[PATCH] Cygwin: Adapt math functions to use 64bit long double on aarch64

[Jon Turney]

On 29/04/2026 12:33, Chandru Kumaresan wrote:
> Hi Corinna,
> The previously mentioned changes are not the only ones required. A few additional files also need to be modified to ensure that Cygwin works correctly on aarch64.
> The approach taken is consistent with mingw-w64. mingw-w64 uses the __SIZEOF_LONG_DOUBLE__ == __SIZEOF_DOUBLE__ macro to implement aarch64-specific code for math functions.
> For reference, a similar approach is used in this upstream mingw-w64 commit:
> https://github.com/mingw-w64/mingw-w64/commit/dbb60ad07c2983027cd09f0f7221505400391caa
> 
[...]
> diff --git a/winsup/cygwin/math/remainderl.S b/winsup/cygwin/math/remainderl.S
> index a69e38296..f05724e94 100644
> --- a/winsup/cygwin/math/remainderl.S
> +++ b/winsup/cygwin/math/remainderl.S
> @@ -7,15 +7,15 @@
> 
>      .file   "remainderl.S"
>      .text
> -#ifdef __x86_64__
> +#if defined(__x86_64__)
>      .align 8
> -#else
> +#elif defined(__i386__)
>      .align 4
>   #endif
>   .globl __MINGW_USYMBOL(remainderl)
>      .def    __MINGW_USYMBOL(remainderl);    .scl    2;  .type   32; .endef
>   __MINGW_USYMBOL(remainderl):
> -#ifdef __x86_64__
> +#if defined(__x86_64__)
>      fldt    (%r8)
>      fldt    (%rdx)
>   1: fprem1
> @@ -27,7 +27,7 @@ __MINGW_USYMBOL(remainderl):
>      movq    $0,8(%rcx)
>      fstpt   (%rcx)
>      ret
> -#else
> +#elif defined(__i386__)
>      fldt    16(%esp)
>      fldt    4(%esp)
>   1: fprem1
> @@ -36,4 +36,9 @@ __MINGW_USYMBOL(remainderl):
>      jp  1b
>      fstp    %st(1)
>      ret
> +#elif _defined(__aarch64__)

Now I have a (minimal) ability to compile this, I've noticed that there 
seems to be typo here.

I'm assuming the attached trivial patch is what's needed.

> +   bl  remainder
> +   ret
> +#else
> +   .error "Not supported on your platform yet"
>   #endif
-------------- next part --------------
From 3156c3c350eb0e9e1a1b65e86414f94ba1b4b7e4 Mon Sep 17 00:00:00 2001
From: Jon Turney <jon.turney@dronecode.org.uk>
Date: Thu, 28 May 2026 22:17:29 +0100
Subject: [PATCH] Cygwin: Fix typo in aarch64 preprocessor conditional in
 remainderl.S
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Fix a typo in the aarch64 preprocessor conditional used in remainderl.S.

> ../../../../src/winsup/cygwin/math/remainderl.S:39:15: error: missing binary operator before token ‘(’
>    39 | #elif _defined(__aarch64__)
>       |               ^

Fixes: a1f347c0d5b8 ("Cygwin: Adapt math functions to use 64bit long double on aarch64")
---
 winsup/cygwin/math/remainderl.S | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/winsup/cygwin/math/remainderl.S b/winsup/cygwin/math/remainderl.S
index f05724e94..9b079eca3 100644
--- a/winsup/cygwin/math/remainderl.S
+++ b/winsup/cygwin/math/remainderl.S
@@ -36,7 +36,7 @@ __MINGW_USYMBOL(remainderl):
 	jp	1b
 	fstp	%st(1)
 	ret
-#elif _defined(__aarch64__)
+#elif defined(__aarch64__)
 	bl	remainder
 	ret
 #else
-- 
2.51.0