[Buildroot] [PATCHv6 4/5] package/bpftrace: new package

[Arnout Vandecappelle <arnout@mind.be>]

On 20/12/2020 16:12, Romain Naour wrote:
[snip]
> diff --git a/package/bpftrace/0001-cmake-link-ast-resources-parser-as-static-libraries.patch b/package/bpftrace/0001-cmake-link-ast-resources-parser-as-static-libraries.patch
> new file mode 100644
> index 0000000000..d99b97c3a4
> --- /dev/null
> +++ b/package/bpftrace/0001-cmake-link-ast-resources-parser-as-static-libraries.patch
> @@ -0,0 +1,71 @@
> +From 960d2b69975462e445733482ac8d6ea319d3c5b2 Mon Sep 17 00:00:00 2001
> +From: Romain Naour <romain.naour@gmail.com>
> +Date: Thu, 5 Nov 2020 01:19:54 +0100
> +Subject: [PATCH] cmake: link ast, resources parser as static libraries

 Why is this needed?

 Has the patch been sent upstream?

> +
> +Signed-off-by: Romain Naour <romain.naour@gmail.com>

[snip]]
> diff --git a/package/bpftrace/Config.in b/package/bpftrace/Config.in
> new file mode 100644
> index 0000000000..f627725afe
> --- /dev/null
> +++ b/package/bpftrace/Config.in
> @@ -0,0 +1,40 @@
> +config BR2_PACKAGE_BPFTRACE_ARCH_SUPPORTS
> +	bool
> +	default y if BR2_aarch64 || BR2_aarch64_be
> +	default y if BR2_x86_64

 You can add

	depends on BR2_PACKAGE_LLVM_ARCH_SUPPORTS # bcc

here as well. The point of _ARCH_SUPPORTS is to make it easier for dependent
packages to express the arch dependency, so that should include recursive arch
dependencies.

> +
> +config BR2_PACKAGE_BPFTRACE
> +	bool "bpftrace"
> +	depends on BR2_PACKAGE_BPFTRACE_ARCH_SUPPORTS
> +	depends on BR2_PACKAGE_LLVM_ARCH_SUPPORTS # bcc
> +	depends on BR2_TOOLCHAIN_USES_GLIBC # bcc
> +	depends on BR2_TOOLCHAIN_GCC_AT_LEAST_4_8 # bcc, clang

 We write this as # bcc -> clang


> +	depends on BR2_INSTALL_LIBSTDCPP # bcc, clang
> +	select BR2_PACKAGE_BCC
> +	select BR2_PACKAGE_LIBBPF
> +	# LLVM RTTI mendatory: https://github.com/iovisor/bpftrace/issues/1156
> +	select BR2_PACKAGE_LLVM_RTTI

 Again, I prefer an explicit select BR2_PACKAGE_LLVM

> +	help
> +	  bpftrace is a high-level tracing language for Linux enhanced
> +	  Berkeley Packet Filter (eBPF) available in recent Linux
> +	  kernels (4.x).
> +
> +	  bpftrace uses LLVM as a backend to compile scripts to
> +	  BPF-bytecode and makes use of BCC for interacting with the
> +	  Linux BPF system, as well as existing Linux tracing
> +	  capabilities: kernel dynamic tracing (kprobes), user-level
> +	  dynamic tracing (uprobes), and tracepoints. The bpftrace
> +	  language is inspired by awk and C, and predecessor tracers
> +	  such as DTrace and SystemTap.
> +
> +	  It's highly recommended to NOT strip bpftrace binary.
> +	  To do so, add "bpftrace" to BR2_STRIP_EXCLUDE_FILES.
> +	  See:
> +	    https://bugzilla.redhat.com/show_bug.cgi?id=1865787
> +
> +	  https://www.github.com/iovisor/bpftrace
> +
> +comment "bpftrace needs a glibc toolchain w/ C++, gcc >= 4.8"
> +	depends on BR2_PACKAGE_LLVM_ARCH_SUPPORTS
> +	depends on !BR2_TOOLCHAIN_USES_GLIBC || !BR2_INSTALL_LIBSTDCPP \
> +		|| !BR2_TOOLCHAIN_GCC_AT_LEAST_4_8
> diff --git a/package/bpftrace/bpftrace.hash b/package/bpftrace/bpftrace.hash
> new file mode 100644
> index 0000000000..b48d9555fd
> --- /dev/null
> +++ b/package/bpftrace/bpftrace.hash
> @@ -0,0 +1,3 @@
> +# locally calculated
> +sha256  cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30  LICENSE
> +sha256  1f08661bac9dfa7d3d76b54b3b4d1738b02c3847049da90a5cff25093a9dc9c9  bpftrace-487dd815bfe4a24908950552f35ef17d6653b2f8.tar.gz
> diff --git a/package/bpftrace/bpftrace.mk b/package/bpftrace/bpftrace.mk
> new file mode 100644
> index 0000000000..0d8640a184
> --- /dev/null
> +++ b/package/bpftrace/bpftrace.mk
> @@ -0,0 +1,21 @@
> +################################################################################
> +#
> +# bpftrace
> +#
> +################################################################################
> +
> +# upcoming 0.12 version.
> +BPFTRACE_VERSION = 487dd815bfe4a24908950552f35ef17d6653b2f8
> +BPFTRACE_SITE = $(call github,iovisor,bpftrace,$(BPFTRACE_VERSION))
> +BPFTRACE_LICENSE = Apache-2.0
> +BPFTRACE_LICENSE_FILES = LICENSE
> +BPFTRACE_DEPENDENCIES = host-bison host-flex bcc libbpf llvm
> +
> +# libbfd, libopcodes

 What happens if target binutils is not built? Is it using a bundled version? In
that case, we prefer to unbundle it (i.e. select binutils).

 Regards,
 Arnout

> +ifeq ($(BR2_PACKAGE_BINUTILS),y)
> +BPFTRACE_DEPENDENCIES += binutils
> +endif
> +
> +BPFTRACE_CONF_OPTS += -DENABLE_MAN=OFF
> +
> +$(eval $(cmake-package))
>