Re: [PATCH libmnl 0/6] Doxygen Build Improvements

[Duncan Roe <duncan_roe@optusnet.com.au>]

Hi Jeremy,

On Wed, Aug 03, 2022 at 09:12:41PM +0100, Jeremy Sowden wrote:
> These changes were prompted by Mark Mentovai's request to remove a hard-coded
> `/bin/bash` path from the rule that creates the man-page sym-links.  Hitherto,
> the doxygen Makefile has jumped through a number of hoops to make sure
> everything works with `make distcheck` and parallel builds.  This patch-set
> makes some doxygen config changes that obviate the need for them, fixes a bug in
> `make clean`, updates .gitignore and moves the shell-script out of the Makefile
> into a separate file for ease of maintenance.  In the process, the hard-coded
> `/bin/bash` is removed.
>
> One thing I have left is the setting of `-p` when running the shell-script.  The
> comment reads "`bash -p` prevents import of functions from the environment".
> Why is this a problem?
>
> Jeremy Sowden (6):
>   build: add `make dist` tar-balls to .gitignore
>   doc: add .gitignore for Doxygen artefacts
>   doc: change `INPUT` doxygen setting to `@top_srcdir@`
>   doc: move doxygen config file into doxygen directory
>   doc: move man-page sym-link shell-script into a separate file
>   doc: fix doxygen `clean-local` rule
>
>  .gitignore                               |  3 +-
>  configure.ac                             | 15 ++++++-
>  doxygen/.gitignore                       |  4 ++
>  doxygen/Makefile.am                      | 53 +++---------------------
>  doxygen.cfg.in => doxygen/doxygen.cfg.in |  4 +-
>  doxygen/finalize_manpages.sh             | 40 ++++++++++++++++++
>  6 files changed, 67 insertions(+), 52 deletions(-)
>  create mode 100644 doxygen/.gitignore
>  rename doxygen.cfg.in => doxygen/doxygen.cfg.in (91%)
>  create mode 100644 doxygen/finalize_manpages.sh
>
> --
> 2.35.1
Thanks for this but ... I think it would be better to use the netfilter-log
model.

That way there is no new shell script to maintain - libmnl would pick up
libnetfilter-queue's copy.

I found the time-consuming part was checking the new SYNOPSIS lines to be
sufficient. I have far less time nowadays (no Covid lockdowns, &c.) but I will
try to get it done soon-ish.

In the meantime I think Mark's 1-line patch to the existing mess is adequate.

The bin/bash thing is easily resolved: I've tested this in libnetfilter-queue:

> --- a/doxygen/build_man.sh
> +++ b/doxygen/build_man.sh
> @@ -1,4 +1,5 @@
> -#!/bin/bash -p
> +#!/bin/sh
> +[ -n "$BASH" ] || exec bash -p $0
>
>  # Script to process man pages output by doxygen.
>  # We need to use bash for its associative array facility.

Cheers ... Duncan.