[12:06] kermitX_, hmm, that be some optics problems [12:14] optics? [12:14] optics is how something appears. Marketing term [12:14] oh [12:14] I was thinking optics as in laser optics :-) [12:14] something like landscape-client may actually be a godo thing and be gpl'ed, but lack of information means that there is a lot of FUD flying around [12:15] personally, I don't see why they don't just announce what it is, now that they have effectively announced it via dapper-changes [12:15] Oh for goodness sake... [12:16] Somebody in #ubuntu is complaining that there is no easy facility to downgrade. [12:16] no, dpkg doesn't support it [12:16] Fujitsu: that's a long standing problem with dpkg, i think [12:17] not that anyone would really want to downgrade. Dapper is so awesome ;-) [12:17] He's doesn't want to accept that he can't downgrade... [12:17] How silly. [12:17] it is a legit use case, just a hard one to implement === DShepherd [n=dwight@port0246-afm-adsl.cwjamaica.com] has left #ubuntu-devel [] [12:18] rolling back upgrades to packages would be tricky to implement. [12:18] We don't support it because it makes packaging instantly 500% harder. [12:19] Very difficult to implement... [12:19] dpkg DOES support downgrades. [12:19] And very silly. [12:19] Individual maintainer scripts don't. [12:19] apt-get dist-downgrade... [12:19] (And Debian Policy clearly tells us we don't have to) [12:20] Heheh. [12:20] Whe nyou have maintainer scripts doing whacky transitional stuff in maintainer scripts (moving files and directories, converting directories to symlinks and vice-versa), your job gets infinitely harder when you have to cope with all possible cases of doing that BACKWARDS. === poningru [n=poningru@ip24-250-225-41.ga.at.cox.net] has joined #ubuntu-devel === LaserJock throws the Debian record player in reverse [12:21] yep, I here breakage [12:21] *hear === ptlo [n=senko@83.131.73.184] has joined #ubuntu-devel [12:24] hey ptlo [12:24] heya neuralis! long time no see :-) [12:25] 'sup, how's the olpc python thingy working out? [12:26] ptlo: i'm in your general area; we should find time to hoist a few pints [12:29] you're on the same continent again? cool :) yeah, i'm looking forward to it === lmanul [n=manu@dan75-4-82-239-58-38.fbx.proxad.net] has joined #ubuntu-devel === zenrox [n=zenrox@71.115.198.118] has joined #ubuntu-devel === _TomB [n=ownthebo@ACCA062F.ipt.aol.com] has joined #ubuntu-devel === dinda [n=dinda@cpe-72-181-85-40.houston.res.rr.com] has joined #ubuntu-devel === sistpoty [n=sistpoty@ubuntu/member/sistpoty] has joined #ubuntu-devel === dinda [n=dinda@cpe-72-181-85-40.houston.res.rr.com] has left #ubuntu-devel [] === imbrandon [n=ubuntu@unaffiliated/imbrandon] has joined #ubuntu-devel === bmonty_away is now known as bmonty === FunnyLookinHat [n=funnyloo@71.57.11.218] has joined #ubuntu-devel === spacey [n=herman@ubuntu/member/spacey] has joined #ubuntu-devel === bddebian [n=bdefrees@71.224.172.103] has joined #ubuntu-devel [01:28] Howdy [01:30] hi [01:30] hello === jane_ [n=JaneW@dsl-146-167-152.telkomadsl.co.za] has joined #ubuntu-devel === Toadstool is now known as ToadZzZztool === j_ack [n=nico@p508D8A76.dip0.t-ipconnect.de] has joined #ubuntu-devel === spacey [n=herman@ubuntu/member/spacey] has joined #ubuntu-devel === _TomB [n=ownthebo@ACCA062F.ipt.aol.com] has joined #ubuntu-devel === _TomB [n=ownthebo@ACCA062F.ipt.aol.com] has joined #ubuntu-devel === spacey [n=herman@ubuntu/member/spacey] has joined #ubuntu-devel === _TomB [n=ownthebo@ACCA062F.ipt.aol.com] has joined #ubuntu-devel === hollowlife1987 [n=hollow@12-214-79-58.client.mchsi.com] has joined #ubuntu-devel === ompaul [n=ompaul@ubuntu/member/ompaul] has joined #ubuntu-devel === spacey [n=herman@ubuntu/member/spacey] has joined #ubuntu-devel [02:48] does anyone have software-controlled brightness? [02:48] for their laptop LCD [02:50] desrt: My Thinkpad does. [02:50] how does gnome handle it? === desrt just wrote a C program to control brightness on macbook but doesn't know the "proper" way to