**** BEGIN LOGGING AT Wed May 25 23:59:56 2005
May 26 01:00:20 <beewoolie-afk> Knock knock?
May 26 10:35:50 <VoodooZ_Work> jbowler: It seems busybox is broken now.
May 26 10:36:27 <VoodooZ_Work> There seems to be an error in line 99: pkg_postinst_${PN} () {
May 26 11:21:28 <beewoolie-afk> Anyone around?
May 26 11:39:10 <beewoolie> Booting from power-off until I see the command prompt with openSlug and APEX takes about 42 seconds.  Looks like part of this time is dedicated to a sleep while the hdd wakes up.  
May 26 11:39:23 <beewoolie> Can anyone contribute a number for the redboot boot?
May 26 11:39:52 <jacques> i haven't timed it recently
May 26 11:40:18 <beewoolie> jacques: hey there man.
May 26 11:40:33 <beewoolie> It looks nice.  Now I just need to install some packages...
May 26 11:41:07 <beewoolie> I'm just curious about the boot time.  Apex doesn't perform the ramdisk copy, so it should be a little faster.  Plus, it doesn't have all of the timeouts that redboot has.
May 26 11:41:40 <beewoolie> jacques: Can you give me a pointer for downloading software?
May 26 11:41:45 <beewoolie> for openslug?
May 26 11:51:48 <jacques> what do you need?
May 26 11:59:44 <beewoolie> jacques: I found what I was looking for.  The feed stuff for ipkg.
May 26 12:00:02 <beewoolie> jacques: I'm going to have to build the packages I need.  I want to make this thing run bind9 and an email mta.
May 26 12:00:23 <beewoolie> ...and an ftp server and probably a light-weight web server.
May 26 12:02:00 <beewoolie> jacques: BTW.  It doesn't power-off when I halt nor does it reboot when I use the reboot command.  Are these supposed to work?
May 26 12:03:15 <jacques> they are supposed to work, and used to
May 26 12:03:22 <jacques> I think that's a recent regression
May 26 12:03:41 <beewoolie> I'm getting some IO errors on the HD as it shuts down.  Perhaps that's what's been hanging it up.
May 26 12:03:46 <jacques> hmm
May 26 12:04:07 <beewoolie> SCSI error : <0 0 0 0> return code = 0x70000
May 26 12:04:07 <beewoolie> end_request: I/O error, dev sda, sector 218628207
May 26 12:04:07 <beewoolie> Buffer I/O error on device sda1, logical block 27328518
May 26 12:04:07 <beewoolie> lost page write due to I/O error on sda1
May 26 12:04:37 <jacques> hmm, I haven't seen that before
May 26 12:06:06 <beewoolie> I pulled the drive from the slug and put it on the notebook.  Looks like the FS was cleanly unmounted.
May 26 12:06:15 <beewoolie> The slug is still not shutdown tho.
May 26 12:06:35 <beewoolie> This may be a USB driver issue on the slug.
May 26 12:06:55 <beewoolie> Or, I suppose, it could be caused by APEX.
May 26 12:17:07 <beewoolie> No problems with it on my laptop.  Hmm.
May 26 12:42:42 <[g2]-away> beewoolie, powerdown should work
May 26 12:46:23 <beewoolie> [g2]: Interesting.
May 26 12:46:29 <beewoolie> I wonder if it's an issue with USB.
May 26 12:46:51 <[g2]> beewoolie, sometimes the reboot and the shutdown hang
May 26 12:47:00 <[g2]> when's that last time you pulled ?
May 26 12:47:04 <beewoolie> Do you know why?
May 26 12:47:23 <beewoolie> I installed the binary firmware image since I was unable to build.
May 26 12:47:45 <beewoolie> OpenSlug-1.12-beta-firmware.zip
May 26 12:47:52 <[g2]> one of jb's theory was the jffs2 wasn't unmounted
May 26 12:48:35 <[g2]> there's been a bunch of changes with the startup and turnup procedures
May 26 12:48:46 <[g2]> so what's the build hangup ?
May 26 12:49:25 <beewoolie> It's the path problem.  For some reason, the build is ignoring the path I set and it is using the path from the default shell.
