[02:23] <BenC> mjg59: I had a bug I wanted you to look at...trying to find it
[02:24] <BenC> 20979
[02:31] <infinity> BenC : Still want that module tested?
[02:32] <infinity> BenC : Should I be suspicous about the drastic file size difference? :)
[02:33] <BenC> nah, it's all good :)
[02:33] <BenC> he reports no difference, but I find that hard to believe
[02:36] <infinity> What did you change?
[02:37] <infinity> Other than shrinking the module. :)
[02:37] <infinity> Mess of printks all over the place with your name on them? :)
[02:37] <infinity> Oh, that's funny.
[02:38] <infinity> You know that assertion I was getting every time I ran update-initramfs?
[02:38] <infinity> Don't see that anymore... (of course, I get the spew of other erros instead)
[02:38] <infinity> Anyhow, ready to reboot.  Back in a few.
[02:43] <infinity> Ah-ha.
[02:43] <infinity> BenC : I guess you just rolled back to an older version of ata_piix? :)
[02:44] <infinity> Scary ATAPI errors gone, failed assertion back.
[02:46] <mjg59> BenC: Weird. Haven't seen that.
[02:56] <BenC> infinity: so the errors are definitely gone
[03:00] <BenC> infinity: can I add you to CC for this bug?
[03:02] <infinity> BenC : Sure.  I didn't add myself when I commented, but go ahead.
[05:50] <fabbione> morning
[06:25] <fabbione> hey BenC 
[06:25] <BenC> hey fabbione
[06:25] <fabbione> i am uploading the kernel right now
[06:25] <fabbione> i thought you went to sleep
[06:25] <BenC> my turn to be up at 3am :)
[06:25] <fabbione> yeah i know
[06:26] <fabbione> i need a smoke
[06:26] <BenC> yeah, me too
[06:27] <infinity> New kernel again?
[06:27] <infinity> New ABI too?
[06:27] <BenC> yep
[06:27] <infinity> Or just bugfixes?
[06:27] <infinity> Feh. :)
[06:27] <BenC> fabbione's kernel upload is sparc64 8.10 build
[06:27] <BenC> my kernel upload is waiting on hppa abi files so I don't ftbfs :)
[06:29] <infinity> !
[06:29] <infinity> Send me one!
[06:29] <BenC> there's a limit of one :)
[06:30] <BenC> 256Mb nvidia card too, 250Gig ATA drive
[06:30] <fabbione> BenC: ah new crack?
[06:30] <BenC> 2xdual-core-2.5ghz cpu's
[06:30] <fabbione> BenC: i know benh bought one
[06:31] <fabbione> the kernel still doesn't really support it in full :)
[06:31] <fabbione> and he is hacking heavily on it
[06:31] <BenC> it comes pre-installed with YDL
[06:31] <BenC> YDL and OSX
[06:31] <BenC> but I have an offer of one machine, and I think I want to get a 17" powerbook
[06:32] <fabbione> BenC: the powerbook i got is sweet
[06:32] <fabbione> but it's a bit fragile
[06:32] <fabbione> keybord is teh sux
[06:32] <fabbione> i need to get a new one already
[06:32] <BenC> damn, that does suck
[06:32] <fabbione> keys are falling off and some of them get stucked.. go figure
[06:33] <BenC> but I need a laptop...the one I had at UBZ, I just took it back and got a refund
[06:33] <BenC> and powerbook 17" display is nice
[06:33] <fabbione> oh yeah
[06:33] <fabbione> the display is COOL
[06:34] <BenC> plus it will give me a chance to test bcm43xx with the airport2 wireless
[06:34] <fabbione> ehhehe
[06:34] <infinity> I don't know very many powerbook owners who say they'd buy one again.
[06:34] <BenC> pitti loves his :)
[06:34] <fabbione> i wanted one becuase i needed a ppc
[06:35] <infinity> Well, Apple fanatics, maybe, but not in the FLOSS world, where we're more hardware agnostic.
[06:35] <fabbione> but yeah.. i heard that too
[06:35] <fabbione> specially from elmo
[06:35] <infinity> pitti is an exception to many rules. :)
[06:35] <fabbione> he had to change his PB.. dunno how many times
[06:35] <infinity> He's also an iBook user, which makes a qorld of different.
[06:35] <infinity> difference, too.
[06:35] <infinity> Only because the iBooks are much cheaper, so you don't feel so ripped off. :)
[06:36] <infinity> OTOH, if you like laying out lots of cash, everyone in the company (me included) seems to be deeply in love with IBM.
[06:36] <infinity> Go figure.
[06:37] <infinity> I'd recommend the G5 desktops over the PowerBooks anyday, if you want/need a PPC machine.
