[03:08] <fabbione> jbailey: did you try to boot your sparc again or did you declare it dead?
[03:08] <jbailey> fabbione: I can boot it fine.  It just doesn't live through a gdb build.
[03:08] <jbailey> And it hung on the loopback self-test.
[03:08] <jbailey> fabbione: It's still sitting here plugged in, though.
[03:09] <fabbione> hmmm
[03:09] <fabbione> weird
[03:09] <jbailey> I had managed to get an oops out of it once, but not when the serial console was plugged into it.
[03:09] <fabbione> did you plug a net cable in it?
[03:09] <jbailey> Yes.
[03:09] <jbailey> Well
[03:09] <jbailey> Into one of the three nics.
[03:09] <fabbione> did you try to unplug it?
[03:10] <fabbione> i wonder if the network card is hunging the test
[03:10] <jbailey> No.
[03:10] <jbailey> I was thinking of trying to find an old stable release (like woody) and putting it on, and doing builds.
[03:10] <jbailey> Just to make sure.
[03:10] <jbailey> But I won't have time for that soon.
[03:10] <jbailey> I'm still worried that it could be just a gcc-4 issue or something.
[03:11] <fabbione> hold on
[03:11] <jbailey> I don't know how many other people are actively testing current dapper kernels. =)
[03:11] <fabbione> halt
[03:11] <fabbione> does it hang the boot or the hw?
[03:11] <jbailey> EPARSE
[03:11] <fabbione> i understood that the hw was broken
[03:11] <jbailey> Which it?
[03:11] <fabbione> dunno.. it sounded that way
[03:11] <jbailey> The machine boots fine.
[03:11] <fabbione> well you can still reinstall breezy and debug stuff :)
[03:11] <fabbione> just stay with breezy kernel ;)
[03:12] <jbailey> Installing breezy on this was such a pain in the arse.  *sigh*
[03:12] <fabbione> well dude
[03:12] <fabbione> it was a pain true
[03:13] <fabbione> now you admit to have a netcable in it
[03:13] <fabbione> so you can netinstall it
[03:13] <fabbione> and do it clean :)
[03:13] <jbailey> I'm not setup for netinstall.
[03:13] <jbailey> That's why I was testing CD installs for you.
[03:13] <jbailey> I don't have a separate network, and shutting up the dhcp server is too much work.
[03:14] <jbailey> That's why I Was thinking Woody.
[03:14] <fabbione> you do have a laptop. don't you?
[03:14] <jbailey> Yes.
[03:14] <jbailey> No crossover cable atm.
[03:14] <jbailey> I gave it back to infinity. =)
[03:14] <fabbione> ok.. and 3 nics in the sparc?
[03:14] <fabbione> ah ok
[03:14] <fabbione> hell dude.. you should have told me
[03:14] <fabbione> i have tons of crossovers and hubs
[03:15] <fabbione> ok listen.. deal.. buy a cross over or an extra hub.. i will pay it ;)
[03:15] <jbailey> Did Sparc participate in flight-1, btw?
[03:15] <fabbione> nope
[03:15] <fabbione> too busy building at that time
[03:16] <fabbione> i might get flight 2 assuming oo2 builds
[03:16] <fabbione> i got all of main builded
[03:16] <jbailey> Is 2.6.12-9 a good test, or do I need to go older?
[03:17] <fabbione> 2.6.12-9 is the kernel that we used for release
[03:17] <fabbione> it's ok, why?
[03:17] <jbailey> I'll try just setting that to the default boot.
[03:17] <fabbione> ah ok
[03:17] <fabbione> sure that should work if udev is not 100% nuts
[03:18] <jbailey> udev won't run on that kernel.
[03:18] <jbailey> It'll just refuse to start.
[03:27] <jbailey> I've just fired up a gdb build on the old kernel, we'll see how she does.
[03:28] <fabbione> cool
[03:29] <fabbione> remind me.. why did you need gdb 64 bit?