May 26 12:49:27 <beewoolie> Very bad,
May 26 12:49:48 <beewoolie> /dev/sda1 on /initrd type jffs2 (rw,noatime)
May 26 12:49:57 <beewoolie> That's in my mount table.  Very strange.
May 26 12:50:12 <[g2]> that's because of the pivot root right ?
May 26 12:50:31 <beewoolie> I suppose.  It should be unmounted.
May 26 12:50:47 <[g2]> or dmesg | grep command and see if you've got the old command line with the bogus initrd
May 26 12:51:13 <beewoolie> Kernel command line: console=ttyS0,115200 root=/dev/mtdblock4 rw rootfstype=jffs2 init=/linuxrc
May 26 12:51:18 <beewoolie> OK?
May 26 12:51:25 <[g2]> good
May 26 12:51:30 <beewoolie> it's what we put in the apex openslug-config
May 26 12:51:40 <beewoolie> BTW.  Nice work to you all.
May 26 12:51:44 <beewoolie> It's really quite good.
May 26 12:52:04 <[g2]> what's quite good ?
May 26 12:52:12 <beewoolie> openslug.
May 26 12:52:32 <beewoolie> What I did was break it into pieces the sizes of the unslug partitions.
May 26 12:52:50 <beewoolie> Then dd if=partition2.image of=/dev/mtdblock2 and so on.
May 26 12:53:05 <beewoolie> I wouldn't have done this without apex being loaded.
May 26 12:53:26 <beewoolie> BTW, do you know how long it takes the openslug image to boot?
May 26 12:56:01 <VoodooZ_Work> mine takes around 60seconds. too damn long! But that includes redboot
May 26 12:56:07 <[g2]> 42 seconds
May 26 12:56:22 <[g2]> but that might have been the first time with APEx
May 26 12:56:26 <beewoolie> 42 seconds is what apex takes.
May 26 12:56:30 <[g2]> subsequent boots are faster
May 26 12:56:38 <beewoolie> How is that?
May 26 12:56:42 <[g2]> Mine might be 30
May 26 12:56:50 <beewoolie> On the slug?  or the avilla?
May 26 12:56:55 <VoodooZ_Work> dropbear key generation
May 26 12:57:13 <beewoolie> Oh.  My successive boots are till 42 seconds.
May 26 12:57:19 <beewoolie> s/till/still/
May 26 12:57:45 <[g2]> it's been a while since I timed
May 26 12:58:00 <[g2]> Ah...
May 26 12:58:14 <[g2]> I'll bet beewoolie still copied the initrd area :)
May 26 12:58:16 <beewoolie> Also, this is with a serial console attached.
May 26 12:58:26 <[g2]> yeah me too
May 26 12:58:27 <beewoolie> [g2]: Nope.
May 26 12:58:40 <beewoolie> The openslug_config doesn't copy initrd.
May 26 12:58:46 <VoodooZ_Work> [g2]: Have you pulled and built today? busybox is broken. I think it's a side-effect of jbowler's changes.
May 26 12:58:55 <[g2]> so you are only coping the 1MB kernel and booting straight to jffs2
May 26 12:58:57 <VoodooZ_Work> Let him know if you see him.
May 26 12:59:12 <beewoolie> [g2]: Yeah.  I cannot show a log right now because I closed the box.
May 26 12:59:32 <[g2]> and what's the autodelay set to ?
May 26 12:59:36 <[g2]> 1 second
May 26 13:00:27 <beewoolie> [g2]: Yeah.
May 26 13:00:46 <beewoolie> I can make it shorter, but it's already almost too fast to catch.
May 26 13:01:06 <[g2]> 1 second is fine
May 26 13:01:11 <beewoolie> Most of the boot delay seems to come from the sda1 mounting.
May 26 13:01:52 <[g2]> Ahh... OK
May 26 13:02:05 <[g2]> you've still got the 10 second delay there
May 26 13:02:08 <beewoolie> is this an artifact of an older build.
May 26 13:02:10 <beewoolie> yeah.
May 26 13:02:16 <beewoolie> 10 second delay.
May 26 13:03:12 <[g2]> I think the old linuxrc waits 10 seconds on mounting the external disk before the pivot root
May 26 13:03:26 <beewoolie> OK.  Then it should boot in 32 seconds.