[06:37] <BenC> well, the pb will be at a reduced price too...get IBM to take %40 of the list price, and I'll go buy one :)
[06:37] <BenC> I already have a G5 and 2 G4's, so I'm not really in need of anymore ppc's, just that I like powerbook looks
[06:38] <fabbione> BenC: you got your bw back
[06:38] <BenC> thanks
[06:39] <fabbione> weird is that i managed to upload up to 40K
[06:39] <fabbione> sometimes even faster
[06:39] <BenC> really? damn, that's more like I am supposed to get
[06:39] <BenC> maybe it's the atmosphere tonight giving me better xmit signal :)
[06:39] <fabbione> linux-source-2.6.15_2.6.15-8.10_sparc.changes 100%   10KB  10.1KB/s   00:00    
[06:39] <fabbione> cdrom-core-modules-2.6.15-8-sparc64-di_2.6.15 100%   54KB  54.2KB/s   00:00    
[06:39] <fabbione> crc-modules-2.6.15-8-sparc64-di_2.6.15-8.10_s 100% 2376     2.3KB/s   00:00    
[06:39] <fabbione> ext2-modules-2.6.15-8-sparc64-di_2.6.15-8.10_ 100%   39KB  38.8KB/s   00:00    
[06:39] <fabbione> ext3-modules-2.6.15-8-sparc64-di_2.6.15-8.10_ 100%  107KB 107.3KB/s   00:00    
[06:39] <fabbione> fat-modules-2.6.15-8-sparc64-di_2.6.15-8.10_s 100%   40KB  39.9KB/s   00:00    
[06:39] <fabbione> ide-modules-2.6.15-8-sparc64-di_2.6.15-8.10_s 100%  101KB 101.4KB/s   00:00    
[06:39] <fabbione> ipv6-modules-2.6.15-8-sparc64-di_2.6.15-8.10_ 100%  144KB 144.2KB/s   00:00    
[06:39] <fabbione> kernel-image-2.6.15-8-sparc64-di_2.6.15-8.10_ 100% 1512KB  40.9KB/s   00:37    
[06:39] <fabbione> linux-headers-2.6.15-8-sparc64-smp_2.6.15-8.1 100%  719KB  51.4KB/s   00:14    
[06:39] <fabbione> linux-headers-2.6.15-8-sparc64_2.6.15-8.10_sp 100%  718KB  47.9KB/s   00:15    
[06:39] <fabbione> linux-headers-2.6.15-8_2.6.15-8.10_sparc.deb  100% 6379KB  36.7KB/s   02:54    
[06:39] <fabbione> linux-image-2.6.15-8-sparc64-smp_2.6.15-8.10_ 100%   13MB  36.2KB/s   06:06    
[06:39] <fabbione> linux-image-2.6.15-8-sparc64_2.6.15-8.10_spar 100%   13MB  36.2KB/s   06:02    
[06:40] <fabbione> loop-modules-2.6.15-8-sparc64-di_2.6.15-8.10_ 100% 7256     7.1KB/s   00:00    
[06:40] <fabbione> md-modules-2.6.15-8-sparc64-di_2.6.15-8.10_sp 100%  240KB 120.2KB/s   00:02    
[06:40] <fabbione> nfs-modules-2.6.15-8-sparc64-di_2.6.15-8.10_s 100%  206KB 205.9KB/s   00:00    
[06:40] <fabbione> nic-modules-2.6.15-8-sparc64-di_2.6.15-8.10_s 100% 1008KB  43.8KB/s   00:23    
[06:40] <fabbione> nic-shared-modules-2.6.15-8-sparc64-di_2.6.15 100%   10KB   9.9KB/s   00:00    
[06:40] <fabbione> parport-modules-2.6.15-8-sparc64-di_2.6.15-8. 100%   39KB  39.3KB/s   00:00    
[06:40] <fabbione> plip-modules-2.6.15-8-sparc64-di_2.6.15-8.10_ 100% 9364     9.1KB/s   00:00    
[06:40] <fabbione> ppp-modules-2.6.15-8-sparc64-di_2.6.15-8.10_s 100%   60KB  59.9KB/s   00:00    
[06:40] <fabbione> reiserfs-modules-2.6.15-8-sparc64-di_2.6.15-8 100%  158KB 158.3KB/s   00:00    
[06:40] <fabbione> MEH
[06:40] <fabbione> sorry
[06:40] <fabbione> i meant to paste only the -images-
[06:40] <fabbione> that are a bit more interesting than the small udebs
[06:41] <fabbione> ok good
[06:41] <fabbione> katie accepted.. or better.. stalled it in NEW
[06:57] <BenC> already copied the abi files
[06:57] <BenC> hppa takes so long to NEW
[07:01] <infinity> Well, we weren't building hppa CDs, so Kamion didn't do "Oh my god, the sky is falling, NEW, NEW, NEW" mojo on it.