tell gnome about it [02:51] Look at gnome-power-manager's source. [02:51] I assume it uses hal. [02:51] Since there's no standard way at the kernel level. [02:52] (Mine, for instance, is in /proc/acpi/ibm, I think. [02:52] ) [02:53] how does g-p-m grab the keyboard keys? [02:54] It doesn't... That's lower-level... hotkey-setup, probably. [02:55] neat. [02:56] I just happen to know that g-p-m twiddles brightness for other reasons, hence the suggestion to look there for how to do it via hal. [02:56] thanks for the pointer [02:57] Heya infinity [02:57] Good work man! :-) [02:58] With anyhting in particular, or just in general? === mdke passes on another generic pat on the back for infinity [03:00] Meh, if you all loved me, you mail me some screwdrivers that don't strip the heads on the first use. === infinity grumbles. [03:00] Haha, [03:00] My kingdom for a hardware store in Australia that doesn't sell third-world crap. === Fujitsu gives infinity some screwdrivers. === Fujitsu hails infinity. [03:00] oh wow [03:00] Hmm. [03:00] hal itself watches the buttons [03:00] You're Adam Conrad, aren't you? [03:00] elite!! [03:00] In Melbourne? [03:01] Fujitsu: Last time I checked. [03:01] infinity, aha. Right near me... [03:01] infinity: why do you find yourself in melbourne anyways? [03:01] :O [03:01] Did mdke just insult Melbourne!? [03:01] no, it looks nice on the tv [03:02] but I was just wondering === bluefoxicy [n=bluefox@c-68-33-112-13.hsd1.md.comcast.net] has joined #ubuntu-devel [03:02] hey nice release with dapper. Looking forward to Edgy devel. [03:02] Anyway. [03:03] I'm looking to take Dapper right now and run some sort of program that magically tells me where %eip goes during execution and graphs it, without doing much more work than install-run-graph [03:03] mdke: I moved here from Canadia because my girlfriend is Australian. [03:03] ... [03:04] Wow, I'm from Canada as well. === licio [n=licio@ubuntu/member/licio] has joined #ubuntu-devel [03:04] Hi, licio. [03:04] It occurs to me this is the only place I'm likely to get a hint on where to look for profiling stuff on Ubuntu. Any ideas? or is this more "rebuild the whole system, patch the kernel, run this analysis software, gather these output files, look at these in this way, run this parser" [03:04] infinity: ah. I googled you yesterday and found your girlfriend's site. it's scary [03:04] hi Fujitsu [03:04] mdke: It doesn't scare me so much. :P [03:04] (in particular I want to see how much time is spent in shared objects versus the main executable; this could be VERY SIGNIFICANT in making an argument on building Edgy with position independent executables or working on that for Edgy+1) [03:05] infinity: glad to hear it [03:05] infinity: my boyfriend is from canada :o [03:05] he just moved back to alberta :( [03:05] from ottawa, so besides 2 timezone difference not much has changed [03:06] infinity: i take it that isn't your still beating heart in her hand then [03:06] you'll need that [03:06] mdke: *laugh*... No, we're past that stage now. 3.5 years together means we're just boring old people now. === mdke nods [03:07] that's the best way [03:07] heh [03:08] I won't deny that "comfortably boring" is generally nicer than "OMG, excitement, WOW" [03:08] Though some of the latter isn't all bad. [03:08] Also, holy crap is this ever offtopic. [03:08] Shall we go babble in #ubuntu-offtopic? [03:09] Is any development actually going to happen here, though> [03:09] *? [03:09] I wouldn't put it past people to attempt some. [03:09] alright [03:09] I'm happy to take the weekend to "have a life" (and catch up on some Debian stuff), but to each their own. === anselmolsm [n=anselmol@200-232-236-25.dsl.telesp.net.br] has joined #ubuntu-devel === Amaranth_ [n=amaranth@ubuntu/member/amaranth] has joined #ubuntu-devel === storm [n=joel@200.208.133.109] has joined #ubuntu-devel [03:13] infinity: For everything :-) === projecteternity [n=broken@CPE000ae675555e-CM000e5ce667a0.cpe.net.cable.rogers.com] has joined #ubuntu-devel === LaserJock [n=mantha@ubuntu/member/laserjock] has joined #ubuntu-devel === spacey [n=herman@ubuntu/member/spacey] has joined #ubuntu-devel [03:29] wow [03:31] ? [03:34] haha === Bluefox_ [n=bluefox@c-68-33-112-13.hsd1.md.comcast.net] has joined #ubuntu-devel [03:35] Fujitsu: Do you know, off the top of your head, where I could buy canned air? [03:35] Fujitsu: It doesn't seem to be a popular product here... [03:35] No, it doesn't. [03:35] (To blow dust/dirt out of components, etc) [03:36] It's annoying. [03:36] dick smith electronics store? [03:36] And, as you say, the complete expensiveness of anything electronic, and the patheticness of most tools, is really annoying. === _TomB [n=ownthebo@ACCA062F.ipt.aol.com] has joined #ubuntu-devel === Fujitsu runs away from Australia. [03:36] ajmitch: Dick Smith is useless. [03:36] Hmm. [03:36] I gave up on DSE ages ago. [03:36] infinity: usually, but that's about the only place I know of here that would have it [03:36] Jaycar I often use... === zul [n=chuck@ubuntu/member/zul] has joined #ubuntu-devel === rpedro [n=rpedro@87-196-42-14.net.novis.pt] has joined #ubuntu-devel [03:37] hey zul :) [03:37] Morning, zul. [03:37] Hi zul [03:37] hey [03:37] :-) === kermitX_ [n=kermit@unaffiliated/cxg] has joined #ubuntu-devel [03:37] are we that bored? [03:37] Yep === infinity idly wonders why this 4-port SATA controller came with *3* SATA cables... [03:38] Heheh. [03:38] I mean, sure, I'm not much good at counting past 3 either, but c'mon... [03:38] 1, 2, 5 [03:38] (3 sire) [03:38] Yay... Two weeks until Dapper is deployed on our mail/web server at school. [03:39] The Holy Hand-grenade! [03:40] mdke: Brits aren't allowed to quote that... [03:40] explain yourself [03:40] it's all we do [03:40] :O [03:40] mdke: It destroys the illusion that only foreigners like Monty Python. === Fujitsu throws the hand-grenade at mdke. [03:40] chicken of bristol? [03:40] heh [03:41] (Much like, only foreigners like Degrassi Jr High, which is so painfully true) [03:41] heh my wife likes degrassi jr high [03:41] Yep, very true. [03:42] mdke: I would consider the quoting of slighty-less-commonly-exported humour to be acceptable, like Black Books, perhaps. [03:42] Black Books is great1 [03:42] *! [03:42] black adder is funny [03:43] infinity: I'll go and order myself a box set immediately [03:43] mdke: Excellent. [03:43] mdke: I get a comission on each DVD sold. [03:43] heh === mikedep333 [n=mikedep3@71.225.254.63] has joined #ubuntu-devel === mdke goes to buy the complete works of monty python and faulty towers [03:45] heh...dont mention the germans [03:45] or the war either [03:45] *Fawlty Towers === spacey [n=herman@ubuntu/member/spacey] has joined #ubuntu-devel [03:48] Is there nothing more horribly ironic (and sad) than installing Windows 2 days after the Ubuntu release? :/ [03:48] ? [03:49] (reinstalling my girlfriend's Windows/Pornoshop beast) [03:49] D: [03:49] Now with a terrabyte RAID. [03:49] :( === Fujitsu throws the hand-grenade at infinity's girlfriend's house. [03:49] Yessir, that's love. [03:49] NO more Windows problem. [03:49] I used to think that a terabyte was quite a bit [03:50] but now we can buy single drives about that size already [03:50] Like Seagate's 750GB. [03:50] ajmitch: Yeah, I'm installing the 4 300GB drives at the same time as reading reviews for 750GB drives. Suck. [03:50] same old thing [03:50] my 3x250GB RAID seems small now === Fujitsu shall be deploying 3x 300GB SATAII drives in a RAID 5 config at school in a couple of weeks. [03:51] OTOH, Australia won't see thoe 750GB drives at a reasonable price for, like, 5 years, so whatever. [03:51] Big server revamp, including partial Ubuntu conversion. [03:51] infinity, only 5 years!? [03:51] Fujitsu: dapper? [03:51] ajmitch, yes. [03:52] Fujitsu: I just buy hardware and sneak it home everytime I visit North America. [03:52] Hehhe. [03:52] Fujitsu: Sadly, it's the only way to get anything new without spending half my salary. [03:52] Yay for being a student admin on a ~200 workstation