Am Freitag, 29. Oktober 2010, 18:49:32 schrieb Andrew:
> 29.10.2010 18:53, davidMbrooke пишет:
> > On Wed, 2010-10-27 at 20:49 +0100, davidMbrooke wrote:
> >> I am now going to investigate some of the suggestions from here:
> >> http://old.nabble.com/pthread-and-daemon%28%29-problem-td27029024.html
> >
> > This did the trick. I have created a simple patch for uClibc 0.9.30.3
> > which makes pthread_initialize() behave like it did with GCC 3.x and now
> > aiccu works perfectly in daemon mode. Another case of many hours of work
> > resulting in a one line change :-)
> >
> > Hopefully we can retire this patch when we upgrade to uClibc 0.9.31 or
> > higher, so I have added comments to that effect in
> > buildenv/buildtool.cfg.
> >
> > I have also been looking at the syslog-ng configuration and there is a
> > relatively simple change (removing some comments) which directs "debug"
> > output to /var/log/debug. (The "verbose" output from aiccu counts as
> > "debug".)
> >
> > dMb
>
> It's good.
Andrew;
I'm currentyl rebuilding from scratch and install hopefully over the weekend
to be shure it's as stable as it had been the past weeks.
What about the switch of gcc you mentioned a week ago? This should be done
before beta1, so we have a fixed environment to work with for the release
cycle.
> So we just need to decide what CPU and what code optimization will be
> used for user-level software for beta release?
I'd say as generic as possible (486 as with 3.x) and to decicde about all the
optimizations later. I have no idea what's needed to build for different cpu's,
no idea how make usweful performance tests.
I expect a slow adoption of the beta and often on test-systems, but it should
be as error-free as possible, performance won't be the issue for the first
release.
> Also I still not know opinions of members about these questions:
> Anybody tries to test dhrystone benchmark with different
> archs/optimizations on different CPUs? Is i486 arch with optimization
> for i686 will be good for all? Or this optimization will cause
> significant performance loss on some archs (Geode, i486)?
>
> Also, maybe it'll be good to compile all packages that are included into
> initrd (busybox, uClibc) with optimization for specific platforms for
> maximal performance - in any case, we must provide kernel+modules+initrd
> for each platform?
Let us pls fix remaining bugs (echo -e in initrd build, apkg -u) and then we
have organizational questions that needs solutions:
- I still miss a leaf-trac mailinglist, to keep track of TRAC entries
- Documentation improves thx to David currently on a daily basis, but still a
lot of work to do (though it doesn't need to be completly ready for beta1)
- Shall we provide compiled packages, as we did with Bering-uClibc ("packages
page"), better ideas?
- What about write permissions in cvs, shall we revive the contrib section
(see http://sourceforge.net/apps/mediawiki/leaf/index.php?title=Talk:Bering-
uClibc_4.x_-_Developer_Guide_-_Policies_and_Guidelines)
Some of this should be solved before beta1, some until beta2, performance and
different kernel etc are something for around beta3, once the dust settles and
we have time to invest work on these.
Currently we are three active developers (and working in our spare time), so
it's necessary to focus on the most needed and important things first.
And I think we have made good progress towards a release, I'm shure it can be
done next month. Would be good to see all the work in the wild.
ymmv
kp
------------------------------------------------------------------------------
Nokia and AT&T present the 2010 Calling All Innovators-North America contest
Create new apps & games for the Nokia N8 for consumers in U.S. and Canada
$10 million total in prizes - $4M cash, 500 devices, nearly $6M in marketing
Develop with Nokia Qt SDK, Web Runtime, or Java and Publish to Ovi Store
http://p.sf.net/sfu/nokia-dev2dev
_______________________________________________
leaf-devel mailing list
leaf-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/leaf-devel