[07:01] <infinity> I assume elmo just did it in his usual NEW runs, which are often, but not instant.
[07:18] <fabbione> yeah Kamion doesn't NEW SCC
[07:38] <fabbione> BenC: do you alraedy have the new kernel ready for upload or are you still working on it?
[07:38] <BenC> still working on it
[07:38] <BenC> should be ready Friday
[07:40] <fabbione> ok perfect
[09:50] <CataEnry> hi :)
[02:15] <zul> heylo
[02:16] <zul> BenC: ping pull
[02:16] <BenC> zul: pong push :)
[02:16] <BenC> already pulled you, just haven't merged
[02:17] <zul> ok...i updated the changelog though
[02:18] <zul> nifty
[02:23] <zul> i almost have  a kick ass build script going on
[02:57] <BenC> share :)
[02:58] <AcidPils> hi
[02:59] <AcidPils> i finally know why acx didnt work... it gets a link, not the ap but to my RTL8180L card 
[03:00] <AcidPils> head -> table
[03:01] <mjg59> AcidPils: Ah...
[03:03] <infinity> hahaha.
[03:03] <infinity> "oops"
[03:03] <AcidPils> :p
[03:08] <BenC> lol
[03:15] <zul> zul: i will when its ready
[03:17] <zul> doh..
[03:17] <zul> im not with it tonight
[03:17] <zul> today even
[03:20] <jbailey> zul: But you clearly mean that you won't get it together in time for tongiht ;)
[03:20] <zul> shaddup :)
[03:24] <jbailey> zul: I'm actually not on the phone at the moment. =)
[03:24] <jbailey> Since you're already clearly stoned, shall we chat? =)
[03:29] <zul> meh...
[03:29] <zul> so whats involved with the grub stuff 
[03:31] <jbailey> Mostly it's closing bugs with "No, grub sucks, won't fix"
[03:31] <zul> thats easy enough...any other stuff
[03:33] <jbailey> Occasionally there's real problems.
[03:34] <jbailey> Not often, though.  In practice we haven't done much to grub since warty, and systems have been installing fine.
[03:34] <jbailey> Future development is happening on grub2, which I want to package and upload to universe at least.
[03:34] <jbailey> But that's such a low priority project compared to my other stuff, I don't actually know that I'll get around to doing it.
[03:34] <zul> yeah but grub2 only supports a few file systems doesnt it?
[03:34] <zul> ill give it a go
[03:35] <zul> so ill start the grub stuff tonight sounds ok?
[03:35] <jbailey> It supports about a dozen now, iirc.
[04:28] <zul> build you joker build...ahahaha!!!!
[04:37] <zul> noooooo...build failed
[05:24] <jbailey> Hmmm.  My usb scanner doesn't make a friendly device for me.
[05:31] <zul> dapper or breezy?
[05:31] <jbailey> dapper.
[05:32] <jbailey> It looks like just udev rule confusion.
[05:32] <jbailey> If I /dev/usb/scanner0 to /dev/bus/usb/005/006 and set the group right, sane seems happier.
[05:33] <crispin> I didn't think you needed devices these days for sane, I don't believe that mine has a device, and it works fine
[05:33] <jbailey> Yup, works in the gimp.
[05:33] <jbailey> crispin: USB Scanner?  Check to see if you have a /dev/usb/scanner* or /dev/usbscan*
[05:34] <jbailey> Certainly mine didn't work until I had actually created the device.  sane-find-scanner saw it on the usb device scan, but scanimage -L didn't see it.
[05:34] <crispin> no usb* at all, but running "scanimage -L" now doesn't appear to show the scanner either
[05:35] <crispin> I haven't tried it in dapper yet, so you are probably right
[05:36] <jbailey> I'm not sure it worked in Breezy, but I don't know for sure.
[05:36] <jbailey> It's probably soething I should test on the liveCDs too.
[05:36] <jbailey> Kamion: Can you add that to a checkbox list on the livecds maybe?  Can you scan an image?
[05:47] <BenC> jbailey: do you know what the status is for ia64/klibc/initramfs?
[05:48] <jbailey> BenC: I've got the build deps on halley to look at it now and haven't had time.
[05:48] <BenC> what is the issue, I can take a look at it
[05:48] <BenC> ?
[05:49] <jbailey> segfault in klibc, apparently.  I haven't debugged it.
[05:49] <jbailey> If you enjoy debugging ia64 asm, you're *welcome* to take it from me.
[05:49] <BenC> is there a way to repro the segv from userspace?
