Another problem with Windows target and `compiler_builtins`

Cloned latest Rust，ran ./x build with the following bootstrap.toml：

profile = "dist"  
change-id = 148636

[build]
build-stage = 1
target = ["x86_64-unknown-linux-gnu", "x86_64-pc-windows-gnu"]

[rust]
codegen-backends = ["gcc"]  

[target.x86_64-pc-windows-gnu]
cc = "x86_64-w64-mingw32-gcc"
cxx = "x86_64-w64-mingw32-g++"
ar = "x86_64-w64-mingw32-ar"
ranlib = "x86_64-w64-mingw32-ranlib"
linker = "x86_64-w64-mingw32-gcc"

and resulted in:

Building stage1 library artifacts{alloc, compiler_builtins, core, panic_abort, panic_unwind, proc_macro, rustc-std-workspace-core, std, std_detect, sysroot, test, unwind} (stage1:x86_64-unknown-linux-gnu -> stage1:x86_64-pc-windows-gnu)
.......(lots of messages)
/tmp/libgccjit-yPspPh/fake.s: Assembler messages:
/tmp/libgccjit-yPspPh/fake.s:51: fatal error: bad .section directive: want a,l,w,x,M,S,G,T in string
libgccjit.so: error: : error invoking gcc driver
error: failed to build archive at `(rust project dir)/build/x86_64-unknown-linux-gnu/stage1-std/x86_64-pc-windows-gnu/release/deps/libcompiler_builtins-4c849587cc5e14fa.rlib`: failed to open object file: no such file or directory (os error 2)

I've learned the bad .section directive error may be caused by invoking a Linux targeted gcc against a Windows .s file, but I don't understand the detail since I've specified linker = "x86_64-w64-mingw32-gcc" already.

By the way, How to keep temporary file(s) of compilation such as /tmp/libgccjit-yPspPh/fake.s ? They simply vanished when I am to inspect them.

Start by reproducing ./x build with the reported bootstrap.toml and inspect the compiler_builtins failure for the x86_64-pc-windows-gnu target. Investigate how the temporary fake.s file is assembled and whether the configured linker is used; done means identifying the cause of the bad section directive and defining a verified fix or clear reproduction details.