[03:30] <jbailey> I think case, to debug sparc64 tls. =)
[03:30] <jbailey> s/think/this/
[03:30] <jbailey> s/I/In/
[03:30] <fabbione> no need to
[03:30] <fabbione> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=341514
[03:30] <jbailey> I don't need to debug it?
[03:30] <fabbione> tls works fine on 64bit assuming you backport all the patches from David Miller
[03:30] <fabbione> that Debian did not
[03:31] <fabbione> we did the right thing disabling it
[03:31] <jbailey> Dude, what do you think I'm doing?
[03:31] <jbailey> I'm in the middle of backporting all of them, and I get a segfault.
[03:31] <fabbione> *cough*
[03:31] <fabbione> ok
[03:31] <jbailey> So now I'm debugging it before I upload it as a working version. =)
[03:31] <fabbione> david said also that tls makes sense on 32bit
[03:31] <fabbione> not on 64
[03:32] <fabbione> it's there almost only for accademic reasons
[03:32] <jbailey> Umm, that's not sensical.
[03:32] <fabbione> at least that's my understanding of what he said to me a couple of days ago
[03:32] <jbailey> NPTL requires TLS support.
[03:32] <jbailey> So if the goal is to eliminate the old LinuxThreads library, I need it working.
[03:32] <fabbione> fabbione i don't think we will push that for dapper
[03:32] <fabbione> davem right, sparc64 libc doesn't need to be TLS
[03:32] <fabbione> davem it's nice that 32-bit can be
[03:33] <fabbione> ok i can ask him for the patch than
[03:33] <jbailey> He won't have it.
[03:33] <jbailey> He did the work against current upstream.
[03:33] <jbailey> You'll be asking him to do the same work that I've mostly done.
[03:33] <fabbione> i didn't connect the 2 things till now
[03:34] <fabbione> right..
[03:34] <jbailey> But we need gdb to work for all the arch combos we can run anyway.
[03:34] <fabbione> well i am sorry.. i got to it too late
[03:34] <jbailey> So I'm getting biarch to work for sparc/sparc64 and ppc/ppc64.
[03:34] <fabbione> otherwise i would have asked him :/
[03:36] <jbailey> Well, but we can't expect Dave to carry the sparc port for us entirely.
[03:36] <fabbione> no no no
[03:36] <fabbione> i understand that
[03:36] <fabbione> but given he wrote most of that code
[03:37] <fabbione> for him doing a backport is much less expensive than for us
[03:37] <jbailey> Not much so, though.
[03:37] <fabbione> specially given he does use debian and ubuntu
[03:37] <fabbione> he cares about it
[03:37] <jbailey> Gettnig all the code was reasonably quick.
[03:37] <jbailey> DEbugging it ought to be once I have a debugger. =)
[03:37] <jbailey> My biggest limitation is waiting 12 hours for a build iteration. =)
[03:38] <jbailey> glibc 2.3 and 2.4 changed how their TLS stuff is done.
[03:38] <jbailey> So I ported all of the stuff to 2.3's style.
[03:38] <fabbione> ah i see
[03:38] <jbailey> I'm guess I've just missed an initialisation somewhere.
[03:39] <fabbione> evi
[03:39] <fabbione> l
[03:39] <jbailey> Right.
[03:39] <jbailey> Because it'll be in a chunk of asm.
[03:39] <fabbione> can we split glibc? ;)
[03:39] <jbailey> That's where he'd be able to go faster than I.
[03:39] <jbailey> He can read the asm at full speed.
[03:39] <fabbione> in a lot of tiny small readable sources?
[03:39] <jbailey> But getting to this point is much harder.
[03:39] <jbailey> It *is* small tiny readable sources. =)
[03:39] <jbailey> That's half the problem.
