On Tue, 11 Feb 2025, WangYuli wrote:
> During make module_install, the need_compiler variable becomes 0,
> so Makefile.compiler isn't included.
>
> This results in call cc-option-yn returning nothing.
>
> Add a check for KBUILD_SYM32=n to avoid the
> "CONFIG_CPU_DADDI_WORKAROUNDS unsupported without -msym32" error
> when KBUILD_SYM32 is unset (meaning it's not 'y' or 'n').
Jeez, just wrap it into `ifdef need-compiler' as I told you previously.
This stuff isn't used with `make modules_install', none of the compiler
flags matter as the compiler isn't ever called, which is obvious from
Makefile.compiler not being included in the first place. I only did not
do it with commit a79a404e6c22 ("MIPS: Fix CONFIG_CPU_DADDI_WORKAROUNDS
`modules_install' regression"), because I was unaware about this LLVM's
limitation and GCC version requirements at the time implied the presence
of this feature at all times.
See also commit 4fe4a6374c4d ("MIPS: Only fiddle with CHECKFLAGS if
`need-compiler'") for a similar change from the same series[1].
Also need-compiler is nil rather than 0 in the relevant case.
Please also double-check your change description before posting, at least
when it comes to referring to codebase artifacts:
s/module_install/modules_install/
s/need_compiler/need-compiler/
This is stuff people may be grepping for.
References:
[1] "MIPS: Fix build issues from the introduction of `need-compiler'",
<https://lore.kernel.org/r/alpine.DEB.2.21.2307180025120.62448@xxxxxxxxxxxxxxxxx/>
Maciej