[05:49] <BenC> like can I chroot to the initramfs when it's unpacked?
[05:50] <jbailey> You don't even need to do that, you can run them from the build tree.
[05:50] <jbailey> You can even add debugging symbols and all that.
[05:50] <BenC> ahh, even easier
[05:50] <jbailey> Then you need to break gdb on whatever that arch calls the starter function and use stepi to walk through the setup.
[05:50] <jbailey> I suspect that the problem is in there.
[05:50] <jbailey> Best to do it with upstream klibc rather than the packge.
[05:51] <jbailey> Upstream has made more changes which will also cause segfaults.
[05:51] <jbailey> Otherwise, I will get to it probably tomorrow.
[05:52] <jbailey> There are only a few places where klibc tends to fail.  Given that it works on amd64, I don't expect this ot be a place where it's a 64 bit issue.
[05:52] <jbailey> Given that cat segfaults, I think it's probably the startup code.
[05:52] <jbailey> REmember not to refer to glibc when fixing the bug.
[05:53] <jbailey> klibc isn't GPLd.
[05:53] <BenC> ok
[05:54] <jbailey> dannf has been helpful when I've had questiosn before, though.
[05:54] <BenC> should I use 2.6.15 headers?
[05:55] <jbailey> Yes, that's what I have installed in there.
[06:03] <BenC> jbailey: I just did a rebuild and "./utils/static.g/cat README" worked
[06:04] <jbailey> That's what someone had reported as failing. 
[06:04] <BenC> gzip works too
[06:04] <jbailey> Try fstype
[06:04] <jbailey> That uses a 5 arg syscall.
[06:04] <jbailey> (llseek)
[06:05] <BenC> works, recognizes my ext3 fs too
[06:05] <jbailey> Hmm, should be fine.  Suggest trying to boot with it.
[06:05] <BenC> ok
[06:05] <BenC> I'm using the 1.1.1 in the archive, FYI
[06:06] <BenC> maybe it just needed a gcc-4.0+linux-headers-2.6.15 rebuild :)
[06:06] <jbailey> It's always possible.
[06:06] <jbailey> do you have a local ia64 to try a reboot on?
[06:07] <BenC> yeah, I have an i2k that I did the build on
[06:07] <BenC> getting ready to test it, 15 minutes and we'll know
[06:08] <jbailey> Is that 15 minutes to get dressed to go out to the barn? =)
[06:08] <BenC> yes, pretty much :)
[06:09] <BenC> I can reboot from here and get out to the barn before elilo starts :)
[06:09] <BenC> wonder if vga16fb+usplash will work
[06:23] <BenC> the only thing I saw was "Illegal Instruction", then a modprobe usage message, and then the /dev/root failure
[06:23] <BenC> couldn't do much else, because keyboard doesn't work when it drops to a shell
[06:25] <BenC> not sure what causes the illegal instruction
[06:26] <Kamion> jbailey: it's in https://wiki.ubuntu.com/Testing/Long already
[06:26] <jbailey> BenC: I think there's a debug mode in there that will sh -x it for you.
[06:26] <jbailey> Or you can break it at the top and step it through by hand.
[06:27] <jbailey> Can you tell me about the /dev/root failure?
[06:27] <jbailey> There shouldn't be one.
[06:27] <jbailey> root= should be passed on the kernel command line.
[06:28] <jbailey> Kamion: Ah nice.  I didn't know this document.
[06:28] <BenC> it's the failure about /dev/root and /dev/.static/root or something similar
[06:28] <jbailey> BenC: Can you do a break in there and tell me what's in /proc/cmdline ?
[06:29] <jbailey> You should never see /dev/root messages when using anything other than lilo.
[06:29] <BenC> like when the scsi module doesn't get loaded, and the device doesn't exist
[06:29] <jbailey> And it's just too broken to try fixing.
[06:29] <BenC> that's not from the kernel
[06:29] <Kamion> jbailey: it's new
[06:29] <BenC> it's from the initramfs, drops to busybox shell
[06:29] <jbailey> /dev/root is the node that we generate when we have to just mount the device by major:minor and don't know its name.
[06:30] <jbailey> BenC: Right, I know.
[06:30] <jbailey> Kamion: I'll forgive my own lack of omniscience this time then ;)
[06:30] <BenC> what's the kernel command line option to get the initramfs to stop?
[06:30] <jbailey> But next time!  Out the third floor window I go.
[06:30] <jbailey> BenC: break
[06:30] <BenC> ok
[06:31] <BenC> break, usplash...
[06:31] <BenC> I know for one I hate hearing about how "nousb" doesn't work on our system :)
[06:32] <jbailey> BenC: My suggestion is that we do it by taking over the world.
