[haiku-development] Re: [haiku-commits] haiku: hrev46863 - headers/private/kernel/arch/arm
- From: "Ithamar R. Adema" <ithamar@xxxxxxxxxxxxxxxxxxx>
- To: haiku-development@xxxxxxxxxxxxx
- Date: Sun, 16 Feb 2014 16:57:00 +0100
On Sun, Feb 16, 2014 at 4:46 PM, Jonathan Schleifer <
js-haiku-development@xxxxxxxxxxx> wrote:
> Am 16.02.2014 um 15:18 schrieb Ingo Weinhold <ingo_weinhold@xxxxxx>:
>
> > You seem to have two major misconceptions:
> >
> > 1. The ports we're talking about aren't "a little broken". They are (in
> some cases largely) incomplete. They are only booting the kernel (or even
> only the boot loader) up to some point. None of them supports a userland.
>
> Having the kernel already boot is quite an achievement already, IMHO. Not
> a lot is missing from there to a booting userland, mostly drivers.
>
Please define "booting the kernel", as just getting to main() in the kernel
is quite trivial. Getting user land working is mostly kernel side
implementation, and more complex then getting the kernel torun a bit.
> > 2. It is not a *huge* task to do an architecture port. It certainly is a
> fair amount of work, but I doubt that it would take a dedicated and able
> developer more than a few month to get a working (basic) userland for a
> completely new architecture. I don't think just stubbing the port would
> take more than a few hours.
>
> A few months of work are still a few months of work, and we are low on man
> power. We should keep them as long as possible to avoid doing that work
> later again.
>
Having a github repository with that code in there, or a separate branch,
does not make the code go away. If we actually keep a list of "experimental
ports" on the wiki with pointers to the repository/branches, it should even
be easy to find.
> No one suggested obliterating the port from the world. It should simply
> live in a separate branch(/clone). I think synchronizing the state of the
> branch only when the time to address the regressions is invested as well is
> a better approach than keeping the port fully in sync the master while
> breaking it more and more.
>
> > Haiku's history disagrees.
>
> So there is a port that was removed form the tree and has been revived?
>
No, but the fact that the only !x86 port that has reached any level of
functionality is still not usable at all, does indicate that the ports are
an afterthought, an experiment, and nobody is seriously making time for
them. Why burden core developers with extra work if the maintainers of the
ports don't have the time to spend on it themselves? (myself included, no
offence intended).
I still think that it would be better to keep them out of the main repo,
simply because it suggests a level of support/development from the core
developer team, which is simply not there. Look at the amount of people in
forums asking about ARM progress, while there is clearly no developer
motivation/time to actually get it done?
Ithamar.
Other related posts:
