arm/neon abs: negating INT_MIN is undefined behavior

- So cast to unsigned int before flipping sign, and then cast back to a signed int

LLVM 19 is making this more prominant: llvm/llvm-project#82112 (comment)
But this was already visible in earlier clang versions with `-O2` #901

- gh-actions: resume testing emscripten using the 'tip of tree' ("tot") builds.
- gh-actions: add clang-17 "-O2" build to confirm the fix

.github/workflows/ci.yml

@@ -168,8 +168,8 @@ jobs:
       run: |
         git clone https://github.com/emscripten-core/emsdk.git /opt/emsdk
         cd /opt/emsdk
-        ./emsdk install 4e7eadf19c76c143e522b535254eb99a9b34fd6d
-        ./emsdk activate 4e7eadf19c76c143e522b535254eb99a9b34fd6d
+        ./emsdk install tot
+        ./emsdk activate tot
         source emsdk_env.sh
     - name: Install v8
       run: |
@@ -532,6 +532,9 @@ jobs:
           distro: ubuntu-22.04
           arch_flags: -ffast-math -Wno-unsafe-buffer-usage
           ccache: 'true'
+        - version: "17"
+          distro: ubuntu-22.04
+          arch_flags: -march=native -Wno-unsafe-buffer-usage -O2
     runs-on: ${{ matrix.distro }}
     env:
       CFLAGS: ${{ matrix.arch_flags }} -Wall -Weverything -Werror -fno-lax-vector-conversions

docker/cross-files/i686-gcc-14.cross

@@ -0,0 +1,17 @@
+[binaries]
+c = 'i686-linux-gnu-gcc-14'
+cpp = 'i686-linux-gnu-g++-14'
+ar = 'i686-linux-gnu-ar'
+strip = 'i686-linux-gnu-strip'
+objcopy = 'i686-linux-gnu-objcopy'
+ld = 'i686-linux-gnu-ld'
+
+[properties]
+c_args = ['-Wextra', '-Werror', '-march=prescott']
+cpp_args = ['-Wextra', '-Werror', '-march=prescott']
+
+[host_machine]
+system = 'linux'
+cpu_family = 'x86'
+cpu = 'prescott'
+endian = 'little'

simde/arm/neon/abs.h

@@ -437,7 +437,7 @@ simde_vabsq_s32(simde_int32x4_t a) {
     #else
       SIMDE_VECTORIZE
       for (size_t i = 0 ; i < (sizeof(r_.values) / sizeof(r_.values[0])) ; i++) {
-        r_.values[i] = a_.values[i] < 0 ? -a_.values[i] : a_.values[i];
+        r_.values[i] = a_.values[i] < 0 ? HEDLEY_STATIC_CAST(int32_t, 0 - HEDLEY_STATIC_CAST(uint32_t, a_.values[i])) : a_.values[i];
       }
     #endif
 
@@ -476,7 +476,7 @@ simde_vabsq_s64(simde_int64x2_t a) {
     #else
       SIMDE_VECTORIZE
       for (size_t i = 0 ; i < (sizeof(r_.values) / sizeof(r_.values[0])) ; i++) {
-        r_.values[i] = a_.values[i] < 0 ? -a_.values[i] : a_.values[i];
+        r_.values[i] = a_.values[i] < 0 ? HEDLEY_STATIC_CAST(int64_t, 0 - HEDLEY_STATIC_CAST(uint64_t, a_.values[i])) : a_.values[i];
       }
     #endif