[06:32] <jbailey> BenC: What do you propose? =)
[06:32] <BenC> if they had called it redhat/nousb, then we'd have an argument
[06:32] <BenC> jbailey: world domination negates my argument :)
[06:33] <jbailey> Given that it's not a techincal problem, I think perhaps we shouldn't take a technical solution.  Let's try my idea first. =)
[06:35] <crispin> jbailey: fyi, my scanner problems in dapper are now bug 21061
[06:35] <jbailey> crimsun: sane is in main, the bug should probably go in bugzilla.
[06:36] <crispin> err, the name's crispin, and it is ....
[06:38] <jbailey> Well, the first three letters are the same, the last is the same, they're the same length and it contains an s.
[06:38] <jbailey> I don't even notice when I've gotten it wrong. =)
[06:39] <crispin> heh, isn't tab completion wonderful (except when it gets it wrong) :-)
[06:41] <jbailey> Try working for mdz and working with mdy. =)
[06:41] <maks_> vorlon said that the statfs from klibc is not 64 bit ready
[06:41] <maks_> altought it works on amd64 it fails on alpha
[06:41] <jbailey> maks_: Did he say why it is?
[06:42] <makx> 10:37 <vorlon> yeah, then I wonder what their kernel people were smoking when 
[06:42] <makx>                they did their statfs implementation. :)
[06:42] <makx> 10:31 <vorlon> cool.  Looks like there's more than one bug in klibc on alpha, 
[06:42] <makx>                though; fixing struct statfs gets me one step farther, but then 
[06:42] <makx>                the darn thing fails trying to clear the rootfs.
[06:42] <makx> 10:32 <vorlon> (which is basically a bunch of unlink() and rmdir() calls, so 
[06:42] <makx>                you'd think this wouldn't be a problem. :P)
[06:42] <jbailey> Ooo, is he hacking on this?
[06:42] <jbailey> BenC: Might be worht syncing up with Steve.
[06:43] <makx> yeah he said that his patch is not yet amd64 safe so didn't send in yet.
[06:45] <makx> but seems that you are chasing the same bug
[06:48] <BenC> stupid ia64 kernel doesn't have AT keyboard built-in
[06:48] <BenC> I need to rebuild a kernel before I can do more testing
[06:49] <jbailey> BenC: AT keyboard?  Not USB?
[06:49] <BenC> nah, I have a ps2 keyboard on it
[06:49] <jbailey> BenC: If you can find a way of probing that in sysfs, there's no reason not to just load it in the initramfs.
[06:50] <BenC> the installer loads it
[06:50] <jbailey> BenC: But I prefer to have everything in modules if possible. =)
[06:50] <jbailey> Right, I was more thinking for boottime detection if there's a port there.
[06:50] <BenC> if it's built-in in i386 and amd64, I want to just match it
[06:50] <jbailey> Right.  Perhaps it shouldn't be built in there either, though?
[06:50] <jbailey> I don't know how many new machines even have ps2 ports anymore.
[06:51] <BenC> ia64 is the only one of hte 6 that has it modular
[06:53] <BenC> I don't know if there's a way to detect it, and the installer for ia64 just loads it regardless, and it was added to /etc/modules after the install regardless
[06:53] <BenC> there's no test, it just does it
[06:53] <BenC> plus, if modprobe is failing in initramfs for my ia64 like I think it is, then I have to build it in
[06:55] <BenC> I'm going to try adding atkbd to /etc/mkinitramfs/modules first and see if that gets me any further
[06:55] <BenC> will it still load modules with "break"?
[06:55] <BenC> if not, that's another good reason for it being built-in :)
[06:57] <BenC> jbailey: if statfs was failing like vorlon said, wouldn't fstype fail?
[06:57] <jbailey> BenC: I don't think fstype actually stats the node.  It just llseeks it.
[06:57] <BenC> ah, true, it reads stdin
[06:57] <jbailey> Right.
[06:59] <makx> BenC: if you look at init you have several stages of "break=$stage"
[07:00] <BenC> makx: ok, thanks
[07:06] <BenC> jbailey: statfs() call in run-init succeeds
[07:07] <jbailey> BenC: And returned sane data?
[07:07] <BenC> tested it by calling manually (not on boot)
[07:07] <BenC> yes
[07:07] <jbailey> I wonder if you can safely step through run-init in a chroot?
[07:09] <BenC> returned 0xef53 for f_type, which is ext3, which is correct
[07:11] <makx> how did you test BenC?
[07:11] <BenC> and f_namelen=255, which is correct (and is the last memebr of the statfs struct, so proves it's not pushing 32-bit/64-bit values in the wrong places
[07:11] <jbailey> Right.
