +397,7 @@ DTB_DEPS := $(BUILD_DTB) $(BUILD_DTB2C)
 endif
 endif
 
-all: config $(DTB_DEPS) $(BUILD_DTB) $(BUILD_DTB2C) $(BIN)
+all: config $(DTB_DEPS) $(BIN)
 
 OBJS := \
     map.o \
@@ -437,7 +442,7 @@ $(OUT):
 
 $(BIN): $(OBJS) $(DEV_OBJS) | $(OUT)
 	$(VECHO) "  LD\t$@\n"
-	$(Q)$(CC) -o $@ $(CFLAGS_emcc) $^ $(LDFLAGS)
+	$(Q)$(CC) -o $@ $(CFLAGS_emcc) $(filter-out %.dtb %.h,$^) $(LDFLAGS)
 
 $(CONFIG_FILE): FORCE
 	$(Q)mkdir -p $(OUT)
diff --git a/mk/toolchain.mk b/mk/toolchain.mk
@@ -9,7 +9,7 @@ ifneq ($(shell $(CC) --version | head -n 1 | grep emcc),)
     EMCC_MINOR := $(shell echo $(EMCC_VERSION) | cut -f2 -d.)
     EMCC_PATCH := $(shell echo $(EMCC_VERSION) | cut -f3 -d.)
 
-    # When the emcc version is not 3.1.51, the latest SDL2_mixer library is fetched by emcc and music might not be played in the web browser
+    # When the emcc version is not 3.1.51, the latest SDL2_mixer library is fetched by emcc and music might not be played in the web browser
     SDL_MUSIC_PLAY_AT_EMCC_MAJOR := 3
     SDL_MUSIC_PLAY_AT_EMCC_MINOR := 1
     SDL_MUSIC_PLAY_AT_EMCC_PATCH := 51
diff --git a/mk/wasm.mk b/mk/wasm.mk
@@ -167,3 +167,21 @@ start-web: $(start_web_deps)
 .PHONY: check-demo-dir-exist start-web
 
 endif
+
+# For SYSTEM mode, DTB needs to be built regardless of whether we're using emcc
+# DTB is only built when SYSTEM=1 and ELF_LOADER=0
+ifeq ($(call has, SYSTEM), 1)
+ifeq ($(call has, ELF_LOADER), 0)
+# Add DTB as dependency for compilation stages
+# This is used by mk/system.mk for device object files
+deps_emcc += $(BUILD_DTB) $(BUILD_DTB2C)
+
+# For emcc builds: ensure DTB exists before emcc embeds it
+# Make BIN directly depend on DTB files as regular prerequisites
+# This will cause them to be built, but they'll also be passed to the linker
+# We need to filter them out in the linker command
+ifeq ("$(CC_IS_EMCC)", "1")
+$(BIN): $(BUILD_DTB) $(BUILD_DTB2C)
+endif
+endif
+endif
diff --git a/src/main.c b/src/main.c
@@ -28,6 +28,7 @@
  *
  * Configuration optimizes for race detection with minimal overhead.
  */
+/* GCC uses __SANITIZE_THREAD__, clang uses __has_feature(thread_sanitizer) */
 #if defined(__SANITIZE_THREAD__)
 const char *__tsan_default_options()
 {
@@ -39,6 +40,19 @@ const char *__tsan_default_options()
            ":history_size=7"          /* Larger race detection window */
            ":io_sync=0";              /* Don't sync on I/O */
 }
+#elif defined(__clang__)
+#if __has_feature(thread_sanitizer)
+const char *__tsan_default_options()
+{
+    return "halt_on_error=0"          /* Continue after errors */
+           ":report_bugs=1"           /* Report data races */
+           ":second_deadlock_stack=1" /* Full deadlock info */
+           ":verbosity=0"             /* Reduce noise */
+           ":memory_limit_mb=0"       /* No memory limit */
+           ":history_size=7"          /* Larger race detection window */
+           ":io_sync=0";              /* Don't sync on I/O */
+}
+#endif
 #endif
 
 /* enable program trace mode */
@@ -304,7 +318,7 @@ int main(int argc, char **args)
         .args_offset_size = ARGS_OFFSET_SIZE,
         .argc = prog_argc,
         .argv = prog_args,
-        .log_level = LOG_INFO,
+        .log_level = LOG_TRACE,
         .run_flag = run_flag,
         .profile_output_file = prof_out_file,
         .cycle_per_step = CYCLE_PER_STEP,