May 26 13:03:30 <beewoolie> That would be nice.
May 26 13:03:38 <[g2]> jb's new option has a sleep in there and is user configurable
May 26 13:04:01 <[g2]> are booting to flash stick ?
May 26 13:04:01 <beewoolie> Why have the delay?
May 26 13:04:18 <beewoolie> No.  I have a 120G drive in an old SCSI enclosure.
May 26 13:04:47 <[g2]> well at one time in the past we had modules that were loaded and we really needed to wait
May 26 13:05:08 <beewoolie> We use a monolithic kernel now?
May 26 13:05:08 <[g2]> the way things are now, the drivers are in the kernel and startup sooner
May 26 13:05:16 <beewoolie> Right.
May 26 13:05:30 <beewoolie> I'll be getting back to the jffs reader pretty soon.
May 26 13:05:31 <[g2]> all Linux kernels are monotlithic :)
May 26 13:05:48 <[g2]> monolithic :)
May 26 13:06:03 <beewoolie> I consider driver in modules as non-nonolithic.
May 26 13:06:14 <[g2]> I'm teasing
May 26 13:06:19 <beewoolie> oh.
May 26 13:06:21 <beewoolie> :-)
May 26 13:06:34 <[g2]> it's a classic linux versus microkernel thing
May 26 13:06:35 <beewoolie> Some people think that the Mach design is better.
May 26 13:06:52 <beewoolie> I'm not convinced since all of the better systems are BSD/Linux types.
May 26 13:06:58 <[g2]> so we've got lots of modules
May 26 13:07:20 <[g2]> it's just that usb/scsi are builtin
May 26 13:07:32 <beewoolie> the Mach thing  is extreme.  The paging code can be in user-space.
May 26 13:07:51 <beewoolie> I used to work on the Accent system, the precursor to Mach. 
May 26 13:07:56 * [g2] is a Linux Lover
May 26 13:07:57 <beewoolie> I thought it was a bloody mess.
May 26 13:08:12 <beewoolie> :-p
May 26 13:08:16 <[g2]> Mach and *BSD are dead in my book
May 26 13:08:49 <[g2]> I'm blown away with the Nokia 770 anouncement
May 26 13:08:52 <beewoolie> I read some good stuff about BSD the other day.  They've got some dedicated developers doing goof work.
May 26 13:08:59 <beewoolie> is that the Linux thing?
May 26 13:09:06 <beewoolie> Nokia supporting Linux?
May 26 13:09:16 <[g2]> Heh...
May 26 13:09:34 <[g2]> http://www.linuxdevices.com/news/NS5409534614.html
May 26 13:09:55 <[g2]> This is one of the biggest announcements in years to me
May 26 13:10:23 <[g2]> right up there with moto and others doing linux cell phones
May 26 13:10:55 <beewoolie> I didn't see that.  Very slick.
May 26 13:10:55 <[g2]> funny thing is koen in OE given a couple days will have most of it running in OE :)
May 26 13:11:09 <beewoolie> Indeed.
May 26 13:11:33 <prpplague> that is mallum's project
May 26 13:11:48 <[g2]> which project the 770 ?
May 26 13:11:56 <prpplague> yea
May 26 13:12:06 <[g2]> who's mallum ?
May 26 13:12:12 <[g2]> #elinux ?
May 26 13:12:15 <prpplague> yea
May 26 13:12:27 <[g2]> you know where I'm going :)
May 26 13:12:32 <beewoolie> No...
May 26 13:12:40 <[g2]> no ?
May 26 13:12:55 <prpplague> http://www.linuxdevices.com/news/NS8608661173.html
May 26 13:12:57 <beewoolie> I'm not kidding.
May 26 13:13:09 <[g2]> to chat with mallum :)
May 26 13:13:30 <prpplague> [g2]: don't pester him too much
May 26 13:13:31 <beewoolie> Are you going there now?
May 26 13:13:55 <[g2]> yeah... I won't pester him much
May 26 13:13:57 <beewoolie> I looked at matchbox and found that it didn't build easily.  I forget now why.
May 26 13:14:20 <prpplague> beewoolie: most of that has been fixed
May 26 13:15:01 <beewoolie> I haven't looked in a while.  There are other problems with the X solutions right now.  Way too much memory and way too complex to build.  The dependencies for gpe are staggering.
