Re: [PATCH v4] kbuild: host: use single executable for rustc -C linker

[Nicolas Schier <nsc@kernel.org>]

On Tue, Mar 17, 2026 at 02:20:21PM +0300, Mohamad Alsadhan wrote:
> rustc's -C linker= option expects a single executable path. When
> HOSTCC contains a wrapper (e.g. "ccache gcc"), passing
> `-Clinker=$(HOSTCC)` results in the shell splitting the value into
> multiple words, and rustc interprets the additional word as an
> input filename:
> 
>   error: multiple input filenames provided ...
> 
> Generate a small wrapper script and pass it to -Clinker e.g.
> ```
> exec sh -c 'exec "$0" "$@"' ccache gcc "$@"
> ```
> 
> This fix should be general enough to address most if not all cases
> (incl. wrappers or subcommands) and avoids surprises of simpler fixes
> like just defaulting to gcc.
> 
> This avoids passing the user command as an environment variable as
> that would be more challenging to trace and debug shell expansions.
> 
> Link: https://github.com/Rust-for-Linux/linux/issues/1224
> Suggested-by: Yoann Congal <yoann.congal@smile.fr>
> Signed-off-by: Mohamad Alsadhan <mo@sdhn.cc>
> ---
> v3 -> v4:
>   - Use filechk instead of if_changed macro to regenerate script
>   - Remove trailing space at EOL
> 
> v2 -> v3:
>   - Scrap previous hacky approaches (e.g. using lastword) and go with
>     a proper fix which turned out not that complex to implement.
>     Apologies Gary, I should have listened to you earlier :/
> 
> v1 -> v2:
>   - Rename HOSTRUSTC_LINKER to HOSTRUSTC_LD for consistency
>   - Introduce explicit HOSTRUSTC_LD override
>   - Warn when falling back due to multi-argument HOSTCC
>   - Error out if a user-specified HOSTRUSTC_LD is not an executable
> 
> v1: https://lore.kernel.org/all/20260225102819.16553-1-mo@sdhn.cc/
> v2: https://lore.kernel.org/all/20260227132713.23106-1-mo@sdhn.cc/
> v3: https://lore.kernel.org/all/20260312002852.11292-1-mo@sdhn.cc/
> ---
>  scripts/Makefile.host | 22 ++++++++++++++++++++--
>  1 file changed, 20 insertions(+), 2 deletions(-)
> 
> diff --git a/scripts/Makefile.host b/scripts/Makefile.host
> index c1dedf646..e41753828 100644
> --- a/scripts/Makefile.host
> +++ b/scripts/Makefile.host
> @@ -87,11 +87,29 @@ hostcxx_flags  = -Wp,-MMD,$(depfile) \
>                   $(KBUILD_HOSTCXXFLAGS) $(HOST_EXTRACXXFLAGS) \
>                   $(HOSTCXXFLAGS_$(target-stem).o)
>  
> +# rustc's `-Clinker=` expects a single executable path, not a command line.
> +# `HOSTCC` may be a multi-word command when wrapped (e.g. "ccache gcc"), which
> +# would otherwise be split by the shell and mis-parsed by rustc.
> +# To work around this, we generate a wrapper script that forwards arguments to
> +# `HOSTRUSTC_LD` so that such commands can be used safely.
> +#
> +# Set `HOSTRUSTC_LD` for a different rustc linker command than `HOSTCC`
> +HOSTRUSTC_LD ?= $(HOSTCC)
> +define filechk_rustc-wrapper
> +	printf "%s\n" \
> +		'#!/bin/sh' \
> +		'exec sh -c '\''exec "$$0" "$$@"'\'' $(HOSTRUSTC_LD) "$$@"'
> +endef

This printf-based solution needs some more tweaking to also support
complex commands in HOSTRUSTC_LD with quoted spaces.  A straight forward
way could be:

	'exec sh -c '\''exec "$$0" "$$@"'\'' $(call escsq,$(HOSTRUSTC_LD)) "$$@"'

to escape single quotes within HOSTRUSTC_LD.  Then the complex example
from v3 works again (make HOSTRUSTC_LD="env 'CCACHE_DIR=/tmp/my cache' ccache gcc").

With the quoting fixed:
Reviewed-by: Nicolas Schier <nsc@kernel.org>
Tested-by: Nicolas Schier <nsc@kernel.org>



I am still not seeing a benefit in calling usage of 'exec' and 'sh -c'.
If we'd skip both, there would be no need to use 'env' for environment
updates:

	'$(call escsq,$(HOSTRUSTC_LD)) "$$@"'

But I may be missing something here.

Kind regards,
Nicolas