[haiku-development] Re: Busted!
- From: Rob Judd <haiqu@xxxxxxxxxx>
- To: haiku-development@xxxxxxxxxxxxx
- Date: Sun, 24 May 2009 21:14:32 +1000
Stephan Assmus wrote:
On 2009-05-24 at 00:11:27 [+0200], Rob Judd <haiqu@xxxxxxxxxx> wrote:
Ryan Leavengood wrote:
How often are you updating your Haiku build? It appears that clearly it
is too often. If you are developing something outside of core Haiku,
and do not need the latest and greatest code DO NOT UPDATE SO OFTEN. It
seems obvious to me. Heck I avoid updating often just so I don't have
to constantly rebuild large parts of Haiku when headers change or
whatever.
If I were you I would get a stable build going, develop my code, get
that working, then update to the latest Haiku to make sure the code
still works. Unless it takes months, then I might update every two
weeks or so.
And therein lies the problem. As yet I haven't been able to get a stable
build, there are always major issues. I'm updating daily in the vague
hope that it happens.
Developing code on a computer that crashes four times an hour and
corrupts its own file system isn't very viable.
My experience is totally different, I do witness crashes or lock ups in
some important kernel threads, but it's probably more in the range of once
per two weeks or so. Usually, I can work all day without an issue. If it's
that bad for you, the most beneficial for Haiku as a project would be if
you tried debugging at least the crashes you are getting. If you rather
keep working on your other Haiku projects, maybe it would save you some
grief to switch to Linux as a host platform for the time being. Although
when you develop a driver, it may be inconvenient, because then you
probably can't use a virtual machine for testing. :-\ In that case, maybe a
setup with two computers could work, where you keep your code and do your
coding on a Linux machine, and keep a second machine running with Haiku and
transfer your test binaries via network all the time. If you want to test
complete images, maybe network booting the second machine may be an option
for "ok" turn-around times.
I'd expect to need to develop on another platform via a network
connection if this were 2003, but the OS has been capable of
self-building for over a year. I think it's time some developers got OFF
Linux and actually used Haiku as a daily system. I'm fairly sure the
bugs would suddenly become apparent.
Besides, running on Linux isn't viable; I'm developing kernel drivers.
The crash I reported earlier isn't related to r30825 at all, I just went
right back to r30782 ans tried a clean build (totally deleting
generated/objects) and I'm still getting problems with `copyattr`
throwing an exception and shutting down the build. This raises two issues:
1. Why can no-one else verify this yet?
2. How can it remain totally undiscovered until a clean build is done?
It's ever so slightly possible that I have a corrupt source file, but
the odds are against it. Just to check that I'm deleting the `copyattr`
sources and doing another `svn up`.
Rob
Other related posts:
