Re: [OE-core] of (morty->zeus) ntp migration, and missing "autogen-native" DEPENDS

Re: [OE-core] of (morty->zeus) ntp migration, and missing "autogen-native" DEPENDS

[Robert P. J. Day]

On Thu, 21 Jan 2021, Robert P. J. Day wrote:

>   colleague hands me the following curiosity ... approximately
> morty-based build, wherein ntp_4.2.8p10 has a bbappend file which
> contains (among other things) the line:
>
>   DEPENDS += "autogen-native perl-native"
>
> i'm always nervous when a bbappend file just adds new build
> dependencies but, apparently, this has worked fine all this time, as
> there was an autogen-native recipe back then.
>
>   dragging this recipe forward, unsurprisingly, it fails as there is
> no autogen-native recipe anymore. so i'm just going to drop that
> dependency, while poking aorund to figure out why it was ever needed
> in the first place.
>
>   and apparently, this is not the first time that a legacy recipe
> which contained an autogen-native DEPENDS line has caused problems, so
> what is the deal with autogen-native? clearly, i need to get rid of
> that, but is there an explanation as to how that migration was done
> elsewhere? i'm about to find out what happens when i go that, while
> going from ntp 4.2.8p10 to 4.2.8p13, and seeing what else breaks.

  ok, i've focused this down a bit more, and here's the problem. one
of the patches brought forward and applied to 4.2.8p13 is to add a
new flag to sntp/sntp-opts.def, where a flag has the format:

  flag = {
    name      = unprivport;
    value     = u;
    descrip   = "Use an unprivileged port";
    doc = <<-  _EndOfDoc_
	Use an unprivilegded UDP port for our queries.
	_EndOfDoc_;
  };

as long as the patch that adds that new flag is not applied, things
build fine. and if one looks at the top of that file as it comes with
ntp-4.2.8p13, one reads:

  autogen definitions options;

  #include autogen-version.def

argh.

  so, as i read it, if you modify that file, you will automatically
kick in the requirements for autogen, which is exactly what i am
trying to avoid.

  am i making sense? is there an alternative approach to doing this
that removes the need for autogen?

rday