network. === Fujitsu throws the school's Exchange CDs away. [03:53] (yes, we're using Exchange at the moment) [03:53] My purchased-in-America Thinkpad still retails here for what I paid for it a year ago. [03:53] Yeah, Australia is silly. [03:53] and NZ is determined to follow [03:53] OTOH, this gives me high hopes for being able to eBay it at a good price when I upgrade. [03:54] Fujitsu: Unless you're in the market for a "guaranteed to work with Ubuntu, cause a core-dev owned it and made sure of that" laptop? :) [03:55] infinity, no. I've got my Dell Inspiron 630m here. Almost works perfectly, only a couple of open bugs. [03:55] infinity, it's good to have a proper Ubuntu dev. in Melbourne :) [03:55] maybe I should flog off my laptop that way.. [03:55] me too [03:56] and anyone else with the same laptop as infinity [03:56] Heh. === Fujitsu goes back to helping in #ubuntu. === infinity suspects that with the Thinkpad ownership in Canonical, you just plain can't go wrong buying IBM/Lenovo, if you want it supported. [03:58] Heheh. Good. [03:58] Though not enough of us have managed to get our hands on T60/X60 hardware to guarantee anything there yet. === mikedep333 [n=mikedep3@71.225.254.63] has left #ubuntu-devel ["Kopete] [03:58] (A T60 is high on my list of "stuff to buy when I get around to it) [03:59] So, how long until Edgy development starts moving properly? [03:59] really moving? after paris, I guess [03:59] It'll slowly start trickling in around Tuesdayish. [03:59] My R50 works perfectly. Can't complain. :) And this was before Ubuntu came out. [03:59] Fujitsu: see https://lists.ubuntu.com/archives/ubuntu-devel-announce/2006-June/000144.html [03:59] That's one list I'm not on... === Fujitsu joins. [03:59] Thanks, crimsun. [04:00] There will be some mass syncs happening next week, but Ubuntu-specific feature devel likely won't happen until post-Paris. [04:00] Aha. [04:00] We should have a large chunk of sid synced into edgy before Paris, though. [04:00] Modulo some outstanding merges here and there. [04:01] so don't bother trying to upgrade then [04:01] ajmitch: MOTU should have a much easier time, with our X coming into sync with Debian's. [04:01] thankfully [04:01] ajmitch: You should be able to drop changes on most of the stuff in universe.. Hopefully. [04:01] Damnit. [04:01] and we'll be able to drop the python2.3/2.4 mess soon [04:01] Mekong has been down for almost 24 hours now. [04:01] ajmitch: If you run into any snags along those lines, let me know, and I'll make sure to push whatever transitional changes we need into sid/etch. [04:01] It's really annoying. [04:02] ajmitch: python2.4 in sid should be happening in the next week or two. [04:02] we agreed to drop zope2.8 from sid last night, iirc [04:02] ajmitch: Of course, doko may decide that he wants to try python2.5 in edgy. :) [04:02] (Probably not, though) [04:02] even if doko wants 2.5, the python-support stuff should avoid having lots of python2.5-* packages [04:03] Hopefully. [04:03] zope2.8 is the last big user of python2.3 specific packages in dapper, we can drop it for edgy [04:03] I had to add a couple of python2.3 packages to universe, taking source from main [04:03] a nasty mess :) [04:04] Why? [04:04] zope2.8 requires python2.3-xml, for example [04:06] zope2.8 should just die anyway. === Fujitsu stabs zope2.8 with a silver stake. [04:07] it will [04:07] soon [04:07] Then again, I think zope in general should just die, but others may disagree. :) [04:07] heh [04:07] haha [04:07] the main use I have of zope is plone [04:07] Which should ALSO die. [04:07] and launchpad :) [04:07] Of course, the main use for zope that I have is launchpad, so whatever. I'm stuck with it. [04:08] lucky [04:08] Heheh. === Fujitsu goes and steps on some bugs [04:09] infinity: so LP should die too? ;-) [04:09] LaserJock: don't ask those questions in a publically logged channel [04:09] Are there enough ubiquity bugs? [04:10] Fujitsu: If you touch ubiquity bugs, Colin *will* hurt you. [04:10] (Unless you're touching