[haiku-development] Re: Haiku Userland on Non-Haiku Kernel
- From: Fano il primo <fanoilprimo@xxxxxxxxx>
- To: haiku-development <haiku-development@xxxxxxxxxxxxx>
- Date: Thu, 19 Feb 2015 21:02:50 +0100
A question this sort of libraryOs would permit binary compatibility with
Haiku OS? That is I could run a Haiku OS HPKG onto Cosmoe without recompile
it?
Obviously if the architecture is the same I don't imagine it would be
possible to run a X86 on a ARM with this...
IMHO this could help the ARM port until we have a working kernel.
On Thu, Feb 19, 2015 at 2:18 PM, Ingo Weinhold <ingo_weinhold@xxxxxx> wrote:
> On 19.02.2015 13:29, Ithamar Adema wrote:
>
>> On Thu, Feb 19, 2015 at 1:19 PM, Axel Dörfler <axeld@xxxxxxxxxxxxxxxx
>> <mailto:axeld@xxxxxxxxxxxxxxxx>> wrote:
>> That depends :-)
>> You might be able to just ignore the PM part of Haiku and get away
>> with it.
>> However, if you wanted to support PM, you'll have to port packagefs
>> to your target OS.
>>
>> Yeah agreed, that was on the mental roadmap, but for now I'd be happy to
>> have a _very_ minimal Haiku userland ;)
>>
>
> Pre PM we only had a few pre-built build dependencies -- zip files --
> needed for building a minimal image. These days there are more of them and
> they are HPKG files. So depending what the immediate goal is -- a minimal
> self-contained system (Linux/*BSD kernel and POSIX libs (and maybe X11 and
> SDL) plus Haiku everything else) or a simple Haikuish test environment
> (libs plus some servers and apps) on a complete Linux/*BSD system -- you'll
> either have to implement a packagefs or hack up the build system (and parts
> of Haiku).
>
> With a pre PM version, I think, a lot less hacking should be necessary to
> reach either goal. For the latter one, concerning the build dependencies
> you might get away with providing empty zip files (respectively ones with
> symlinks to the build host's directories).
>
> Regarding system interfaces there isn't that much that has changed since
> pre PM (some system/CPU info structures/syscalls), so updating a working
> pre PM based version to PM would mainly consist of getting the build going
> again (bootstrap build with haikuporter...) and implementing packagefs.
>
> To sum that up, I think starting with a pre PM version will get you
> initial results quicker and won't really add to the total amount of work.
> It might even save a little work, since some hacks around the package
> aspects of the PM build system won't be necessary at all.
>
> CU, Ingo
>
>
>
Other related posts:
