riscv: allow single-precision float ABI (lp64f / ilp32f) (#349)

* riscv: allow single-precision float ABI (lp64f / ilp32f)

The header rejected __riscv_float_abi_single with an #error, but
the F and D extensions share the same FCSR. The hard-float inline
implementations below the guard use only csrr/csrw/csrc on fcsr
and fflags, both of which are present whenever F is present. So
lp64f / ilp32f can use the same code path as the double ABI.

Closes #262.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* add CI coverage for the RISC-V single-float ABI (#262)

The lp64f / ilp32f ABI used to be rejected by openlibm_fenv_riscv.h with
an #error. Add a compile-only cross job that rebuilds the static library
with -mabi=lp64f so the removed #error stays removed. A full lp64f
shared-lib link or qemu run is impossible because Ubuntu's
gcc-riscv64-linux-gnu ships only the lp64d sysroot; the static archive
needs no libc link, so it still compiles riscv64/fenv.c (and every other
source) under the single-float ABI.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: ViralBShah

.github/workflows/cross.yml

@@ -66,3 +66,35 @@ jobs:
           QEMU_LD_PREFIX: /usr/${{ matrix.config.triple }}
           LD_LIBRARY_PATH: .
         run: make test ARCH=$ARCH TOOLPREFIX=$TRIPLE- ${CC:+CC=$CC} ${AR:+AR=$AR}
+
+  # Regression test for #262 / #349: the RISC-V single-float ABI (lp64f /
+  # ilp32f) used to be rejected by include/openlibm_fenv_riscv.h with an
+  # `#error`. This job rebuilds the whole library with `-mabi=lp64f` to make
+  # sure the header (and the rest of openlibm) accepts that ABI.
+  #
+  # This is a COMPILE-ONLY check (build the static libopenlibm.a, no qemu run).
+  # Ubuntu's gcc-riscv64-linux-gnu ships only the default lp64d sysroot, so a
+  # full shared-library link or a qemu execution under lp64f is impossible
+  # without a custom single-float sysroot the distro does not provide. Building
+  # the static archive needs no libc/CRT link, so it compiles every source file
+  # (including riscv64/fenv.c, which pulls in the fixed header) under the
+  # single-float ABI and directly regresses the removed `#error`. Adding a
+  # bespoke lp64f sysroot would be exactly the kind of fragile, unmaintainable
+  # toolchain dependency that got mips64 disabled in #347, so we deliberately
+  # avoid it.
+  build-cross-riscv64-lp64f:
+    runs-on: ubuntu-24.04
+    timeout-minutes: 10
+    name: build-cross-riscv64-lp64f
+    steps:
+      - uses: actions/checkout@v7
+      - name: Install RISC-V toolchain
+        run: |
+          sudo apt update
+          sudo apt install -y gcc-riscv64-linux-gnu
+      - name: Build libopenlibm.a for the single-float ABI (lp64f)
+        run: |
+          make -j$(nproc) ARCH=riscv64 TOOLPREFIX=riscv64-linux-gnu- \
+            CFLAGS_arch='-march=rv64imafc -mabi=lp64f' \
+            SFLAGS_arch='-march=rv64imafc -mabi=lp64f' \
+            libopenlibm.a

include/openlibm_fenv_riscv.h

@@ -74,13 +74,10 @@ __BEGIN_DECLS
 extern const fenv_t	__fe_dfl_env;
 #define	FE_DFL_ENV	(&__fe_dfl_env)
 
-#if !defined(__riscv_float_abi_soft) && !defined(__riscv_float_abi_double)
-#if defined(__riscv_float_abi_single)
-#error single precision floating point ABI not supported
-#else
+#if !defined(__riscv_float_abi_soft) && !defined(__riscv_float_abi_double) && \
+    !defined(__riscv_float_abi_single)
 #error compiler did not set soft/hard float macros
 #endif
-#endif
 
 #ifndef __riscv_float_abi_soft
 #define	__rfs(__fcsr)	__asm __volatile("csrr %0, fcsr" : "=r" (__fcsr))