Native compiler wrong multiarch

[concatime]

Hi. I’ve built a cross compiler, x86_64-linux-musl as target, then a native compiler using the cross one. When running gcc -print-multiarch on the native, I get i386-linux-gnu.

Step to reproduce

Create a config.mak with:

TARGET := x86_64-linux-musl

ifneq ($(NATIVE),)
	OUTPUT := /foo
	TOP := $(abspath $(dir $(lastword $(MAKEFILE_LIST))))
	PATH := $(TOP)/output/bin:/usr/bin
endif

MAKEFLAGS := --jobs=$(shell nproc)

COMMON_CONFIG += --disable-nls
GCC_CONFIG += --enable-languages=c,c++
GCC_CONFIG += --disable-libquadmath --disable-decimal-float
GCC_CONFIG += --disable-multilib

then

make
make install
make NATIVE=1
sudo make NATIVE=1 install
sudo ln -fs libc.so /foo/lib/ld-musl-x86_64.so.1 # https://github.com/richfelker/musl-cross-make/issues/82
sudo ln -s -t /lib $_
/foo/bin/gcc -print-multiarch

I discovered this bug when building python (./configure --build=x86_64-linux-musl). Here is the relevant part from python’s configure.

MULTIARCH=$($CC --print-multiarch 2>/dev/null)
[…]
if test x$PLATFORM_TRIPLET != x && test x$MULTIARCH != x; then
  if test x$PLATFORM_TRIPLET != x$MULTIARCH; then
    as_fn_error $? "internal configure error for the platform triplet, please file a bug report" "$LINENO" 5
  fi
elif test x$PLATFORM_TRIPLET != x && test x$MULTIARCH = x; then
  MULTIARCH=$PLATFORM_TRIPLET
fi

In the snippet, PLATFORM_TRIPLET is x86_64-linux-musl and MULTIARCH is i386-linux-gnu.

That being said, gcc -dumpmachine does print the right x86_64-linux-musl.

From https://gcc.gnu.org/install/configure.html:

--enable-multiarch: Specify whether to enable or disable multiarch support. The default is to check for glibc start files in a multiarch location, and enable it if the files are found. The auto detection is enabled for native builds, and for cross builds configured with --with-sysroot, and without --with-native-system-header-dir.

[richfelker]

Based on the description in the GCC documentation, I don't see why multiarch support would be used without explicitly requesting it. Does adding --disable-multiarch to GCC_CONFIG fix the problem?

[concatime]

I was building with --disable-multiarch while filling this issue. Note, the issue is only present in the native compiler. gcc -print-multiarch on the cross one returns nothing. And yes, --disable-multiarch fixes the issue. It now returns nothing just like the cross one.

[concatime]

~~Also, why have you removed --disable-multilib in 2954bf4 ?~~

this is now mandatory and always passed to configure

~~Details ?~~ EDIT: (7a6ed3a) Ignore this comment.

[concatime]

This bug can be solved with an argument to gcc. I don’t know if the argument should be used or not automatically by litecross when making a native compiler. I will close it for now.

[richfelker]

I don't think this should be closed until --disable-multiarch is committed.

[jheaff1]

I have also hit this issue when trying to build Python3 using x86_64-linux-musl-native.

Is there any update on a fix?