May 26 13:16:26 <[g2]> beewoolie, speaking of dependencies....
May 26 13:16:35 <beewoolie> ...
May 26 13:17:06 <[g2]> I need a really sharp guy like you to look at the runtime dependencies :)
May 26 13:17:28 <[g2]> or point me in the proper direction to figure some stuff out
May 26 13:17:29 <beewoolie> What of?
May 26 13:17:42 <beewoolie> I'm an espert at the moment.  :-)
May 26 13:18:05 <[g2]> Did I talk to you about the custom Knoppix LiveCD/DVDs I've done ?
May 26 13:19:59 <[g2]> I've got a custom DVD that's got everything to build Openslug/Unslung and the openslug-packages.
May 26 13:20:14 <[g2]> It just needs a swap space and disk space for "tmp"
May 26 13:27:40 <[g2]> beewoolie, so after a minor bug fix or two... I'd like to have the LiveDVD launch a web browser to the local machine running apache/php/mysql.
May 26 13:27:59 <beewoolie> OK.
May 26 13:28:02 <[g2]> The scripts will allow the users to build OpenSlug, Unslung via pull downs
May 26 13:28:14 <[g2]> Select libc (uclibc / glibc)
May 26 13:28:14 <beewoolie> do they have to have an HD available?
May 26 13:28:45 <[g2]> well we could to all pre-built and just packages stuff
May 26 13:28:56 <[g2]> but that isn't any different than the binary release
May 26 13:29:06 <beewoolie> Can you tell me what you're going for?
May 26 13:29:13 <[g2]> I'm thinking custom builds
May 26 13:29:44 <beewoolie> OK.  So this is a build CD.  Where is the output going to go?
May 26 13:30:49 <[g2]> it's a build DVD
May 26 13:31:00 <beewoolie> And the DVD is r/w?
May 26 13:31:04 <[g2]> the metadata is 100M and the tarballs are 350MB
May 26 13:31:33 <[g2]> no right now I mount a partition for swap and for build/"tmp"
May 26 13:31:50 <[g2]> A r/w would "work" but be very slow
May 26 13:32:00 <beewoolie> There's enough RAM to hold everything...?  I guess there should be.
May 26 13:32:16 <beewoolie> Since the target is so small.  But they could want to build a lot, no?
May 26 13:32:18 <[g2]> the "tmp" files are 3-5 GB
May 26 13:32:57 <beewoolie> That's a lot of RAM.
May 26 13:33:43 <[g2]> that's why one should have a disk :)
May 26 13:34:33 <beewoolie> Oh.  So you mount a partition for swap *and* you mount a partition for the output?  Or are you just using swap?  Really, that's not such a bad way to go.
May 26 13:34:54 <[g2]> Right now it's two
May 26 13:35:06 <beewoolie> I really like this idea.  It could go far.  Think about the setup of a build server.  Super easy.
May 26 13:35:07 <[g2]> but it could easily be 1
May 26 13:35:17 <[g2]> with a swap file on that partition
May 26 13:35:33 <[g2]> it seems the NTFS writing is 100% safe yet
May 26 13:35:36 <beewoolie> If the build prunes the build output as it goes, It could be reasonable to do this with just a swap partition.
May 26 13:35:38 <[g2]> which is an issue
May 26 13:35:53 <beewoolie> I wouldn't recommend that.
May 26 13:36:07 <[g2]> NTFS writing
May 26 13:36:22 <beewoolie> Not yet at least.  But all we need is a bunch of blocks.  We may not really care about saving it when the power is off.
May 26 13:36:51 <beewoolie> If the user can perform the build quickly and then fetch it with the web server, who cares about saving the result?
May 26 13:37:12 <[g2]> quickly is a relative term
May 26 13:37:27 <[g2]> openslug builds on my P4 in about 90 minutes from scratch
May 26 13:37:39 <beewoolie> If my build takes only 1.5 hours, there is no way that the slug build could take more than that.
May 26 13:38:08 <beewoolie> Hmm.  I wonder what is the hold-up.  I'm build about 50MiB of stuff.  Openslug is only 8MiB.