them to include patches) [04:11] infinity, aha. I dealt with a few yesterday, asking them to attach the appropriate logs. [04:11] ajmitch: hmm, hadn't thought about that [04:11] And Colin modified all of them shortly afterwards. [04:11] How many Canonical-based Ubuntu devs are there? [04:11] Fujitsu: Oh, asking people to attach logs is fine. Just don't go duping or rejecting or whatnot. [04:12] infinity, I thought so. [04:12] Fujitsu: 99% of d-i/ubiquity bugs that looks like dupes aren't. [04:12] infinity, I know :) [04:12] Fujitsu: (Because "stuff doesn't install right" always looks the same to most people) [04:12] Because they all look practically the same. [04:12] I have the same problem with initramfs-tools. :/ [04:12] It'd be nice if they didn't die with what looked like the same error... [04:12] "Hey, it can't find MY root filesystem either, this is totally the same bug!" [04:12] ARGH. === spacey [n=herman@ubuntu/member/spacey] has joined #ubuntu-devel [04:13] I just wish people would file new bugs instead of commenting in ones that they think are "the same bug" [04:13] Makes it impossible to sort out when one of the 30 bugs in the comment list has been fixed. [04:13] Yes, it's irritating. [04:13] infinity: very irritating [04:13] Although it must be bad for the people that actually have to deal with them. [04:14] my favourite is inlining huge pastes of semi-relevant information. [04:14] YEs. [04:14] I like the ricer kids that need to tell you every piece of hardware in their system. [04:14] Like, multi-hundred line ones in the root. [04:14] Hmm [04:15] "I have an ASUS whizzbang motherboard and an Athlon64 4000+, is my CPU too fast for Ubuntu?" [04:15] Are reports that things don't install icons valid bug reports? [04:15] Heheheheh [04:15] Fujitsu: Sure, icon and desktop file reports are valid wishlists. [04:15] OK. [04:15] I shall mark them as such (there are two new ones). [04:17] No they aren't :-) === stub [n=stub@ppp-58.8.2.92.revip2.asianet.co.th] has joined #ubuntu-devel [04:18] Fujitsu: Okay, to ME they're wishlists. Alternately, you could make them critical and assign them to bddebian. :P [04:19] YES! [04:19] Hmm. [04:19] Why do we have to use Launchpad? [04:19] It's got no `Grave' severity. [04:19] infinity: Well as many as I have done you can if you want :-) [04:19] I think the debbugs severities were deemed unfriendly. === stub [n=stub@ppp-58.8.2.92.revip2.asianet.co.th] has joined #ubuntu-devel [04:19] Unless of course they are for main packages [04:20] bddebian: You're welcome to provide .desktop and icon (and icon transparency, and, and) patches in bug reports for main packages. [04:20] bddebian: I'm pretty sure I've done a few that other people left patches in. [04:20] Why, so they don't get uploaded like my kscd one? :-) [04:20] bddebian: (And done a few without patches when they really irritated me, like the gnome-bluetooth stuff) [04:20] Sorry, that was pissy [04:21] bddebian: You'd have to whine to Riddell about that one. I don't touch KDE, except when it's FTBFS. [04:21] bddebian: you assume that developers have time to look at bugs for every main package [04:21] Aye, I know [04:21] ajmitch: I don't assume anything === _TomB [n=ownthebo@ACCA062F.ipt.aol.com] has joined #ubuntu-devel [04:22] I did ask a couple people to take a look. But when I ask too much I become a pain in the ass :) [04:22] Great. [04:22] Bug #48150 [04:22] Malone bug 48150 in linux-kernel-headers "2.6.15-23 headers are not in the repositories" [Normal,Unconfirmed] http://launchpad.net/bugs/48150 === Fujitsu rejects [04:23] That's special.. [04:23] Of course, he searched for kernel-headers. [04:23] probably following a howto [04:24] Fujitsu: That's actually a dupe of bug #35186 [04:24] Malone bug 35186 in glibc "linux-kernel-headers should be renamed to linux-libc-headers for clarify" [Normal,Confirmed] http://launchpad.net/bugs/35186 === bluefoxicy is still having trouble getting oprofile to profile individual binaries and list library usage in them >< [04:24] Fujitsu: