fix(pkg): compat.gmp 的 MSVC 分支改用 static __inline(上一版守卫无效)

上一版按"clang 打 MSVC ABI 同时定义 __GNUC__ 与 _MSC_VER"的假设,只给 _MSC_VER
分支补了 `! defined (__GNUC__)` 守卫。CI 结果一字未变(gmp.h 行号从 1845 移到 1852,
证明新头确实生效),这反过来证明:**该编译器不定义 __GNUC__**,守卫恒真,分支照走。

真正的解法是换掉这条分支的答案:`static __inline`,即 GMP 自己给另外两个
"extern inline 会漏出全局符号"的编译器(DEC C、SCO OpenUNIX)的答案。

关键耦合是 __GMP_INLINE_PROTOTYPES:GCC 分支把它设为 1(发 extern 原型,与
gnu_inline 定义相容),MSVC 分支不设 → 默认 0 → 不发 extern 原型,static 定义才不会
和它冲突。本地第一次模拟只换 __GMP_EXTERN_INLINE 不动 __GMP_INLINE_PROTOTYPES,
立刻得到 `static declaration of '__gmpz_abs' follows non-static declaration` ——
忠实复现 MSVC 分支后实测:非 FORCE TU 里那些符号一个都不发射(全内联),FORCE TU
发射唯一的全局,六个 TU 合并链接无重复符号。

守卫保留:将来若有既 GNU 兼容又设 _MSC_VER 的编译器,仍走 gnu_inline 那条。

本地两条 linux 腿重跑 gmp / gmp-gmpxx 仍全绿(_MSC_VER 在那边根本没定义,严格 no-op)。

Author: Sunrisepeak

.agents/docs/2026-08-10-gmp-full-platform-no-build-system.md

@@ -149,29 +149,38 @@ reference。所以做成 feature:11 个 `cxx/*.cc` 合并成**一个** TU(`mcpp_
 
 除 configure 的 8 处 substitute 外,生成的 `gmp.h` 还打了**一个**补丁,由首轮 windows CI 抓出来。
 
-`gmp.h` 用一串厂商分支决定 "inline" 怎么写。GCC 分支给的是
-`extern __inline__ __attribute__ ((__gnu_inline__))` —— GNU89 语义:**头里的函数体只用于内联,
-永远不作为独立符号发射**,而独立那份由带 `-D__GMP_FORCE_<fn>` 的那个 TU 提供
-(`mpz/get_ui.c`、`mpn/generic/sub.c` …)。整个库靠这个分工成立。
+`gmp.h` 里约十二个小函数(`mpz_get_ui`、`mpn_add`、`mpn_cmp` …)有两份定义:头里一份走
+`__GMP_EXTERN_INLINE`,库里一份由带 `-D__GMP_FORCE_<fn>` 的那个 TU 提供
+(`mpz/get_ui.c`、`mpn/generic/add.c` …;那里 gmp.h 会**手工抑制** `__GMP_EXTERN_INLINE`,
+所以那份一定是普通外部定义)。这套分工成立的前提是:**头里那份永远不能变成独立符号**。
+GCC 分支用 `extern __inline__ __attribute__ ((__gnu_inline__))` 保证了这点。
 
-问题在于 `_MSC_VER` 那条分支是整串里**唯一没有** `! defined (__GMP_EXTERN_INLINE)` 守卫的
-(SunPro / SCO 两条都有)。而 clang 打 MSVC ABI 时**同时**定义 `__GNUC__` 与 `_MSC_VER`,于是它
-默默把 GCC 的答案改写成裸 `__inline`。MS 的 inline 语义下,每个包含 `gmp.h` 的 TU 都会发射一份外部
-定义,链接直接死在十个重复符号上:
+`_MSC_VER` 分支给的是裸 `__inline`,MS 语义下**每个**包含 gmp.h 的 TU 都会发射一份外部定义;
+而且它是整串厂商分支里**唯一没有** `! defined (__GMP_EXTERN_INLINE)` 守卫的(SunPro / SCO 两条都有)。
+本索引 windows 用的 clang 打 MSVC ABI **定义 `_MSC_VER` 但不定义 `__GNUC__`**(实测:先按"两个都定义"
+的假设只补 `! defined (__GNUC__)` 守卫,CI 结果一字未变),于是落到 MSVC 那条,链接死在十个重复符号:
 
 ```
 lld-link: error: duplicate symbol: __gmpz_get_ui
->>> defined at gmp.h:1793   obj/.../mpz/pprime_p.o
+>>> defined at gmp.h:1800   obj/.../mpz/pprime_p.o
 >>> defined at              obj/.../mpz/get_ui.o
 ```
 
