Re: Release Schedule for 1.8

2007-01-04 Thread Miguel Sousa Filipe

On 1/3/07, Matthew Dillon [EMAIL PROTECTED] wrote:


:
: Include ALTQ in generic?  Hmm. I would personally rather not, unless
: a really good case is made for including it.
:
:The only reason why I say this is the fact that many would like to use BSD 
flavors as home routers with transfer rate limiting, and regular lines don't 
require a too powerful computer, an 586 or 686 will do (we were NATing 10Mbit 
with a PI-133 and 48M EDO RAM.) On these machines, kernel building takes hours.

Well... you do the build on a faster machine and copy the binary?



that requires some other dfly machine, does it not?



: We *are* including
: kernel sources on the ISO now, so perhaps a better solution would
: be to create a section on how to set up a kernel build infrastructure
: given just a freshly installed ISO.
:
:Do you mean the nativekernel target?
:
:--
:Gergo Szakal [EMAIL PROTECTED]

Yes.  It should theoretically only require the sources tar'd up on the CD
to build a new kernel.  That is one of the things that we have to test
for this release... can you unpack the sources stored on the ISO in a
chroot'd environment and actually build the kernel without the rest of
/usr/src ?  Inquiring minds want to know!

-Matt




--
Miguel Sousa Filipe


Re: Release Schedule for 1.8

2007-01-04 Thread Gergo Szakal
On Thu, 4 Jan 2007 19:09:33 +
Miguel Sousa Filipe [EMAIL PROTECTED] wrote:

 On 1/3/07, Matthew Dillon [EMAIL PROTECTED] wrote:
 
  :
  : Include ALTQ in generic?  Hmm. I would personally rather not, unless
  : a really good case is made for including it.
  :
  :The only reason why I say this is the fact that many would like to use BSD 
  flavors as home routers with transfer rate limiting, and regular lines 
  don't require a too powerful computer, an 586 or 686 will do (we were 
  NATing 10Mbit with a PI-133 and 48M EDO RAM.) On these machines, kernel 
  building takes hours.
 
  Well... you do the build on a faster machine and copy the binary?
 
 
 that requires some other dfly machine, does it not?

No, it's enough to put the HDD into another, fast machine for that time. I have 
done it a couple times.


-- 
Gergo Szakal [EMAIL PROTECTED]
University Of Szeged, HU
Faculty Of General Medicine

/* Please do not CC me with replies, thank you. */


Re: Release Schedule for 1.8

2007-01-03 Thread Karthik Subramanian

On 1/2/07, Justin C. Sherrill [EMAIL PROTECTED] wrote:

On Tue, January 2, 2007 1:45 am, Karthik Subramanian wrote:

 Is there someplace where there's a list of release-oriented code
 bits/features, so that people can pick off stuff that they're
 interested in testing? I'm moving to a different city/state in a week
 or so, so i'm not sure if I can contribute to any testing right away -
 but I'd like to take a look all the same.

That would be a good thing for the wiki.  In fact, if you want to help out...



I'd love to - but I'm going to be off the internet for around a couple
of weeks starting Friday, and I'm not sure how much I can do before
that :(


Re: Release Schedule for 1.8

2007-01-03 Thread Matthew Dillon

:Matt, what about the SMP kernels?
:Oh, and may I request ALTQ be compiled into it/them by default?
:
:-- 
:Gergo Szakal [EMAIL PROTECTED]
:University Of Szeged, HU
:Faculty Of General Medicine

I don't think we're going to be able to include SMP kernels on the
release ISO for this release, when it was brought up last time 
a number of issues were pointed out that would require us to have
at least two and maybe even four different kernel binaries (and related 
module directories) to cover all the bases.

Include ALTQ in generic?  Hmm. I would personally rather not, unless
a really good case is made for including it.  We *are* including 
kernel sources on the ISO now, so perhaps a better solution would
be to create a section on how to set up a kernel build infrastructure
given just a freshly installed ISO.

-Matt
Matthew Dillon 
[EMAIL PROTECTED]


Re: Release Schedule for 1.8

2007-01-03 Thread Gergo Szakal
On Wed, 3 Jan 2007 12:58:11 -0800 (PST)
Matthew Dillon [EMAIL PROTECTED] wrote:

 Include ALTQ in generic?  Hmm. I would personally rather not, unless
 a really good case is made for including it.   

The only reason why I say this is the fact that many would like to use BSD 
flavors as home routers with transfer rate limiting, and regular lines don't 
require a too powerful computer, an 586 or 686 will do (we were NATing 10Mbit 
with a PI-133 and 48M EDO RAM.) On these machines, kernel building takes hours.

 We *are* including 
 kernel sources on the ISO now, so perhaps a better solution would
 be to create a section on how to set up a kernel build infrastructure
 given just a freshly installed ISO.

Do you mean the nativekernel target?