Well, sort of. :) === Burgundavia [n=corey@ubuntu/member/burgundavia] has joined #ubuntu-devel [04:24] Fujitsu: The fact that he filed it on "linux-kernel-headers" confirms 35186's assumption that we should really rename that package. [04:25] Hmm. Yes. [04:25] #35186 isn't /that/ hard to fix, is it? [04:25] Lacks round tuits. === _TomB [n=ownthebo@ACCA062F.ipt.aol.com] has joined #ubuntu-devel [04:26] I'm sure I'll do it for edgy, but I'd like to get the same change in etch/sid, to avoid even more confusion. === ctd [i=ctd@incubus.progsoc.uts.edu.au] has joined #ubuntu-devel [04:27] Heheh. [04:27] next you'll end up as glibc maintainer.. [04:27] ajmitch: Too late. === _TomB [n=ownthebo@ACCA062F.ipt.aol.com] has joined #ubuntu-devel [04:27] heh nice [04:27] my condolences [04:28] ajmitch: I already joined the Debian glibc team, and jbailey and I are co-maintaining it in Ubuntu. [04:28] Then get to work on the hurd bits of glibc will ya? ;-P [04:28] bddebian: Uhm, no. That's not my thing. :) [04:28] didn't realise you'd joined that team [04:28] bddebian: I have my own useless port (m68k) to worry about, thanks. [04:28] heh [04:28] Heheh. [04:29] today would be a good day to fix some of my debian bugs [04:30] Okay, the guy at Sony/Ericsson who decided to put a really bright white LED on my phone was a GENIUS. [04:30] (It's the only flashlight I own, and works great in tight spaces) [04:30] hm [04:30] Hehe. Not bad. [04:31] system wide total it seems 10% of code is executed in actual elf fixed position main executables. === Hobbsee [n=Hobbsee@ubuntu/member/hobbsee] has joined #ubuntu-devel [04:31] 8.2% of that so far is X === RadiantFire [n=ryan@c-69-180-43-27.hsd1.ga.comcast.net] has joined #ubuntu-devel [04:32] does anyone think it's unreasonable to place a 6% performance hit here (on x86)? Total impact appears to be (at a glance) 0.6% === RadiantFire [n=ryan@c-69-180-43-27.hsd1.ga.comcast.net] has left #ubuntu-devel ["Ex-Chat"] [04:32] however I need to figure out htf to get oprofile to work better and do more fine-grained tests >:| [04:33] bluefoxicy: pax? [04:33] bluefoxicy: let me guess you used gentoo at one point? [04:33] ajmitch: mm, actually just doing PIE [04:34] ajmitch: compiling a program as position independent executable imposes a 0.99% performance overhead. Compiling with -fomit-frame-pointer gives a 5% performance increase; however, this is ineffective with PIE on x86. [04:34] ajmitch: assuming we rely on the 5% performance gain from -fomit, that's total 6% slow-down from PIE; but it only affects the main executable [04:34] c [04:34] sorry... [04:36] The result of using a position independent executable is that the main executable .text segment and heap become segments in a dynamic elf binary, identical to a shared object. This allows the base of the main executable AND the heap to take full advantage of existing mmap() randomization in vanilla linux since 2.6.12, as well as higher-entropy PaX randomization if a grsecurity kernel is used (which there is discussion of on [04:36] the wiki) [04:36] in other words you get added security for free === LeeJunFan [n=junfan@adsl-69-210-207-5.dsl.klmzmi.ameritech.net] has joined #ubuntu-devel [04:37] I'm trying to quantify 'free' by determining how much code actually executes in the main executable; so far, Xorg's main executable accounts for 8% of my CPU usage here for the whole system. Using oprofile it seems that total (grepping out any image with .so in it) every main executable including Xorg uses about 10% of the CPU [04:38] Hmm. [04:38] the next most intensive thing was The Battle for Wesnoth (0.54%), oprofiled (0.477%), gtk-gnutella (0.2581%), and metacity (0.0757%) [04:38] however [04:39] I have nfc how to really use oprofile [04:39] If a bug was caused by a dodgy BIOS, and was fixed by an update, is it invalid? [04:39] I just started it up and then ran oreport to see wtf is going on :P [04:40] zul: Yes, I used gentoo with full PaX, PIE, and propolice. Fedora Core 5 includes full FORTIFY_STACK, ProPolice (gcc 4.1 merged), light address space randomization via mainline and exec shield (with heap randomization not in mainline-- this breaks things occasionally), and a few daemons like Apache built as PIE === lemsto [n=salim@ANantes-154-1-62-181.w81-53.abo.wanadoo.fr] has joined #ubuntu-devel [04:41] zul: I believe this may be a good direction for Ubuntu to take, with Edgy or Edgy+1; PIE is a hard argument because Theo de Raadt, Ingo Molnar, and Arjan van de Ven will all tell you that it's extremely expensive and makes the system unbearably slow without providing any use cases or numbers to back this up [04:41] Theo, Ingo, and Arjan are all respected names in the community so everyone takes their statements as gold. === milli [n=milli@phantom.acmeps.com] has joined #ubuntu-devel [04:42] (Theo's actual argument against me when I challenged him on the real overhead of PIE was, "I don't even know who you are. Everyone knows who I am, I created openbsd, we invented this stuff") === Aegir [n=richard@d58-105-2-145.dsl.vic.optusnet.com.au] has joined #ubuntu-devel === spacey [n=herman@ubuntu/member/spacey] has joined #ubuntu-devel === jsgotangco [n=jsg123@ubuntu/member/jsgotangco] has joined #ubuntu-devel === j_ack [n=nico@p508D8A76.dip0.t-ipconnect.de] has joined #ubuntu-devel === _TomB [n=ownthebo@ACCA062F.ipt.aol.com] has joined #ubuntu-devel === bmonty is now known as bmonty_away === spacey [n=herman@ubuntu/member/spacey] has joined #ubuntu-devel === _TomB [n=ownthebo@ACCA062F.ipt.aol.com] has joined #ubuntu-devel [05:24] jdub: ping === _TomB [n=ownthebo@ACCA062F.ipt.aol.com] has joined #ubuntu-devel === didymo [n=ashley@CPE-61-9-197-223.nsw.bigpond.net.au] has joined #ubuntu-devel === shenki [n=shenki@ppp163-51.lns3.adl4.internode.on.net] has joined #ubuntu-devel === spacey [n=herman@ubuntu/member/spacey] has joined #ubuntu-devel === irvin [n=irvin@ubuntu/member/irvin] has joined #ubuntu-devel [05:43] is there a similar mechanism to the replaces: header that allows the overwritten files to remain installed when the 2nd package is removed? [05:43] (so that the overwritten files are only removed by removing the package to which the original file belonged) === _TomB [n=ownthebo@ACCA062F.ipt.aol.com] has joined #ubuntu-devel [05:45] desrt: You mean, you want package B to say "when I'm installed, package A's files should go away, but they should come back when I'm removed"? [05:46] no. that's impossible === _TomB [n=ownthebo@ACCA062F.ipt.aol.com] has joined #ubuntu-devel [05:46] No, that's not, that's diversions. [05:46] i mean i want a package to effectively upgrade a single file of another package [05:46] actually [05:46] that's quite fine, i guess [05:46] how do i do this? [05:46] man dpkg-divert [05:46] thanks [05:47] If you're just looking to upgrade package A, though, do that. ;) [05:47] i want to replace a single file [05:47] dpkg-divert is only sane if you really need package B to ship a file that shouldn't be in package A, but is needed by package B. [05:48] what happens if A gets upgraded in the meantime? [05:48] (See, for install, nvidia-glx and fglrx using dpkg-divert to use their own libGL.so.1.2) === _TomB [n=ownthebo@ACCA062F.ipt.aol.com] has joined #ubuntu-devel [05:48] divert puts the new version of the stock file in its little storage cabinet for safe-keeping? [05:48] That's the point of a diversion. If you divert libGL.so.1.2 to /usr/lib/fglrx/libGL.so.1.2, the next time libgl1-mesa is upgraded, it's copy goes to the bogus location. [05:49] nice. [05:49] When the diversion is removed, the right file is moved back into place. [05:49] It's a sketchy process, and best avoided in most cases, but useful when unavoidable. [05:51] i like to learn. === jcape [i=jcape@71.194.177.222] has joined #ubuntu-devel === Amaranth [n=amaranth@ubuntu/member/amaranth] has joined #ubuntu-devel === _TomB [n=ownthebo@ACCA062F.ipt.aol.com] has joined #ubuntu-devel === spacey [n=herman@ubuntu/member/spacey] has joined #ubuntu-devel === Chipzz [n=chipzz@ace.ulyssis.student.kuleuven.be] has