May 26 13:38:31 <[g2]> there's 350MB of tarballs
May 26 13:38:50 <beewoolie> Are you talking about the output packages?
May 26 13:39:03 <[g2]> it's both core and packages
May 26 13:39:25 <[g2]> but I mean the kernel is 36MB and GCC/binutions is another 25
May 26 13:39:34 <[g2]> binutils
May 26 13:40:12 <beewoolie> I'm talking about the runtime.  The kernel isn't 36MiB in the final form.
May 26 13:40:54 <[g2]> no our final form weighs in around 1MB kernel 16K bootloader and 4.2MB jffs2
May 26 13:40:59 <beewoolie> I'm building two compilers when I do this, the 2.95 and the 3.41.  I build support tools such as busybox, a heap of native packages to support the build, and then OPIE which is pretty big.
May 26 13:41:21 <beewoolie> That's my point.  How can it take so long to build it?  what is the clock speed of your p4?
May 26 13:41:39 <beewoolie> My machine is a 2GHz Athlon64.
May 26 13:42:04 <[g2]> 2.8 1G memory fast disks
May 26 13:42:18 <[g2]> I've got a quicker setup 
May 26 13:42:43 <beewoolie> Mine is rated at the 3200 level which is supposed to be faster than the P4 at 2.8.  No proof tho.
May 26 13:43:19 <beewoolie> I've got 1G of ram, too as well as a scsi raid5.
May 26 13:43:21 <[g2]> I've got 3200 nforce4 ultra too and it's feels faster but I haven't timed or build there yet
May 26 13:43:45 <beewoolie> Well, I'm just wondering if there is something that OE does that is taking longer.
May 26 13:43:56 <[g2]> I'm pullin 60MBs on hdparm on the maxtor driver with the 16MB cachce
May 26 13:43:57 <beewoolie> I discovered something relaly interesting when I tuned my builder.
May 26 13:44:14 <[g2]> it's got NCQ and the NForce4 ultra supports that
May 26 13:44:18 <beewoolie> Limiting the PATH to *only* include my approved directories made the build significantly faster.
May 26 13:44:50 <beewoolie> I did this in order to guarantee that I was building all of the support tools, but it had the side effect of making the build run faster.
May 26 13:44:53 <[g2]> do you build with ccache ?
May 26 13:44:57 <beewoolie> No.
May 26 13:45:07 <[g2]> then you're building slowly
May 26 13:45:08 <beewoolie> It's a waste unless you rebuild the same things over and over.
May 26 13:45:16 <[g2]> nod.
May 26 13:45:24 <[g2]> but most often you are rebuiding stuff
May 26 13:45:30 <beewoolie> No. It bloats disk space without helping me much.
May 26 13:45:42 <beewoolie> When I test, I want it to rebuild everything.
May 26 13:45:59 <beewoolie> I use cache files instead.  Once a package is built, it is archived and can be reused from there.
May 26 13:46:10 <[g2]> Ok... In OpenSlug we often only rebuild a handful of 350 packages 
May 26 13:46:15 <beewoolie> ccache is, IMHO, like the autotools.  Lots of hype, not necessarily worth the cost.
May 26 13:46:32 <beewoolie> why rebuild?  New source version?
May 26 13:46:49 <[g2]> it just like buiding modules in the kernel
May 26 13:46:54 <[g2]> you only rebuild what required
May 26 13:47:02 * [g2] is typing too fast
May 26 13:47:05 <beewoolie> The kernel is really smart about rebuilds and only recompiles what it needs.
May 26 13:49:53 <[g2]> well the point is the eventually I'd like to build some runtime depenencies via walk all the used objects and libraries
May 26 13:50:03 <[g2]> walking
May 26 13:50:11 <beewoolie> [g2]: that isn't hard.
May 26 13:50:28 <[g2]> with the LiveDVD all the stuff is built so it's just the walking of thats the issue
May 26 13:50:40 <beewoolie> Much of the info can be retrieved from ldd.
May 26 13:50:50 <beewoolie> I gotta take a break.
May 26 13:51:00 <beewoolie> ttfn
May 26 13:51:06 <[g2]> cheers!
**** BEGIN LOGGING AT Thu May 26 23:17:54 2005
**** ENDING LOGGING AT Thu May 26 23:59:56 2005