-- 
Gergo Szakal [EMAIL PROTECTED]
University Of Szeged, HU
Faculty Of General Medicine

/* Please do not CC me with replies, thank you. */


Re: Release Schedule for 1.8

2007-01-03 Thread walt
Timothy wrote:
 On Tuesday 02 January 2007 12:11, Matthew Dillon wrote:
 Well, my biggest issue for this release is that all the packages that
 compiled and ran on 1.6 also compile and run now...

 ...I'll do a compile of kde and gnome, that should wring it out pretty good.

Yes!  And thanks for joining the DragonFly community, Timothy.

There are a few chronic difficulties with a few pkgsrc packages, caused
by Matt's ongoing changes to the data structures in the kernel.

These changes have affected mostly apps which require access to kernel
statistics like CPU/memory usage, network load, etc. Things, e.g., that
'top' displays -- and therefore similar userland packages like gnome's
libgtop2, and system-monitor-like packages from kde.

The other big category of problems I've noticed is with any package that
needs access to CD hardware.  This is due to changes in the location of
header files in the DF source tree.  (These are trivial to fix once you
know where to find the header files.)

So, please build away, and report back with problems.

Thanks!


Re: Release Schedule for 1.8

2007-01-02 Thread Justin C. Sherrill
On Tue, January 2, 2007 1:45 am, Karthik Subramanian wrote:

 Is there someplace where there's a list of release-oriented code
 bits/features, so that people can pick off stuff that they're
 interested in testing? I'm moving to a different city/state in a week
 or so, so i'm not sure if I can contribute to any testing right away -
 but I'd like to take a look all the same.

That would be a good thing for the wiki.  In fact, if you want to help out...



Re: Release Schedule for 1.8

2007-01-02 Thread Matthew Dillon
Well, my biggest issue for this release is that all the packages that 
compiled and ran on 1.6 also compile and run now.  One thing I
would like people with extra (test) machines to do is to do a clean
install and then build their packages starting with a clean slate and
see what problems pop up.

People running HEAD can do the same thing without reinstalling by
constructing a chroot environment, if you happen to have enough disk
space.

I'd also like to see that packet filter bug found and fixed.  I went
through the code but I couldn't find a smoking gun.

On the driver front, NATA is still considered experimental and will
not be the default for this release, so it doesn't need any serious
testing until after the release cycle.  I would like to make it the
default for the 2.0 release 6 months from now.

The network drivers all need testing.  Both wireless and hard
wired adapteres.  Sepherosa has done a phenominal amount of work
on the network front.

GCC-4.1 is in the tree but it is not the default for this release
and does not need to be tested.  We will probably make 4.1 or 4.2
the default for the 2.0 release 6 months from now.  So just don't 
worry about it for now.

dhclient needs some testing, because none of the patches were being
applied after we made the changes to the auto patch system months ago.
I just fixed that a day or two ago.  It seems to work on my test box.

-Matt



Re: Release Schedule for 1.8

2007-01-02 Thread Justin C. Sherrill
On Tue, January 2, 2007 1:11 pm, Matthew Dillon wrote:
 I would like to make it the
 default for the 2.0 release 6 months from now.

Is it going to be 2.0 so that we don't have a 1.10, or are there other
targets for that release?  I was hoping (without any idea of feasibility)
that 2.0 would mark the arrival of ZFS or the departure of the BGL.



Re: Release Schedule for 1.8

2007-01-02 Thread Timothy
On Tuesday 02 January 2007 12:11, Matthew Dillon wrote:
 Well, my biggest issue for this release is that all the packages that
 compiled and ran on 1.6 also compile and run now.  One thing I
 would like people with extra (test) machines to do is to do a clean
 install and then build their packages starting with a clean slate and
 see what problems pop up.

Ok, I have a amd k7 3000 with a 36gb scsi hd and 1gig of mem.  I'll be glad to 
volunteer it for a clean install.  The question is should I download and 
build from a snapshot or do you want a full cvs build?  I'll do a compile of 
kde and gnome, that should wring it out pretty good.

Timothy


Re: Release Schedule for 1.8

2007-01-02 Thread Matthew Dillon

:On Tue, January 2, 2007 1:11 pm, Matthew Dillon wrote:
: I would like to make it the
: default for the 2.0 release 6 months from now.
:
:Is it going to be 2.0 so that we don't have a 1.10, or are there other
:targets for that release?  I was hoping (without any idea of feasibility)
:that 2.0 would mark the arrival of ZFS or the departure of the BGL.

I don't want to have a 1.10.  We'll see how much gets done in 6 months
but it is going to be 2.0 regardless.

-Matt
Matthew Dillon 
[EMAIL PROTECTED]


Release Schedule for 1.8

2007-01-01 Thread Matthew Dillon
* The CVS tree will be branched in two weeks, on Sunday, January 14th

* The 1.8 release is slated for January 28th.

