refactor(example): 板子的内存布局不再由这个例子陈述

它此前带着一份链接脚本 —— OpenSBI 把控制权交到哪里、多少栈、堆在哪 —— 挂在一个
关于目标的条件后面。那个条件是诚实的部分:它们是关于**一台机器**的陈述,不是关于
这个程序的。

⭐ 而这恰恰说明它们也不属于这里。openkal-opensbi 是知道那台机器的包,它现在用自己的
构建程序供给这份地图,到达每个使用者的链接线 —— 与它供给 kal_* 定义的方式相同。
于是这个目录所主张的「同一份源码、不同的机器、什么都不用改」对构建文件也成立。

⚠️ 两份并存时实测(板级事实**可传递**地到达使用者,而这个例子是使用者):

    ld.lld: error: section .eh_frame file range overlaps with .debug_str_offsets
    ld.lld: error: section .debug_str file range overlaps with .eh_frame_hdr

两份脚本都被应用了,而诊断一个字都没提「有两份」。

复验四个目标全部产出;裸机(qemu+OpenSBI)、Linux、Windows(wine)三个跑通,
输出逐字相同。

Author: speak-agent

examples/same-source/build.mcpp

@@ -1,26 +1,30 @@
 import mcpp;
-#include <cstring>
 
-int main() {
-    // ⭐⭐ THE LINKER SCRIPT IS FOR THE MACHINE WITH NO OPERATING SYSTEM, AND
-    // ONLY FOR IT. THAT CONDITION IS WHAT MAKES THIS ONE PROJECT INSTEAD OF TWO.
-    //
-    // The claim the design document makes is that a program above openkal is
-    // built without regard to which implementation is beneath — same source,
-    // different machine, nothing edited. A directory that could only be built
-    // one way would illustrate that claim rather than demonstrate it.
-    //
-    // So: `mcpp run` puts this on the host over openkal-linux, and
-    // `mcpp run --target riscv64-none-elf` puts the SAME source on riscv64 over
-    // OpenSBI. Both print the same four lines. The only thing that differs is
-    // the memory layout, which is a statement about the machine — and it is
-    // stated here, once, behind the condition that says which machine.
-    if (std::strcmp(mcpp::target_os(), "none") == 0) {
-        // A relative path in `ldflags` would resolve against the build
-        // directory; `link-script` resolves against the package root.
-        mcpp::link_script("link.ld");
-    }
-    mcpp::rerun_if_changed("link.ld");
-    mcpp::rerun_if_env_changed("MCPP_TARGET_OS");
-    return 0;
-}
+// ⚠️ THIS FILE STATES NOTHING, AND ITS EMPTINESS IS WHAT THE EXAMPLE IS ABOUT.
+//
+// Until 2026-08-23 it carried a linker script for the machine with no operating
+// system — where OpenSBI hands control over, how much stack, where the heap is
+// — behind a condition on the target. The condition was the honest part: those
+// are statements about a MACHINE and not about this program.
+//
+// ⭐ Which is exactly why they do not belong here either. `openkal-opensbi` is
+// the package that knows that machine, and it supplies the map the same way it
+// supplies its `kal_*` definitions: through its own build program, reaching
+// every consumer's link line. A program adds nothing and gets a layout that
+// works, and the claim this directory makes — same source, different machine,
+// nothing edited — becomes true of the build files as well as of the source.
+//
+// ⚠️ Measured while both copies existed, because a board fact reaches its
+// consumers TRANSITIVELY and this example is a consumer:
+//
+//     ld.lld: error: section .eh_frame file range overlaps with
+//                    .debug_str_offsets
+//     ld.lld: error: section .debug_str file range overlaps with .eh_frame_hdr
+//
+// Two scripts, both applied, and a diagnostic that says nothing about there
+// being two. ⇒ A program that also states a board fact states it twice.
+//
+// The file remains rather than being deleted: a build program is how this
+// example would say anything at all, and its absence would read as "this was
+// never considered" rather than "there is nothing to say".
+int main() { return 0; }