[07:11] <jbailey> Hmm
[07:12] <BenC> ./utils/static.g/run-init /mnt /sbin/init
[07:12] <jbailey> Can you drop strace in the initramfs and just watch the strace for stange values?
[07:12] <BenC> where /mnt existed, and /init was /bin/true
[07:12] <BenC> I don't think the Illegal Instruct is coming from the klibc stuff
[07:12] <BenC> I think it may be busybox
[07:13] <BenC> perhaps busybox needs to be rebuilt for ia64?
[07:13] <jbailey> Wait a sec.  
[07:13] <makx> hmm only static klibc is said to work on alpha
[07:13] <jbailey> We never actually demonstrated there was a problem on ia64 for run-init.
[07:13] <BenC> so maybe it's an issue of dynamic linking?
[07:14] <jbailey> The /dev/root messages mean that initramfs-tools is getting bad info from elilo
[07:14] <makx> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=341181
[07:14] <jbailey> BenC: Can you cat /proc/cmdline for me?
[07:16] <BenC> root=/dev/sda2
[07:16] <BenC> jbailey: no, the /dev/root is because the scsi module never gets loaded
[07:17] <jbailey> Ah, okay.
[07:17] <BenC> so /dev/sda neven exists
[07:17] <jbailey> Try adding it to /etc/initramfs/modules, rebuild and see if it  works?
[07:17] <BenC> modprobe fails just before that error
[07:17] <BenC> it spits out a usage message
[07:17] <BenC> just after the Illegal Instruction
[07:18] <jbailey> The modprobe should be just the system modprobe, I think.
[07:18] <makx> can you modprobe from shell?
[07:18] <BenC> I think the failed modprobe is it trying ot load the scsi module
[07:18] <BenC> shell doesn't work without atkbd, which is a module too :)
[07:19] <BenC> jbailey: busybox doesn't have it's own? I can't see the system one working with klibc :)
[07:19] <jbailey> There's glibc in the initramfs, though.
[07:19] <jbailey> The system one is linked against that.
[07:20] <BenC> hmm, what's the point of klibc is glibc is there too?
[07:20] <jbailey> The idea is that eventually everything should be against klibc and it should be possible to eliminate glibc.
[07:20] <jbailey> But correctness over optimisation.
[07:20] <jbailey> klibc is sufficiently small that it doesn't actually matter that both are there for now, and there's some reall convenient tools that come with it.
[07:21] <makx> would ease dist upgrades as no glibc dependency for kernel
[07:21] <makx> i mean indirect dep
[07:21] <BenC> how can I run klibc shared progs from normal userspace?
[07:21] <makx>  sudo chroot /usr/lib/klibc /bin/cat
[07:22] <BenC> segv
[07:22] <BenC> so it is a shared lib issue
[07:24] <BenC> wonder how I can run gdb for this
[07:24] <jbailey> On the shared lib?  No idea.
[07:24] <jbailey> gdb won't understand their shared library style, it's not elf.
[07:27] <makx> strace that dannf posted doesn't show anything too..
[07:28] <jbailey> makx: It really sounds like there's nothing wrong with the static libs.
[07:28] <jbailey> Or with the syscall mechanism.
[07:28] <jbailey> Probably just shared library initialisation.
[07:28] <BenC> it segv's immediately following execve of /bin/cat
[07:29] <BenC> I've replaced the shared klibc-utils with static ones, let's see if that changes anything
[07:29] <BenC> if I can boot with that, we know it's shared lib issues
[07:29] <jbailey> Cool.
[07:29] <jbailey> I've also just started a build here to look as well.
[07:29] <jbailey> But making lunch, too.
[07:30] <BenC> "We've secretly replaced the ia64's shared binaries with static ones...let's see if it notices"
[07:31] <makx> dannf reproduced it with a shared build by gcc-3.3
[07:35] <zul> BenC: lol
[07:37] <BenC> ok, that made a big difference, still not there, but a lot further along
[07:38] <BenC> here's what I noticed this time, my scsi/ide modules got loaded, and it got to the point of mounting the roofs, and then the SIGILL happened
[07:38] <BenC> I think the SIGILL isn't klibc related
[07:38] <BenC> unless there's some other tool in the initramfs that is compiled against klibc (not from klibc-utils)
[07:39] <BenC> what else uses klibc?
[07:40] <jbailey> Nope, just klibc itself.
[07:40] <BenC> is mount busybox and does mkinitrd use busybox too?
[07:41] <jbailey> Mmm.  mount is probably busybox.
[07:41] <jbailey> It's a simple syscall.
[07:42] <jbailey> Do you mean does the mkinitramfs tool itslef use busybox?
