| author | 2026-09-06 17:23:50 +0700 | |
|---|---|---|
| committer | 2026-09-06 19:56:37 +0700 | |
| commit | 6672410d04e47b3b455b1ca40b7320e65af8fb2c (patch) | |
| tree | 1ba5e16344a7b0ddb3e20f8bef111e67d1d82c59 /config/no-debug-info.config | |
| download | kernel-surface-6672410d04e47b3b455b1ca40b7320e65af8fb2c.tar.gz kernel-surface-6672410d04e47b3b455b1ca40b7320e65af8fb2c.zip | |
initial commit
Diffstat (limited to 'config/no-debug-info.config')
| -rw-r--r-- | config/no-debug-info.config | 21 |
1 files changed, 21 insertions, 0 deletions
diff --git a/config/no-debug-info.config b/config/no-debug-info.config new file mode 100644 index 0000000..502fa50 --- /dev/null +++ b/config/no-debug-info.config @@ -0,0 +1,21 @@ +# Config fragment that turns off DWARF and BTF generation. +# +# The package throws the debug info away anyway: the spec sets +# `%global debug_package %{nil}` and installs modules with +# INSTALL_MOD_STRIP=1, so no -debuginfo subpackage is ever produced. +# Compiling it only costs build RAM, disk and time, which is what constrains +# the 6 CPU / 8 GB arm64 builder. +# +# What this gives up: BTF is the one part of the debug info that would have +# shipped, in vmlinux and in the modules. Without it there is no +# /sys/kernel/btf/vmlinux, so eBPF CO-RE tooling (bpftrace, libbpf programs) +# does not work, and CONFIG_BPF_LSM turns itself off because it depends on +# CONFIG_DEBUG_INFO_BTF - which in turn disables systemd unit settings built +# on the BPF LSM, such as RestrictFileSystems=. +CONFIG_DEBUG_INFO_NONE=y +# CONFIG_DEBUG_INFO is not set +# CONFIG_DEBUG_INFO_DWARF_TOOLCHAIN_DEFAULT is not set +# CONFIG_DEBUG_INFO_DWARF4 is not set +# CONFIG_DEBUG_INFO_DWARF5 is not set +# CONFIG_DEBUG_INFO_BTF is not set +# CONFIG_DEBUG_INFO_BTF_MODULES is not set |