[Buildroot] [PATCH v1 2/3] package/iwd: add basic configuration file

[Peter Seiderer]

Hello Yann,

On Sun, 14 Jun 2020 22:57:15 +0200, "Yann E. MORIN" <yann.morin.1998 at free.fr> wrote:

> On 2020-06-14 21:07 +0200, Peter Seiderer spake thusly:
> > On Sun, 14 Jun 2020 11:56:29 +0200, "Yann E. MORIN" <yann.morin.1998 at free.fr> wrote:
> > > On 2020-06-13 01:24 +0200, Peter Seiderer spake thusly:
> [--SNIP--]
> > > > +	select BR2_PACKAGE_OPENRESOLV if !BR2_INIT_SYSTEMD
> > > I think this is preferable to use BR2_PACKAGE_SYSTEMD as the dependency,
> > > rather than BR2_INIT_SYSTEMD, even if the two are currently tightly
> > > intertwinned.
> > I would prefer the !BR2_INIT_SYSTEMD as I believe iwd needs
> > an service of the running systemd (or an alternative implementation)
> > and not a feature/library of the systemd package (only runtime not
> > a compile time dependency)...
>
> I'm still not convinced...
>
> What if one has:
>
>     BR2_INIT_SYSTEMD=y
>     # BR2_PACKAGE_SYSTEMD_RESOLVED is not set
>
> Does systemd still provides resolv.conf?
>
> I guess not. So in fact the dependency is really not about the init
> system. but about a package that provides resolv.conf. In which case it
> should be changed to:
>
>     select BR2_PACKAGE_OPENRESOLV if !BR2_PACKAGE_SYSTEMD_RESOLVED

Or:
	select BR2_PACKAGE_OPENRESOLV if !BR2_INIT_SYSTEMD
	select BR2_PACKAGE_SYSTEMD_RESOLVED if BR2_PACKAGE_SYSTEMD
>
> Or did I miss something else?
>
> [--SNIP--]
> > > > +ifeq ($(BR2_INIT_SYSTEMD),y)
> > > > +define IWD_CONFIG_FILE_NAME_RESOLV_SERVICE
> > > > +	echo -e "[Network]\nNameResolvingService=systemd" >> $(TARGET_DIR)/etc/iwd/main.conf
> [--SNIP--]
> > > However, this will not be working nicely on rebuilds, as it will keep
> > > appending this to the file. See below for a better approach (I think):
> > On re-builds the common IWD_INSTALL_CONFIG_FILE (copy plain main.conf will
> > be executed first, so no problem...
>
> Indeed, but still that's not obvious, as it splits the install of that
> one file in two places: a generic one, and a per-init-system one.
>
> > > Also, the condiotion on BR2_INIT_SYSTEMD is already handled by the
> > > infra, by using the appropriate INIT_INSTALL hooks:
> > >
> > >     define IWD_INSTALL_INIT_SYSV
> > >         sed -r -e '$s/$/\n[Network]\nNameResolvingService=resolvconf/' \
> > >             $(IWD_PKGDIR)main.conf \
> > >             >$(TARGET_DIR)/etc/iwd/main.conf
> > >     endef
> >
> > Does not work, gives:
> >
> > 	sed: -e expression #1, char 46: missing command
>
> This is Makefile, I forgot to duplicate the $ signs when copy-pasting
> from my shell experiment...
>
> > And did not figure out what would be the right sed exprecssion (the leading $
> > seems to be wrong, but without another error occures)...
>
> The leading $ means 'on the last line', so the replacement is only
> attemped on the last line.

Did not know this one, will test...

>
> > >     define IWD_INSTALL_INIT_SYSTEMD
> > >         sed -r -e '$s/$/\n[Network]\nNameResolvingService=systemd/' \
> > >             $(IWD_PKGDIR)main.conf \
> > >             >$(TARGET_DIR)/etc/iwd/main.conf
> > >     endef
> > Nice shortcut, but misses the enable-the-one-option-for-systemd and
> > enable-the-other-option-for-all-other-init-systems...
>
> Sorry, I don't follow you.
>
> For systemd, only the IWD_INSTALL_INIT_SYSTEMD hook is run, and for the
> others, only the IWD_INSTALL_INIT_SYSV hook is run. And each does append
> a different NameResolvingService key, systemd in the first case,
> resolvconf in the second case.
>
> Ah, there is *one* case where the file would not be copied and expanded
> with my proposal, when the user uses BR2_INIT_NONE. But this really is
> the user's reponsibility to catter for the requirements of their init
> system...

And what is about the LIBFOO_INSTALL_INIT_OPENRC mentioned in [1]?

Regards,
Peter

[1] https://buildroot.org/downloads/manual/manual.html#_infrastructure_for_packages_with_specific_build_systems

>
> [--SNIP--]
> > > > +IWD_POST_INSTALL_TARGET_HOOKS += \
> > > > +	IWD_INSTALL_CONFIG_FILE \
> > > > +	IWD_CONFIG_FILE_NAME_RESOLV_SERVICE
> > > And thus those post-install hooks are not longer required.
> > Nice, but see downside above...
>
> Except for BR2_INIT_NONE, I don't see the problem...
>
> Regards,
> Yann E. MORIN.
>