[07:42] <jbailey> No, it uses posix shell #!/bin/sh
[07:43] <BenC> calling busybox mount from userspace works
[07:43] <jbailey> BenC: Can you break in the initramfs and step through the rest of it by hand?
[07:43] <BenC> if keyboard worked
[07:44] <BenC> which means I'll need to do a recompile
[07:44] <jbailey> Add the keyboard modules to /etc/initramfs/modules and reboot
[07:44] <jbailey> well, regenerate the initramfs.
[07:44] <BenC> I did
[07:44] <BenC> didn't work
[07:44] <makx> you need to load it too
[07:45] <makx> add an modprobe somewhere at the beginning of init
[07:46] <BenC> ok
[07:46] <jbailey> Things in the modules file do get modprobed...
[07:46] <jbailey> You just need to make sure you break after that point. =)
[07:46] <makx> aahh zut forget that indeed.
[07:47] <makx> but atkb or i8042 dont get modprobed by initramfs if build modular atm.
[07:58] <jbailey> They should if included in the modules file.
[07:58] <jbailey> Everything in there should get modprobed.
[08:00] <makx> sure was speaking out of the box.
[08:24] <BenC> you know what would be super neato?
[08:25] <BenC> if I remembered you have to run elilo after updating initrd's
[08:25] <BenC> after that it booted
[08:25] <BenC> so it's just shared vs. static in klibc that is stopping ia64 from using initramfs
[08:27] <makx> wasn't there an optimisation patch that would only unzip some bytes of the initramfs head
[08:28] <BenC> maybe, haven't checked
[08:28] <BenC> btw, just adding atkbd doesn't do much without some more modules
[08:28] <BenC> just going to make it static
[08:28] <BenC> built-in I mean
[08:28] <jbailey> BenC: Hmm.  so it's not declaring dependancies correctly?
[08:29] <BenC> well, it depends on enough to load, but not really present itself as useful
[08:29] <BenC> either that or it wasn't really loading
[08:29] <jbailey> Hmm. Have to rerun elilo?
[08:29] <BenC> I put "modprobe atkbd" right after depmod -a in  /init, and break=top, and it didn't work
[08:29] <jbailey> do you think that update-initramfs should have a hook for that in it?
[08:29] <BenC> can't check if it was loaded
[08:30] <BenC> not sure, I mentioned that to lamont
[08:30] <BenC> I think hppa will have to have the same
[08:30] <jbailey> Right, for palo.
[08:30] <jbailey> So does lilo FWIW.
[08:31] <makx> btw on alpha we dont land into console too
[08:31] <makx> even with atkb build in.
[08:31] <makx> i assumed that to be a bb bug.
[08:32] <BenC> how do shared klibc bins declare the dynamic loader?
[08:32] <BenC> I think other things need to be built-in too, like serio
[08:33] <jbailey> BenC: I think they declare klibc.so to be the dynamic loader itself.
[08:34] <BenC> jbailey: I don't see anything about it from objdump
[08:34] <jbailey> BenC: readelf -a will tell you.
[08:34] <jbailey> I think objdump assumes too much.
[08:35] <BenC> if I do strings, I see it
[08:35] <jbailey> HIghtech debugging tools. =)
[08:35] <fabbione> hey guys
[08:35] <jbailey> I think I need to go lie down and get this headache gone.
[08:35] <jbailey> back in a short while.
[08:36] <makx> bonne sante' jbailey
[08:41] <BenC> hey fabbione
[08:45] <zul> how did the thing go?
[08:47] <fabbione> very well
[08:47] <zul> good to hear
[09:11] <zul> jbailey: pinger
[09:13] <jbailey> zul: On phone.
[09:14] <BenC> data point, shared bins on ia64 atleast get to __libc_init
[09:15] <jbailey> How are you tracing it? 
[09:17] <BenC> write(0, "DBG-X\n", 6); in libc_init.c
[09:18] <BenC> changeing X as I add more
[09:18] <BenC> err, 1
[09:26] <zul> jbailey: grub2 is already in universe
[09:28] <BenC> segv occurs right when libc_init calls main()
[09:29] <BenC> which it does differently for shared than static
[09:29] <BenC> for static, it just calls the main symbol directly, but for shared, it grabs the point to main from the pointer just after the env vars
[09:29] <BenC> and that pointer is correct, as far as readelf says (they match)
[09:30] <BenC> but it's wrong when you take relocation into account
[09:39] <jbailey> zul: Really out of date, though.
[09:41] <BenC> aha, I found the cause!
[09:42] <BenC> well, found the symptom that actually makes it crash
[09:42] <BenC> the correct address for main is 0x4000000000000ec0
[09:42] <BenC> note the 0x4 to start
[09:43] <BenC> it is only seeing, 0xec0
[09:44] <jbailey> Interesting.
