6.10.0 - Odd Compile Errors Under LLVM18

[ThisNekoGuy]

Describe the Bug:

• OS: Gentoo (LLVM18-built)
• Running Kernel: 6.10.9-tkg-eevdf-gentoo-llvm18-generic_v3
• Logs.zip

Description:
Yesterday, I compiled version 6.10.9 and installed it just fine. Today, with the same patches, same compile options, and same configuration (I touched nothing except changing the kernel version number to 6.10.10), I get 20 errors now when attempting to compile. Even if I switch the kernel version back to 6.10.9, set the config fragment to pick up the running tkg kernel, rm -rf the kernel git repo, and git pull the linux-tkg repo.

I've got absolutely no idea what suddenly changed overnight for LLVM to suddenly claim that there are 20 uses of "undeclared identifiers".

User Patch Sources:
For clarification on what user patches were used, it was primarily Gentoo's official patches:

• https://dev.gentoo.org/~mpagano/genpatches/trunk/6.10

(I also wrote a script for grabbing them, since they're hassle to grab individually)

Quick Downloader:

#!/bin/bash

# Get the script path
SCRIPT_DIR="$0"

# Check if the path is relative and convert it to absolute
if [ "$(printf "%s" "${SCRIPT_DIR}" | cut -c 1)" = "/" ]; then
    # Absolute path; trim any potentially trailing filename name off
    SCRIPT_DIR="$(echo "${SCRIPT_DIR}" | xargs -I [] dirname [])"
else
    # Relative path, convert to absolute
    abs_script_path="${PWD}/${SCRIPT_DIR}"
    # Extract the directory; second step is necessary to remove a trailing '.'
    SCRIPT_DIR="$(dirname "${abs_script_path}")"
    SCRIPT_DIR=$(echo "${SCRIPT_DIR}" | xargs -I [] dirname [])
fi

DOWNLOADER=""
if [ -x /usr/bin/wget2 ]; then
    DOWNLOADER="wget2"
else
    DOWNLOADER="wget"
fi

# Function to download a single file
download_patch() {
    local url="${1}"
    local filename="${url##*/}"
    local new_extension=".mypatch"  # Change this to your desired extension

    # Remove the existing .patch extension and add the new one
    local new_filename="${filename%.patch}${new_extension}"

    echo "Downloading ${new_filename}..."

    ${DOWNLOADER} "$url" -O "${SCRIPT_DIR}/${new_filename}" || return
}

# KERNEL_VERSION=$(basename "${SCRIPT_DIR}" | cut -f 1 -d 'x' --complement | cut -f 1 -d '-')
# if [ -n ${KERNEL_VERSION} ] && [ ${KERNEL_VERSION} -gt 0 ]; then
#     ## TODO
#     # Problem: linux-tkg userpatch subfolders are named like '610'; which is an issue for determining the currect major and minor version dynamically on a per-folder basis
#     # and changing the genpatches URL to follow suit.
#     # The current method is just to copy this script to the respective folder and change the values here to match.
# fi

# Capture URLs into an array
# shellcheck disable=SC2207
# shellcheck disable=SC1001
patch_names=($(${DOWNLOADER} -O - https://dev.gentoo.org/~mpagano/genpatches/trunk/6.10/ | grep img | grep href | grep \.patch | grep \: | grep -v -E "[A-Za-z0-9]+\\.[A-Za-z0-9]+\\.[A-Za-z0-9]+\\.patch" | cut -f 1 -d '>' --complement | cut -f 1 -d '"' --complement | cut -f 1 -d '"'))

# Download each file
for patch in "${patch_names[@]}"; do
    download_patch "https://dev.gentoo.org/~mpagano/genpatches/trunk/6.10/${patch}" || return
done

printf "/nDone!"

for patch in "${SCRIPT_DIR}"/*; do
    case "${patch}" in
        *"_enable-cpu-optimizations-universal.mypatch"|*"_BMQ-and-PDS-"*".mypatch")
            rm "${patch}" || return
            ;;
        *)
            :
            ;;
    esac
done

And just a few of CachyOS's:

• https://github.com/CachyOS/kernel-patches/tree/master/6.10

Regardless, though, I can't imagine what could be causing this if nothing changed between yesterday and today.
I didn't even run a system update, either.

[AdelKS]

Have you tried without any user patches just to confirm ?

[Tk-Glitch]

The only thing that comes to mind is Graysky2's cpu opt patch. Since it's external to this repo it gets updated outside of our control and stuff might change/break. There were pretty large changes recently regarding that patch. First about 3 weeks ago, when the issue happened, then up to last week where it seemingly stabilized. Did it get fixed along the way @ThisNekoGuy ?

[ThisNekoGuy]

I haven't checked yet, but I can try again tomorrow @Tk-Glitch

[ThisNekoGuy]

@Tk-Glitch Sorry, I think I had responded when I was half-asleep before and I forgot;
I'm building 6.10.13 right now with the same configuration (did have to update the generic_v3 value to match the new config file update handling that) and it seems to be going fine... mostly:

. . .
LD [M]  drivers/gpu/drm/amd/amdgpu/amdgpu.o
  AR      drivers/gpu/built-in.a
  AR      drivers/built-in.a
  AR      built-in.a
  AR      vmlinux.a
  GEN     .tmp_remove-ctf.lds
ld.lld: error: no input files
  LD      vmlinux.o
  OBJCOPY modules.builtin.modinfo
  GEN     modules.builtin
  GEN     modules.builtin.objs
  MODPOST Module.symvers
  CC      .vmlinux.export.o
  UPD     include/generated/utsversion.h
  CC      init/version-timestamp.o
  KSYMS   .tmp_vmlinux0.kallsyms.S
  AS      .tmp_vmlinux0.kallsyms.o
  LD      .tmp_vmlinux1
  BTF     .tmp_vmlinux1.btf.o
  NM      .tmp_vmlinux1.syms
  KSYMS   .tmp_vmlinux1.kallsyms.S
  AS      .tmp_vmlinux1.kallsyms.o
  LD      .tmp_vmlinux2
  NM      .tmp_vmlinux2.syms
  KSYMS   .tmp_vmlinux2.kallsyms.S
  AS      .tmp_vmlinux2.kallsyms.o
  LD      vmlinux
  BTFIDS  vmlinux
  NM      System.map
  SORTTAB vmlinux
  RELOCS  arch/x86/boot/compressed/vmlinux.relocs
  RSTRIP  vmlinux

The errors from before aren't an issue anymore, but I did get an lld error for some reason?
The build continues, regardless, for some reason which surprises me because I would have expected a kernel linkage error to just halt compilation... Though, I'm not sure if I should use the kernel if it failed to link?

[Tk-Glitch]

Strange that it didn't stop, indeed 🤔 And the error itself is also weird.