Re: [gentoo-dev] [PATCH] Transitional changes to the kernel-2 eclass to support future CPU OPT

[Ulrich Mueller]

>>>>> On Thu, 29 Apr 2021, mpagano  wrote:

> --- a/eclass/kernel-2.eclass
> +++ b/eclass/kernel-2.eclass
> @@ -1241,8 +1241,32 @@ unipatch() {
>  			local GCC_MAJOR_VER=$(gcc-major-version)
>  			local GCC_MINOR_VER=$(gcc-minor-version)
>
> -			# optimization patch for gcc < 8.X and kernel > 4.13
> -			if kernel_is ge 4 13 ; then
> +			# this section should be the target state to handle the cpu opt
> +			# patch for kernels > 4.19.189, 5.4.115, 5.10.33 and 5.11.17,
> +			# 5.12.0 and gcc >= 9  The patch now handles the
> +			# gcc version enabled on the system through the Kconfig file as
> +			# 'depends'. The legacy section can hopefully be retired in the future
> +			# Note the patch for 4.19-5.8 version are the same and the patch for
> +			# 5.8+ version is the same
> +			# eventually we can remove everything except the gcc ver <9 check
> +			# based on stablization, time, kernel removals or a combo of all three
> +			if ((kernel_is eq 4 19 && kernel_is gt 4 19 189) ||
> +				(kernel_is eq 5 4 && kernel_is gt 5 4 115) ||
> +				(kernel_is eq 5 10 && kernel_is gt 5 10 33) ||
> +				(kernel_is eq 5 11 && kernel_is gt 5 11 17) ||
> +				(kernel_is eq 5 12 && kernel_is gt 5 12 0)); then

Looks like the outermost pair of parentheses ( ) isn't needed here.

Also, when writing nested parentheses without a space, bash may
sometimes (but not always!) interpret them as an arithmetic expression.
This can cause unexpected results:

   $ ((true) && (true)); echo $?
   0
   $ ( (true && true) ); echo $?
   0
   $ ((true && true)); echo $?
   1
   $ true=42
   $ ((true && true)); echo $?
   0

> [...]