Hi,
On Wed, Jan 26, 2022 at 01:47:42AM +0100, Ben Hutchings wrote:
> Control: tag -1 patch
>
> The patch add_cmake_libbpf_enabled_option.patch is indeed the problem,
> because it doesn't consistently use the same macro name. The fix is
> below.
>
> If you're not able to upload with this today, can I please NMU to
> unblock linux?
>
> Ben.
>
> --- a/debian/patches/add_cmake_libbpf_enabled_option.patch
> +++ b/debian/patches/add_cmake_libbpf_enabled_option.patch
> @@ -5,6 +5,8 @@ Non-linux systems also can make good use of these tools. This
> patch
> adds option LIBBPF_ENABLED (default: on) so to opt-out BPF when needed.
>
> Signed-off-by: Domenico Andreoli <domenico.andre...@linux.com>
> +[bwh: Only check the macro HAVE_BPF, not HAVE_BTF]
> +Signed-off-by: Ben Hutchings <b...@debian.org>
> ---
> CMakeLists.txt | 61
> +++++++++++++++++++++++++++++++++------------------------
> dwarves.c | 2 ++
> @@ -193,7 +195,7 @@ Index: b/pahole.c
> cu__fprintf_ptr_table_stats_csv(cu, stderr);
> }
>
> -+#ifdef HAVE_BTF
> ++#ifdef HAVE_BPF
> if (btf_encode) {
> static pthread_mutex_t btf_lock = PTHREAD_MUTEX_INITIALIZER;
>
> @@ -241,7 +243,7 @@ Index: b/pahole.c
> type_instance__delete(header);
> header = NULL;
>
> -+#ifdef HAVE_BTF
> ++#ifdef HAVE_BPF
> if (btf_encode) {
> err = btf_encoder__encode(btf_encoder);
> if (err) {
> @@ -257,7 +259,7 @@ Index: b/pahole.c
> #ifdef DEBUG_CHECK_LEAKS
> cus__delete(cus);
> structures__delete();
> -+#ifdef HAVE_BTF
> ++#ifdef HAVE_BPF
> btf__free(conf_load.base_btf);
> conf_load.base_btf = NULL;
> #endif
Tested against our current master (5.16.2-1~exp1) and fixes the issue.
Tested-by: Salvatore Bonaccorso <car...@debian.org>
Regards,
Salvatore