-上游从来没碰到,因为它压根没有 MSVC ABI 的构建。补丁就是给这条分支补上邻居们已经有的守卫
-(`#if defined (_MSC_VER) && ! defined (__GNUC__)`),在别处是严格 no-op:非 windows 目标不定义
-`_MSC_VER`,真 MSVC 不定义 `__GNUC__`。生成器里对补丁点做了"必须恰好命中一次"的断言,
-上游改动了这段会直接报错而不是静默漏打。
+**解法是把这条分支改成 `static __inline`** —— 正是 GMP 自己给另外两个"extern inline 会漏出全局符号"
+的编译器(DEC C、SCO OpenUNIX)的答案。关键耦合在于 `__GMP_INLINE_PROTOTYPES`:GCC 分支会把它设为 1
+(于是发 extern 原型,与 gnu_inline 定义相容),而 MSVC 分支不设 → 后面默认 0 → **不发 extern 原型**,
+`static` 定义才不会和它打架。第一次本地模拟只换 `__GMP_EXTERN_INLINE` 不动 `__GMP_INLINE_PROTOTYPES`,
+立刻得到 `static declaration of '__gmpz_abs' follows non-static declaration` —— 忠实复现 MSVC 分支
+(两者都换)之后本地实测:非 FORCE TU 里那些符号**一个都不发射**(全内联),FORCE TU 发射唯一的全局,
+六个 TU 合并链接无重复。守卫照旧补上,这样将来若有既 GNU 兼容又设 `_MSC_VER` 的编译器,仍走
+gnu_inline 那条。
+
+生成器里对补丁点做了"必须恰好命中一次"的断言,上游改动了这段会直接报错而不是静默漏打。
 
 值得记一笔的是**首轮 windows CI 的 516 个 TU 全部编译通过**,只挂在链接 —— 也就是说
-`config.h`/`gmp.h`/内联表在 LLP64 + MSVC ABI 上本来就是对的。
+`config.h`/`gmp.h`/内联表在 LLP64 + MSVC ABI 上本来就是对的。另外由于 windows 上 `__GNUC__` 未定义,
+`config.h` 里 `HAVE_ATTRIBUTE_*` 与 `longlong.h`/`MPN_IORD_U` 的 inline asm 在那边自动全关,
+走纯 C 路径 —— 正确但比 linux/macOS 慢,这是可接受的取舍。
 
 ### 3.6 `-fPIC`
 

pkgs/c/compat.gmp.lua

@@ -931,15 +931,20 @@ typedef const __gmp_randstate_struct *gmp_randstate_srcptr;
 #endif
 
 /* Microsoft's C compiler accepts __inline */
-/* mcpp-index (compat.gmp): `&& ! defined (__GNUC__)`, which upstream's
-   neighbouring vendor branches already carry in spirit.  clang targeting the
-   MSVC ABI defines BOTH __GNUC__ and _MSC_VER; without the guard this line
-   replaces the GCC branch's gnu_inline spelling with MS inline semantics, the
-   header starts emitting an external definition of every __GMP_EXTERN_INLINE
-   function in every TU that includes it, and the link fails with ten
-   duplicate symbols against the -D__GMP_FORCE_<fn> copies. */
-#if defined (_MSC_VER) && ! defined (__GNUC__)
-#define __GMP_EXTERN_INLINE  __inline
+/* mcpp-index (compat.gmp): `static __inline`, not `__inline`, and guarded.
+   Plain `__inline` has MS inline semantics -- the header's copy of every
+   __GMP_EXTERN_INLINE function becomes an external definition in EVERY TU
+   that includes gmp.h, which collides with the out-of-line copy the
+   -D__GMP_FORCE_<fn> TU emits (ten `lld-link: duplicate symbol` errors on
+   clang targeting the MSVC ABI, which defines _MSC_VER but not __GNUC__).
+   `static __inline` is the answer GMP already gives for the other compilers
+   whose "extern inline" leaks a global (DEC C, SCO OpenUNIX): the header's
+   copy stays file-local and still inlines, while the forced TU -- where
+   gmp.h suppresses this macro by hand -- remains the single external
+   definition. The guard mirrors the SunPro/SCO branches, so a GNU-compatible
+   compiler that also sets _MSC_VER keeps the gnu_inline answer above. */
+#if defined (_MSC_VER) && ! defined (__GMP_EXTERN_INLINE)
+#define __GMP_EXTERN_INLINE  static __inline
 #endif
 
