Re: [dpdk-dev] [PATCH 17/20] app/chkincs: add chkincs app to verify headers

On Thu, Jan 14, 2021 at 01:16:58PM +0100, David Marchand wrote:

On Thu, Jan 14, 2021 at 12:09 PM Bruce Richardson
[off-list ref] wrote:

quoted

To verify that all DPDK headers are ok for inclusion directly in a C
file, and are not missing any other pre-requisite headers, we can
auto-generate for each header an empty C file that includes that header.
Compiling these files will throw errors if any header has unmet
dependencies.

Some drivers expose APIs to applications, their headers would need checks too.

Yes, that is something that should be checked too, but I've left it as
"further work" for now, and I'm not planning it in this set, as I view it
as less important and this set is big-enough as it is for now. :-)

quoted

Signed-off-by: Bruce Richardson <redacted>
---
 app/chkincs/gen_c_file_for_header.py | 49 ++++++++++++++++++++++++++++
 app/chkincs/main.c                   |  4 +++
 app/chkincs/meson.build              | 28 ++++++++++++++++
 app/meson.build                      |  1 +
 lib/meson.build                      |  1 +
 meson.build                          |  1 +
 meson_options.txt                    |  2 ++
 7 files changed, 86 insertions(+)
 create mode 100755 app/chkincs/gen_c_file_for_header.py
 create mode 100644 app/chkincs/main.c
 create mode 100644 app/chkincs/meson.build

diff --git a/app/chkincs/gen_c_file_for_header.py b/app/chkincs/gen_c_file_for_header.py
new file mode 100755
index 0000000000..f92f2b412c
--- /dev/null
+++ b/app/chkincs/gen_c_file_for_header.py

@@ -0,0 +1,49 @@
+#! /usr/bin/env python3
+# SPDX-License-Identifier: BSD-3-Clause
+# Copyright(c) 2020 Intel Corporation

2021*

Code actually was written in 2020 hence the date there.

quoted

+
+from sys import argv
+from os.path import abspath
+
+empty_contents = 'static const char *empty __attribute__((unused)) = "empty";'
+# files which are not used directly, but included via others
+exceptions = [
+    'rte_cmp_arm64.h',
+    'rte_cmp_x86.h',
+    'rte_crc_arm64.h',
+    'rte_eal_interrupts.h',
+    'rte_eth_ctrl.h',
+    'rte_ethdev_core.h',
+    'rte_ipsec_group.h',
+    'rte_lpm_altivec.h',
+    'rte_lpm_neon.h',
+    'rte_lpm_sse.h',
+    'rte_lpm_x86.h',
+    'rte_lru_arm64.h',
+    'rte_lru_x86.h',
+    'rte_regexdev_core.h',
+    'rte_ring_core.h',
+    'rte_ring_generic.h',
+    'rte_ring_hts_c11_mem.h',
+    'rte_ring_hts.h',
+    'rte_ring_peek_c11_mem.h',
+    'rte_ring_peek.h',
+    'rte_ring_peek_zc.h',
+    'rte_ring_rts_c11_mem.h',
+    'rte_ring_rts.h',
+    'rte_stack_lf_c11.h',
+    'rte_stack_lf_generic.h',
+    'rte_stack_lf.h',
+    'rte_stack_std.h',
+    'rte_table_hash_func_arm64.h',
+    ]

Can we instead flag those headers from the libraries themselves?
In addition of the headers current variable, something like a
internal_headers or private_headers variable?

Yes, we could do that, though we need a suitable name, since they aren't
internal only or private. :-) Maybe just have it literally called
"skip_chkincs_headers"? Open to more ideas...

Thanks for the review!

/Bruce