[03:39] <jbailey> In which frigging source file is the problem? =)
[03:39] <fabbione> eh
[03:40] <fabbione> exactly
[03:40] <jbailey> But gdb will tell me what part of the initialisation is segfaulting.
[03:40] <fabbione> right
[03:40] <jbailey> Then I'll do the trace against a modern glibc compile and track both backwards.
[03:41] <fabbione> you make me hot when you talk to me so dirty!
[03:41] <jbailey> *lol*
[03:41] <fabbione> ok time for a break
[03:42] <jbailey> fabbione: It might be worth asking Znarl if he can handle it - he's been really quick on things.
[05:50] <BenC> lamont: did you see what I said wrt elilo?
[05:51] <lamont__> BenC: did not.
[05:52] <BenC> scroll back in ubuntu-kernel
[05:53] <lamont__> BenC: doh
[05:53] <lamont__> thanks
[05:53] <lamont__> and awaiting your reply there
[05:58] <jbailey> fabbione: I notice that the gdb testsuite includes a number of thread failures on sparc.  I wonder how many of these are related to them making assumptions about debugging linuxthreads.
[05:58] <jbailey> fabbione: I should check the source to see what its assuming.
[06:29] <jbailey> gdb build finished on sparc.
[06:30] <jbailey> So that hardware could still be broken, but not exercised by 2.6.12, or 2.6.15 could be broken.
[06:30] <jbailey> No opinion offered. =)
[06:33] <BenC> none given either :)
[06:33] <BenC> I have no idea
[06:33] <BenC> it's an odd lockup
[06:33] <jbailey> BenC: Maybe when there's less snow, I can give you my patch to gdb and you can try building it?
[06:34] <jbailey> I don't know that it hangs the machine if I don't try a biarch gdb build.
[06:34] <BenC> if you can put it somewhere I can download it, I can do a build
[06:36] <jbailey> Sure.  I was more thinking that you don't want to have to go out there if it crashes. =)
[06:37] <jbailey> It's a 3 line patch to debian/rules.  Lemme put the .dsc/.diff.gz up somewhere.
[06:39] <BenC> it's ok, I can just flip the breaker for the whole barn from in the house, and reboot all the machines :)
[06:39] <jbailey> You have the whole thing on one circuit out there?
[06:39] <jbailey> Tell me it's a subpanel circuit. =)
[06:39] <jbailey> And not a 15amp. =
[06:39] <jbailey> )
[06:45] <jbailey> BenC: http://people.ubuntu.com/~jbailey/sparc/
[06:55] <BenC> yeah, it's a 100Amp breaker to a subpanel
[06:56] <BenC> might be a bit before I can do that gdb build
[06:57] <BenC> Fair Access Policy just kicked in because I was downloading edubuntu iso
[06:57] <jbailey> Fair Access Policy?
[06:57] <BenC> pay $120/month for business grade access, three times the cost of cable, and they didn't tell me about this FAP bullshit
[06:58] <jbailey> Hmm.  Direcway.  Isn't that DirecTv's Internet service?
[06:58] <BenC> yeah, if you consume high b/w over an extended period, they b/w limit you for awhile
[06:58] <BenC> yeah, I can't get normal broadband here
[06:58] <jbailey> Just funny.  I helped review some of the design specs for it.
[06:59] <jbailey> I wonder how different the design from back then was to what they have now.
[06:59] <jbailey> It was 1994. =)
[06:59] <BenC> other than FAP and the 400ms latency, it's not too bad
[06:59] <BenC> it's way different
[06:59] <BenC> I have high-power xmit, for .5mbs upload speed
[07:00] <jbailey> Ah, yeah.  The original design didn't include any up.  It was telephone connect for up.