joined #ubuntu-devel === LeeJunFan [n=junfan@adsl-69-210-207-5.dsl.klmzmi.ameritech.net] has joined #ubuntu-devel === stub [n=stub@ppp-58.8.2.92.revip2.asianet.co.th] has joined #ubuntu-devel [06:21] infinity; woo. go dpkg-divert. === freeflying|away [n=freeflyi@ubuntu/member/freeflying] has joined #ubuntu-devel === bluefoxicy [n=bluefox@c-68-33-112-13.hsd1.md.comcast.net] has joined #ubuntu-devel [06:34] here's a question [06:34] 1.8% execution performance overhead [06:34] For every 60 minutes of continuous maxed 100% CPU usage without a break, 1 additional minute of execution time is needed to complete the task. [06:34] Honestly, do we care? [06:38] bluefoxicy: It depends. It's a bit of a loaded question you're asking. [06:39] Foreground or background task? Interactive or batch? Latency or bandwidth constrained? I/O or CPU driven? [06:40] sfllaw: system wide, compiler modification that allows a security guarantee to be made on the probability that any given instance of an attack can actually result in exploitation of a known security hole. [06:40] sfllaw: CPU driven. [06:40] You mean you're adding in code to compiled applications? [06:40] no [06:40] Or is this static analysis? [06:40] ok. next task [06:40] learn how to install an initscript [06:41] sfllaw: open a shell and do: /lib/libc.so.6 [06:41] sfllaw: (just trust me on this one, it'll work) [06:41] I know it does. [06:41] ok [06:41] You know why that works right? [06:41] libc.so.6 has a _main(), and is thus an executable program, but is position independent and all because it's a library [06:42] Yes. [06:42] well on IA-32, doing this to a program causes a 0.99% decrease in execution speed of the code in the main executable of the program; also, using -fomit-frame-pointer normally gives a 5% increase in execution speed, but this is completely ineffective with position independent code. [06:43] 6% total decrease in execution speed of the code [06:43] crimsun: you around? [06:43] However, system-wide, according to oprofile only 10% of the code executed on my system in a half hour was in any main executable; 8% of that was in Xorg (spends a lot of time in the main executable huh?) and the rest were all around 1/2 percent except battle for wesnoth [06:44] Overall, that's a 0.6% decrease in execution speed (10% of the time you're in code that's 6% slower?) [06:44] sfllaw: However, I am guessing that the time spent in the main executable ranges from 1% to 30% and saying this is 0.06% to 1.8% performance hit [06:45] (percent of time spent in affected code times slowdown on execution of said code) [06:46] sfllaw: Randomization of the stack and mmap() base is nice, but a really crafty attacker can inject his stack frames in the heap (at a known base address) and return to a 'call system()' instruction with SFP putting the pointer somewhere appropriate in the heap [06:46] this would give a pretty damn workable method of completely nullifying the protection afforded by mmap()/stack randomization with 100% efficiency [06:46] Are you proposing that we take out -fPIC for IA-32? [06:46] uh hell no. [06:47] Opposite way [06:47] (and also you can't take -fPIC out for IA-32, libraries break massively) [06:48] By making the main executable a PIE, it suffers the noted performance slow-down; however, it also makes the described attack impossible, as the main .text segment is randomized via mmap() and the heap is an anonymous mapping randomized via mmap() (I *THINK* you need extra kernel code to actually randomize the heap though... I'm not sure it gets stuffed into the GOT) [06:49] thus you can make a security guarantee that security attacks based on buffer overflows, double free()s, etc are likely to succeed 1 in every 2^27 times (134 million fail states, 1 success state) [06:49] in theory [06:50] I'm proposing that the main executables be built -fPIE (-fPIC for main executable) on all architectures, I just want to get a heads up on what the reaction to the predicted performance impact will be like :) [06:52]