anyways, i am building llvm native musl nixos rn (cc Tristan Ross , its mostly on top of your PR). Findings so far:

• a bunch of things are pulling libgcc via non-strict deps and fail on llvm native
• bison is bork
• a couple tests are bork (mpfr, fmt)
• pfm in llvm causes issues (either disable in llvm, or allow it if cc for stdenv is llvm)
• python3 minimal fails (not allowed to refer to libunwind)

there is probably more

[root@asterinas:~]# cat /proc/version
Linux version 5.13.0 (root@kiryca) (rustc 1.94.0-nightly) #1 Wed Dec 17 02:57:56 UTC 2025

Hi all,
I'm trying to write a derivation for an embedded C++ microcontroller application.
For this, I must use a copy of libstdc++ with something like --enable-cxx-flags='-fno-exceptions -fno-rtti' (docs here).
Is there a way to configure Nixpkgs to use these settings for libstdc++ for the host platform? Something to add to the following, for example?

import <nixpkgs> {
  crossSystem = {
    config = "arm-none-eabihf";
    libc = "newlib-nano";
    gcc = {
      cpu = "cortex-m7";
      fpu = "fpv5-d16";
      float-abi = "hard";
      thumb = true;
    };
    rust.rustcTarget = "thumbv7em-none-eabihf";
    isStatic = true;
  }
}

I'm looking at something like this patch to Nixpkgs at the moment, but hopefully there's an existing mechanism I'm yet to discover

--- a/pkgs/development/compilers/gcc/common/platform-flags.nix
+++ b/pkgs/development/compilers/gcc/common/platform-flags.nix
@@ -17,6 +17,7 @@ lib.concatLists [
   (lib.optional (p ? fpu) "--with-fpu=${p.fpu}")
   (lib.optional (p ? float) "--with-float=${p.float}")
   (lib.optional (p ? mode) "--with-mode=${p.mode}")
+  (lib.optional (p ? enableCxxFlags) "--enable-cxx-flags=${builtins.concatStringsSep ''\ '' p.enableCxxFlags}")
   (lib.optionals targetPlatform.isPower64
     # musl explicitly rejects 128-bit long double on
     # powerpc64; see musl/arch/powerpc64/bits/float.h