[PATCH] newlib: add dummy implementations of fe{get,set}prec for Aarch64 Cygwin

[Radek Barton]

Hello.

Albeit Aarch64 does not support setting floating point operations precisions at runtime `cygwin1.dll` exports `fegetprec` and `fesetprec` functions leading to undefined references when linking. This patch adds their dummy implementations in a similar way how they are defined in `newlib/libc/machine/shared_x86/sys/fenv.h`, resp. in `/newlib/libm/machine/shared_x86/fenv.c`.

Alternative fix would be to modify `winsup/cygwin/scripts/mkimport` and `winsup/cygwin/scripts/gendef` scripts so they would exclude those from being exported to the DLL upon specific label in `winsup/cygwin/cygwin.din`, e.g.:

```
fegetprec NOSIGFE NOARM64
fesetprec NOSIGFE NOARM64
```

The same options applies to `_fe_nomask_env` which would be for the first option sent as a separate patch as it changes only `winsup` codebase.

Which one would you prefer?

Radek

---
>From 0ea739b8eef37d433712c185e0562c9f3142b86c Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Radek=20Barto=C5=88?= <radek.barton@microsoft.com>
Date: Thu, 3 Jul 2025 12:02:03 +0200
Subject: [PATCH] newlib: add dummy implementations of fe{get,set}prec
 for Aarch64 Cygwin
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Signed-off-by: Radek Bartoň <radek.barton@microsoft.com>
---
 newlib/libc/machine/aarch64/sys/fenv.h | 24 ++++++++++++++++++++-
 newlib/libm/machine/aarch64/fenv.c     | 30 ++++++++++++++++++++++++++
 2 files changed, 53 insertions(+), 1 deletion(-)

diff --git a/newlib/libc/machine/aarch64/sys/fenv.h b/newlib/libc/machine/aarch64/sys/fenv.h
index 212612725..0e4e7f492 100644
--- a/newlib/libc/machine/aarch64/sys/fenv.h
+++ b/newlib/libc/machine/aarch64/sys/fenv.h
@@ -61,7 +61,14 @@ typedef	__uint64_t	fexcept_t;
 			 FE_UPWARD | FE_TOWARDZERO)
 #define	_ROUND_SHIFT	22
 
-
+/* Only Solaris and QNX implement fegetprec/fesetprec.  As Solaris, use the
+   values defined by http://www.open-std.org/jtc1/sc22//WG14/www/docs/n752.htm
+   QNX defines different values. */
+#if __MISC_VISIBLE
+#define FE_FLTPREC	(0)
+#define FE_DBLPREC	(2)
+#define FE_LDBLPREC	(3)
+#endif
 
 /* Default floating-point environment */
 extern const fenv_t	*_fe_dfl_env;
@@ -115,6 +122,21 @@ fegetexcept(void)
 
 #endif /* __BSD_VISIBLE */
 
+#ifdef __cplusplus
+extern "C" {
+#endif
 
+#ifdef __CYGWIN__
+
+#if __MISC_VISIBLE
+int fegetprec (void);
+int fesetprec (int __prec);
+#endif
+
+#endif /* __CYGWIN__ */
+
+#ifdef __cplusplus
+}
+#endif
 
 #endif	/* !_FENV_H_ */
diff --git a/newlib/libm/machine/aarch64/fenv.c b/newlib/libm/machine/aarch64/fenv.c
index 3ffe23441..3f3459e9d 100644
--- a/newlib/libm/machine/aarch64/fenv.c
+++ b/newlib/libm/machine/aarch64/fenv.c
@@ -55,3 +55,33 @@ extern inline int feupdateenv(const fenv_t *__envp);
 extern inline int feenableexcept(int __mask);
 extern inline int fedisableexcept(int __mask);
 extern inline int fegetexcept(void);
+
+#if defined(__CYGWIN__)
+
+/* Returns the currently selected precision, represented by one of the
+   values of the defined precision macros. */
+int
+fegetprec (void)
+{
+  /* AArch64 doesn't have configurable precision.
+     Return a fixed value indicating double precision (most common). */
+  return FE_DBLPREC;
+}
+
+/* http://www.open-std.org/jtc1/sc22//WG14/www/docs/n752.htm:
+
+   The fesetprec function establishes the precision represented by its
+   argument prec.  If the argument does not match a precision macro, the
+   precision is not changed.
+
+   The fesetprec function returns a nonzero value if and only if the
+   argument matches a precision macro (that is, if and only if the requested
+   precision can be established). */
+int
+fesetprec (int prec)
+{
+  /* Aarch64 doesn't support changing precision at runtime. */
+  return 0; // return failure
+}
+
+#endif
-- 
2.49.0.vfs.0.4
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-newlib-add-dummy-implementations-of-fe-get-set-prec-for-Aarch64-Cygwin.patch
Type: application/octet-stream
Size: 2929 bytes
Desc: 0001-newlib-add-dummy-implementations-of-fe-get-set-prec-for-Aarch64-Cygwin.patch
URL: <https://sourceware.org/pipermail/newlib/attachments/20250703/c5c411e3/attachment-0001.obj>