[09:44] <BenC> nm
[09:44] <BenC> my printf was using %x and not %lx
[09:45] <BenC> so it has the correct address, why is it crashing when jumping to it
[09:50] <zul> jbailey: okie dokie ill re-write it then ;)
[09:50] <jbailey> zul: =)
[09:51] <jbailey> I don't remember if the packaging was any good or not.
[09:51] <zul> its using cdbs
[09:51] <jbailey> There are good ways to use cdbs and ugly ways to use cdbs, my friend.
[09:51] <zul> meh...wouldnt know..ill just do what i know
[09:51] <fabbione> it's *CDBS*! what are the good ways?
[09:52] <jbailey> fabbione: Come over here with a tube of vasseli... wait.  This is a public channel isn't it?
[09:52] <fabbione> ahaha
[09:53] <zul> and a family channel fuckers
[09:55] <fabbione> you guys are so tempting me to goatse all of you
[09:56] <paulproteus|lapt> Is "mem=128M" supposed to work?  It seems not to have any effect on powerpc at least with linux-2.6.15-8 .  See http://bugzilla.ubuntu.com/show_bug.cgi?id=21073
[09:57] <BenC> think I figured it out jeff
[09:57] <jbailey> !!! =)
[09:57] <BenC> compare klibc/arch/{x86_64,i386}/MCONFIG with the ia64 one
[09:58] <BenC> you'll see there's no -T text section var
[09:58] <BenC> klibc.so and the binaries code space are colliding
[10:00] <BenC> doing a recompile to see if I can guess a good location for it
[10:02] <jbailey> Ah, interesting.
[10:03] <zul> fabbione: dont you mean goatcheese?
[10:03] <fabbione> no
[10:03] <fabbione> i mean goatse
[10:05] <jbailey> zul: You don't want the goatse cheese.
[10:05] <jbailey> It's a little.. ripe.
[10:06] <zul> hehe..
[10:06] <zul> baahh..
[10:07] <fabbione> ahah
[10:07] <fabbione> night guys
[10:08] <zul> later
[10:08] <zul> im going to head home as well
[10:08] <jbailey> BenC: Any luck?
[10:09] <zul> later
[10:09] <BenC> jbailey: just trying to find the right location, I'm trying glibc's libc.so location for trial
[10:25] <BenC> not having any luck finding the right offset
[10:33] <jbailey> BenC: Maybe ask in #parisc?
[10:34] <BenC> ia64? :)
[10:34] <jbailey> It's all the same people.
[10:35] <jbailey> dannf hangs out in #parisc, he's the one who fixed elilo
[10:35] <jbailey> Or lamont, lamont-away might be able to recommend a person.
[10:46] <BenC> can't find a #parisc channel
[10:53] <BenC> paulproteus|lapt: mem= is arch specific, it's not supposed to work
[10:53] <paulproteus|lapt> BenC: I see, that's sad.  Is there any ppc way to limit how much memory the ppc kernel uses, then?
[10:54] <BenC> the bcm43xx thing is a non-issue for ppc
[10:54] <paulproteus|lapt> BenC: "non-issue"?
[10:54] <BenC> limiting memory like that is mainly for highmem bugs on i386
[10:54] <BenC> 1024megs isn't highmem on ppc
[10:54] <BenC> so that isn't your problem
[10:55] <paulproteus|lapt> Hmm, if you say so.  The latest Apple update for the Airport Extreme driver seemed to mention something about >1024M, so I figured it was maybe a bcm43xx problem instead of a PPC problem.
[10:55] <paulproteus|lapt> Or a problem with the DMA code written by the bcm43xx people.
[10:55] <BenC> just remove a memory stick :)
[10:56] <paulproteus|lapt> BenC: After all the trouble I had with trying to put one in, I kinda don't want to do that. :)
[10:59] <jbailey> BenC: OFTC.
[11:39] <paulproteus|lapt> BenC: I marked the bug as INVALID since mem= is irrelevant for this arch.  "Thanks" ;-).
[12:00] <makx> vorlon found that is supposed to be pulling the errno from a1, which is *true* for most other calls; for unlink() the errno is apparently in v0, which is normally the return value (and glibc seems to handle this fine).
[12:00] <makx> so it stomps over the false register on alpha for run-init..
[12:00] <makx> so the shared failure seems very diff.
[12:00] <jbailey> makx: Okay.  It's easy enough to write a little wrapper around it that grabs errno from the other register just for unlink.
[12:01] <jbailey> makx: What channel is he hacking in?
[12:01] <makx> oftc #debian-kernel