tiny package --

[Patrick Venture]

On Mon, Sep 10, 2018 at 5:39 AM Brad Bishop <bradleyb at fuzziesquirrel.com> wrote:
>
> > On Sep 8, 2018, at 12:01 PM, Patrick Venture <venture at google.com> wrote:
> >
> > Brad;
> >
> > I was looking at publishing a single-header file from a package
> > containing only that header file.  In this case, "oemgoogle.hpp" into
> > the same place that phosphor-host-ipmid installs its headers so that
> > any google-ipmi-* packages could store their OEM numbers in one file
> > (upstream).
> >
> > Downstream we store that file in our phosphor-host-ipmid mirror.
> >
> > The advice I seek is -- it doesn't seem worth creating a separate
> > repository for one file, but I know it's a frowned upon practice to
> > store code in the bitbake repository -- but in this case, I feel like
> > it may be appropriate.  Does that make sense?
>
> Obviously it will work just fine but I probably wouldn’t allow it in
> the IBM or phosphor layers, out of fear of it being used as a counter
> example or justification for a future change that isn’t quite so harmless.
> But it is truly _your_ layer, so you can do whatever you want.
>
> >
> > I'm thinking:
> > meta-google/recipes-google/ipmi/google-ipmi-header.bb
> > meta-google/recipes-google/ipmi/google-ipmi-header/configure.ac
> > meta-google/recipes-google/ipmi/google-ipmi-header/Makefile.am
> > meta-google/recipes-google/ipmi/google-ipmi-header/oemgoogle.hpp
> >
> > Patrick
>
> Why is the number hardcoded at all?  Could it be a configure option or
> command line option?

I could definitely add the number to each daemon or library, however
having a single source of the numbers is valuable to avoid accidental
collisions.

>
> If those don’t work, as a fall-back can we just put the header in the
> upstream phosphor-host-ipmid repository as you’ve done downstream?

I started with this approach and it was rejected by the maintainers,
because the numbers reserved for Google code (even shared upstream)
"didn't belong" in phosphor-host-ipmid.

>
> My two cents is that one of these three options would be simpler.

I agree that having the header file just in phosphor-host-ipmid would
be easier, and I'd like to do effectively that (I'd really really
rather just do that exact thing).  Having a single file to point to is
a nice solution.  I'm working on upstreaming more Google-"only" code
for others.  I imagine, some of it will stay Google-only, while others
will get phosphor-ized over time.  It's difficult to know what today
you'll really need tomorrow.  And if nobody ever uses something but
us, I still would rather the world have a chance to use it or comment
on it, or improve it.

>
> -brad