[07:00] <BenC> you can't stand in front of my dish, else you risk eltromagnetic health hazards :)
[07:00] <BenC> first time I had it, they just started the 56kps upload to avoid the phone line
[07:00] <jbailey> But it was going to be basically 38,4k up, 100 megs down. =)
[07:00] <BenC> they've been building up business class service for high-power upload
[07:01] <BenC> shit, they don't sell 100mbs down, highest is 2mbs I believe
[07:01] <BenC> I max out at 300kbytes/sec
[07:01] <BenC> generally it's 200kbytes/sec
[07:01] <BenC> under FAP, I get 12kbytes/sec
[07:01] <jbailey> Hmm.  I wonder how much of that is just a limit on how fast you can ack packets.
[07:01] <jbailey> *ouch*
[07:02] <BenC> I think most of it has to do with dish size, the higher d/l rate you get, the larger the dish gets
[07:02] <jbailey> Yeah.  I just looked in awe at the specs that actually mentioned how the birds and dishes worked.
[07:03] <jbailey> Totally no clue on that side.
[07:03] <BenC> the xmit is configurable via the DW7000 modem/router, so they can bump that from the NOC
[07:03] <BenC> I think I'm at 2 watts on xmit
[07:04] <BenC> wish there was a way to bump that power without it telling the NOC :)
[07:04] <jbailey> So the day beore thanksgiving they up the power and roast you a turkey in flight? =)
[07:04] <BenC> lol
[07:05] <BenC> I actually thought these new DW7000's were supposed to be linux OS on the inside
[07:05] <BenC> Hughes had a Linux black-box beta test a few years ago
[07:05] <BenC> but they are VxWorks (I can telnet to it)
[07:06] <jbailey> It's interesting seeing Linux lose out a bit to VxWorks.
[07:06] <jbailey> I don't follow embedded that closely, but it seems to be turning up more and more.
[07:07] <BenC> probably had to do with real-time stuff
[07:07] <jbailey> ALso that an embedded more Linux kernel with all the size options turned on is still 700k uncompressed.
[07:07] <jbailey> (2.6.14)
[07:08] <BenC> might have been something with the GPL too
[07:08] <BenC> all the press that Linksys got about WRT54G's being hacked up probably scared them
[07:09] <jbailey> Yeah.
[07:09] <jbailey> It would be sad to see Linux bumped out of the embedded market.
[07:09] <jbailey> So many fun applications waiting for us. =)
[07:09] <BenC> yeah, I love doing embedded work
[07:10] <BenC> so much freedom
[07:10] <BenC> of course the only embedded system I ever did, I wrote the OS from scratch (multitasking, preemptive OS, with TCP/IP stack, video decode, and gigabit ethernet driver, in 512k RAM)
[07:10] <BenC> linux wouldn't be able to do that
[07:11] <jbailey> BenC: Anything you want me to do on this sparc box befor eI wander off?
[07:11] <jbailey> 512k, sweet.
[07:11] <BenC> nah, I'll get gdb compiling to see if I can reproduce
[07:13] <jbailey> Cool.  If you can't, I can try running test kernels or something that just spits out log information more often or something.
[07:13] <jbailey> I've got a serial console wired up to it now, so if there's anything that can be got usefully from that, lemme know.
[08:09] <fabbione> jbailey: ok
[08:12] <jbailey> EUNMATCHED: ok
[08:12] <jbailey> fabbione: Was that to the gdb thread failures comment?
[08:13] <fabbione> jbailey: yes
[08:13] <fabbione> (catching up on the backlog)
[08:13] <jbailey> Yeah, it was just far enough back I didn't remember what I'd said to you. =)
[08:14] <fabbione> no suriprise :)
[09:24] <jbailey> Hmm.
[09:24] <jbailey> BenC: Might still be my sparc hardware.
[09:24] <jbailey> BenC: It died while sitting idle.
[09:24] <jbailey> on 2.6.12-9
[09:25] <BenC> ah, sadly, I'm happy about that :)
[10:00] <jbailey> lamont, lamont__ : Any word on libc6-i386 for ia64 yet?
[11:21] <jbailey> lamont__: Thanks. =)