Re: [PATCH bpf v3 01/15] selftests/bpf: Pass through build flags to bpftool and resolve_btfids

[Ihor Solodrai <ihor.solodrai@linux.dev>]

On 2/20/26 2:57 PM, bot+bpf-ci@kernel.org wrote:
>> diff --git a/tools/bpf/resolve_btfids/Makefile b/tools/bpf/resolve_btfids/Makefile
>> index 1733a6e93a07..ef083602b73a 100644
>> --- a/tools/bpf/resolve_btfids/Makefile
>> +++ b/tools/bpf/resolve_btfids/Makefile
>> @@ -65,6 +65,9 @@ $(BPFOBJ): $(wildcard $(LIBBPF_SRC)/*.[ch] $(LIBBPF_SRC)/Makefile) | $(LIBBPF_OU
>>  LIBELF_FLAGS := $(shell $(HOSTPKG_CONFIG) libelf --cflags 2>/dev/null)
>>  LIBELF_LIBS  := $(shell $(HOSTPKG_CONFIG) libelf --libs 2>/dev/null || echo -lelf)
>>
>> +ZLIB_LIBS  := $(shell $(HOSTPKG_CONFIG) zlib --libs 2>/dev/null || echo -lz)
>> +ZSTD_LIBS  := $(shell $(HOSTPKG_CONFIG) libzstd --libs 2>/dev/null || echo -lzstd)
> 
> The "|| echo -lzstd" fallback makes zstd an unconditional link
> dependency. On systems where libzstd-dev is not installed, pkg-config
> fails and the fallback provides -lzstd, but the linker cannot find the
> library and the build fails.
> 
> bpftool handles the same transitive dependency conditionally via a
> feature test:
> 
> tools/bpf/bpftool/Makefile:
>     ifeq ($(feature-libelf-zstd),1)
>     LIBS += -lzstd
>     LIBS_BOOTSTRAP += -lzstd
>     endif
> 
> Would it be safer to drop the fallback here so ZSTD_LIBS is empty
> when zstd is not available?
> 
>     ZSTD_LIBS := $(shell $(HOSTPKG_CONFIG) libzstd --libs 2>/dev/null)
> 
> This would still link zstd when present (needed for the static linking
> case where libelf.a pulls in ZSTD symbols from elf_compress.o), but
> avoid breaking dynamic builds on systems without libzstd-dev.

I tried to reproduce the situation like follows:

  $ git remote -v
  origin	git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf.git (fetch)
  origin	git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf.git (push)

  $ git log --oneline -n1
  8bf22c33e7a1 (HEAD -> master, origin/master, origin/HEAD) Merge tag 'net-7.0-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net

  $ b4 shazam 20260220182011.802116-1-ihor.solodrai@linux.dev   # apply strscpy series
  $ b4 shazam 20260220222604.1155148-1-ihor.solodrai@linux.dev  # apply ASAN series

  $ podman pull debian:latest
  $ podman run -it --rm -v $(pwd):/linux debian:latest /bin/bash

  # in the container
  $ apt update -y && apt install -y make gcc python3 libelf-dev
  # note: no libzstd-dev

  $ cd /linux/tools/bpf/resolve_btfids/
  $ make -j8
  # build is successful

  $ ldd resolve_btfids
	linux-vdso.so.1 (0x00007f3eecc81000)
	libelf.so.1 => /lib/x86_64-linux-gnu/libelf.so.1 (0x00007f3eecbe2000)
	libz.so.1 => /lib/x86_64-linux-gnu/libz.so.1 (0x00007f3eecbc2000)
	libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f3eec9cd000)
	libzstd.so.1 => /lib/x86_64-linux-gnu/libzstd.so.1 (0x00007f3eec903000)
	/lib64/ld-linux-x86-64.so.2 (0x00007f3eecc83000)


So, while AI may be technically correct that the build will fail if
libelf is installed and libzstd is not, it seems a far-fetched
situation, given that libelf depends on libzstd.

I think we can leave the default -lzstd to have an explicit
dependency in the Makefile.


> 
> [ ... ]
> 
> 
> ---
> AI reviewed your patch. Please fix the bug or email reply why it's not a bug.
> See: https://github.com/kernel-patches/vmtest/blob/master/ci/claude/README.md
> 
> CI run summary: https://github.com/kernel-patches/bpf/actions/runs/22243753316
> 
> AI-authorship-score: low
> AI-authorship-explanation: Human-characteristic typo in commit message, natural lore discussion with incremental debugging, and standard Makefile pattern reuse all indicate human authorship.
> issues-found: 1
> issue-severity-score: low
> issue-severity-explanation: Unconditional zstd link fallback in resolve_btfids Makefile can break builds on systems without libzstd-dev, but the affected scenario is narrow and limited to the selftests build system.