 /* Recent enough Sun C compilers want "inline" */

tools/gmp/generate_descriptor.py

@@ -301,45 +301,58 @@ def prepare(workdir):
 # --------------------------------------------------------------------------
 # One patch to gmp.h beyond configure's substitutions.
 #
-# gmp.h picks how to spell "inline" through a vendor cascade.  The GCC branch
-# gives `extern __inline__ __attribute__ ((__gnu_inline__))` -- GNU89 inline
-# semantics: the body in the header is for inlining only and is NEVER emitted
-# as a standalone symbol, because the out-of-line copy comes from the one TU
-# that compiles with -D__GMP_FORCE_<fn> (mpz/get_ui.c, mpn/generic/sub.c, ...).
-# The whole library depends on that division.
+# gmp.h picks how to spell "inline" through a vendor cascade.  About a dozen
+# small functions (mpz_get_ui, mpn_add, mpn_cmp, ...) are then defined in the
+# header under __GMP_EXTERN_INLINE, AND out of line by the one TU that
+# compiles with -D__GMP_FORCE_<fn> (mpz/get_ui.c, mpn/generic/add.c, ...).
+# That only works if the header's copy never becomes a standalone symbol.
+# The GCC branch guarantees it with
+# `extern __inline__ __attribute__ ((__gnu_inline__))`.
 #
-# The `_MSC_VER` branch further down is the only one in the cascade WITHOUT a
-# `! defined (__GMP_EXTERN_INLINE)` guard (the SunPro and SCO branches have
-# one), so on a compiler that is BOTH -- clang targeting the MSVC ABI, which
-# is what this index uses on Windows -- it silently overwrites the GCC answer
-# with plain `__inline`.  Under MS inline semantics that emits an external
-# definition in every TU that includes gmp.h, and the link dies on ten
-# duplicate symbols against the forced copies:
+# The `_MSC_VER` branch answers plain `__inline`, whose MS semantics DO emit
+# an external definition in every TU that includes the header -- and it is the
+# only branch in the cascade without a `! defined (__GMP_EXTERN_INLINE)` guard
+# (SunPro and SCO both have one).
+#
+# This index compiles Windows with clang targeting the MSVC ABI, and that
+# compiler defines _MSC_VER but NOT __GNUC__ (measured: guarding the branch on
+# `! defined (__GNUC__)` changed nothing), so it lands on the MSVC answer and
+# the link dies on ten duplicate symbols:
 #
 #   lld-link: error: duplicate symbol: __gmpz_get_ui
-#   >>> defined at gmp.h:1793  obj/.../mpz/pprime_p.o
+#   >>> defined at gmp.h:1800  obj/.../mpz/pprime_p.o
 #   >>> defined at            obj/.../mpz/get_ui.o
 #
-# Upstream never hit this because it has no MSVC-ABI build at all.  Adding the
-# same guard the neighbouring branches already carry is the minimal fix, and
-# it is a no-op everywhere else: non-Windows targets never define _MSC_VER,
-# and real MSVC never defines __GNUC__.
+# Fix: `static __inline`, which is what GMP itself answers for the other two
+# compilers whose "extern inline" leaks a global (DEC C and SCO OpenUNIX).
+# Non-forced TUs get a file-local copy that still inlines; the forced TU is
+# unaffected because gmp.h suppresses __GMP_EXTERN_INLINE there by hand
+# (`#if ! defined (__GMP_FORCE_<fn>)`), so the external definition still comes
+# from exactly one object.  The added guard keeps a future GNU-compatible
+# compiler that also sets _MSC_VER on the gnu_inline answer.
+#
+# Upstream never hit any of this: it has no MSVC-ABI build at all.
 GMP_H_PATCHES = [(
     """/* Microsoft's C compiler accepts __inline */
 #ifdef _MSC_VER
 #define __GMP_EXTERN_INLINE  __inline
 #endif
 """,
     """/* Microsoft's C compiler accepts __inline */
-/* mcpp-index (compat.gmp): `&& ! defined (__GNUC__)`, which upstream's
-   neighbouring vendor branches already carry in spirit.  clang targeting the
-   MSVC ABI defines BOTH __GNUC__ and _MSC_VER; without the guard this line
-   replaces the GCC branch's gnu_inline spelling with MS inline semantics, the
-   header starts emitting an external definition of every __GMP_EXTERN_INLINE
-   function in every TU that includes it, and the link fails with ten
-   duplicate symbols against the -D__GMP_FORCE_<fn> copies. */
-#if defined (_MSC_VER) && ! defined (__GNUC__)
-#define __GMP_EXTERN_INLINE  __inline
+/* mcpp-index (compat.gmp): `static __inline`, not `__inline`, and guarded.
+   Plain `__inline` has MS inline semantics -- the header's copy of every
+   __GMP_EXTERN_INLINE function becomes an external definition in EVERY TU
+   that includes gmp.h, which collides with the out-of-line copy the
+   -D__GMP_FORCE_<fn> TU emits (ten `lld-link: duplicate symbol` errors on
+   clang targeting the MSVC ABI, which defines _MSC_VER but not __GNUC__).
+   `static __inline` is the answer GMP already gives for the other compilers
+   whose "extern inline" leaks a global (DEC C, SCO OpenUNIX): the header's
+   copy stays file-local and still inlines, while the forced TU -- where
+   gmp.h suppresses this macro by hand -- remains the single external
+   definition. The guard mirrors the SunPro/SCO branches, so a GNU-compatible
+   compiler that also sets _MSC_VER keeps the gnu_inline answer above. */
+#if defined (_MSC_VER) && ! defined (__GMP_EXTERN_INLINE)
+#define __GMP_EXTERN_INLINE  static __inline
 #endif
 """)]