FreeLists / haiku-development / [haiku-development] Re: Haiku alpha 1 release (draft)

[Simon Taylor <simontaylor1@xxxxxxxxxxxx>]

Hi Rene,

Rene Gollent wrote:
> > I'm not convinced this is actually necessary - I wouldn't expect the
> > devs to shift their development platform to Haiku after the alpha, so
> > self-hosting is more a symbolic milestone that demonstrates the
> > stability of lots of different OS subsystems. Linux is a very nice
> > development platform for Haiku - rock solid, quick at compiling, and
> > allows fast turn-around using emulators.
> The problem is that viewpoint is a bit of a copout. You will not find
> anywhere near as many bugs in the base system if you don't force it to
> do any heavy work. Building the entire thing is one of the heaviest
> tasks you can throw at it, as it stresses the VM, disk and scheduler
> subsystems quite heavily, and will result in a lot of latent bugs
> being discovered that would otherwise remain hidden in day-to-day use.
I agree it's a great stress test, but having this as one of the key 
things underlying the alpha seems unnecessary. Though I think someone 
should have successfully built Haiku under Haiku to demonstrate 
stability, there are other tasks before we can truly call Haiku 
self-hosting. I presume the build system will need updating (copying new 
files over the top of a running system probably isn't a good idea), the 
compiler and development environment needs to be built for Haiku, all of 
the Be Inc. headers need to be replaced, and probably other stuff.
The first thing people are going to do when they boot the alpha will not 
be to check out the entire Haiku tree and attempt to build the whole 
thing - they will boot it, have a look around, and play with whatever 
demo apps we put on the image. If their experience is positive they 
might begin following the project more closely or even look into what 
they can do to help. 95% of them would probably follow whatever 
development setup the core developers use (presuming it is well 
documented somewhere) ie Linux/BeOS.
> Also, it would increase the incentive to improve performance in the
> areas that are lacking with respect to build speed, which isn't
> anywhere near as high a priority if you just use Linux to build
> instead.
Is that a major problem for the alpha?

> [...] If I want to try and start finding bugs in Haiku's
> kits by writing apps against them, the last thing I want to do is have
> to reboot a VM every time I want to make some changes to an app.
Haiku is already perfectly stable enough to build and run small test 
apps like that.
So again I think we need to better define who the alpha is for and what 
we expect them to do with it.

Simon