I would like people to start testing release oriented bits if possible.  
We have about two weeks before the CVS tree is branched, and two weeks
after it is branched for fine-tuning and package building.  So if there
are things you want to get into the 1.8 release but haven't gotten to yet,
now is the time.

-Matt



Re: Release Schedule for 1.8

2007-01-01 Thread Victor Balada Diaz
On Mon, Jan 01, 2007 at 01:00:31PM -0800, Matthew Dillon wrote:
 * The CVS tree will be branched in two weeks, on Sunday, January 14th
 
 * The 1.8 release is slated for January 28th.
 
 I would like people to start testing release oriented bits if possible.  
 We have about two weeks before the CVS tree is branched, and two weeks
 after it is branched for fine-tuning and package building.  So if there
 are things you want to get into the 1.8 release but haven't gotten to yet,
 now is the time.

What's the status of VKERNEL stuff? was one of the major things for
this release.

We are having a lot of problems with our IRQ routing code, i remember
you proposing a solution for it some time ago, do you think that it's
doable in time for the release? I wouldn't like to ship with stability
problems and in my opinion we should delay the release until that is
fixed.

About testing, i think that we should start looking to do automated
regression testing of the system as much as we can. TET[1] looks like
a very good solution and FreeBSD is probably going to use it[2], so we
could get a lot of regression tests for free that apply to DragonFly.


[1]: http://tetworks.opengroup.org/
[2]: http://wikitest.freebsd.org/TetIntegration

-- 
La prueba más fehaciente de que existe vida inteligente en otros
planetas, es que no han intentado contactar con nosotros. 


Re: Release Schedule for 1.8

2007-01-01 Thread Matthew Dillon

:What's the status of VKERNEL stuff? was one of the major things for
:this release.

I am going to try to get it done in 2 weeks.  No guarentees.  It
can't hold up the release.  

But it won't be production for at least two months after that.  e.g.
we'll need to write a network driver and do some other bits to make
it truely useful.  Once I get the initial virtual kernel working
I will impose on other developers to work on some drivers for it :-)

:We are having a lot of problems with our IRQ routing code, i remember
:you proposing a solution for it some time ago, do you think that it's
:doable in time for the release? I wouldn't like to ship with stability
:problems and in my opinion we should delay the release until that is
:fixed.

The IRQ issues are the same we've had since inception.  i.e. no better
and no worse then before.  We have workarounds (build SMP without
APIC_IO,and use the emergency interrupt poll sysctl).  I don't think
we can rush a solution in two weeks.

:About testing, i think that we should start looking to do automated
:regression testing of the system as much as we can. TET[1] looks like
:a very good solution and FreeBSD is probably going to use it[2], so we
:could get a lot of regression tests for free that apply to DragonFly.
:
:[1]: http://tetworks.opengroup.org/
:[2]: http://wikitest.freebsd.org/TetIntegration

I dunno.  I'm always skeptical about such frameworks and wonder 
whether porting them actually saves us any time over building
our own.  I do think testing will become a lot easier with the
virtual kernel giving us a totally self contained environment for
three files (kernel binary, memory image file, disk image file).

-Matt
Matthew Dillon 
[EMAIL PROTECTED]


Re: Release Schedule for 1.8

2007-01-01 Thread Matt Emmerton
 I would like people to start testing release oriented bits if
possible.
 We have about two weeks before the CVS tree is branched, and two weeks
 after it is branched for fine-tuning and package building.  So if
there
 are things you want to get into the 1.8 release but haven't gotten to
yet,
 now is the time.

I'm hoping to have the [get|set|make|swap]context functions ported to DFly
in time for the release.  I have all the code bits in place (kernel, libc,
libc_r) and am working on debugging right now.

Regards,
--
Matt Emmerton



Re: Release Schedule for 1.8

2007-01-01 Thread Karthik Subramanian

:About testing, i think that we should start looking to do automated
:regression testing of the system as much as we can. TET[1] looks like
:a very good solution and FreeBSD is probably going to use it[2], so we
:could get a lot of regression tests for free that apply to DragonFly.
:
:[1]: http://tetworks.opengroup.org/
:[2]: http://wikitest.freebsd.org/TetIntegration

I dunno.  I'm always skeptical about such frameworks and wonder
whether porting them actually saves us any time over building
our own.  I do think testing will become a lot easier with the
virtual kernel giving us a totally self contained environment for
three files (kernel binary, memory image file, disk image file).


Just my 2 cents as a guy who wrote part of a test framework ages ago :)

I agree with Matt, rolling your own is the best way to go - simply
because it's very difficult to get hold of something that does
everything that you want, and does it well. Far better to write your
own test framework in Python or the like, and get away with it. It
doesn't take a long time, and it's worth the effort.

Is there someplace where there's a list of release-oriented code
bits/features, so that people can pick off stuff that they're
interested in testing? I'm moving to a different city/state in a week
or so, so i'm not sure if I can contribute to any testing right away -
but I'd like to take a look all the same.

Cheers,
K.