| MTeck-Linux | what code is it that requires initramfs support now in 9.10 that wasn't there in 9.04? | 01:05 |
|---|---|---|
| RAOF | MTeck-Linux: You mean 10.04 rather than 9.10? IIRC there was a “parallelise populating of root filesystem" patch for bootspeed that broke non-initramfs. | 01:07 |
| MTeck-Linux | ya.. | 01:07 |
| MTeck-Linux | oh... | 01:07 |
| MTeck-Linux | RAOF: You have any idea if there's intention to have that fixed by release time? | 01:09 |
| RAOF | MTeck-Linux: It'd want to be fixed before sending upstream, but I don't belive that the kernel team cares about the non-initramfs case for Ubuntu. | 01:12 |
| RAOF | MTeck-Linux: I think there was discussion about this on the kernel-team@ mailing list, for what it's worth. | 01:12 |
| MTeck-Linux | RAOF: any idea how long ago, I could make an attempt to try to fix it - although it's not likely I have that skill | 01:13 |
| RAOF | MTeck-Linux: It'd be about a month ago, I think. | 01:17 |
| RAOF | I can't seem to find the exact mail. | 01:17 |
| MTeck-Linux | RAOF: thanks, I can look for it later too; once I get that other thing that's missing on my config I can see if I can figure out how to fix it :) | 01:19 |
| === jMyles_ is now known as jMyles | ||
| === Whoopie_ is now known as Whoopie | ||
| * lamont tries to remember - in general could one run a dapper userspace on a hardy kernel? | 12:02 | |
| smb | lamont, With luck but with no guaranties | 12:40 |
| lamont | right | 12:41 |
| lamont | our test of PPC_CELL=n didn' | 12:41 |
| lamont | t pan out, so it's back to dapper | 12:41 |
| lamont | but it'd be nice test a hardy kernel with tweaks from time to time to see if we can get to hardy | 12:42 |
| smb | The problem likely is with userspce, as the kernel between lts releases is likely prone to some kernel-userspace abi changes | 12:43 |
| smb | And actually I got the feeling we probably do not care enough to make much effort there as Dapper goes away soon if I did not remember the wrong year | 12:44 |
| smb | lamont, Ok, sorry I remembered the wrong year. Maybe more hoping than remembering. :-P | 12:50 |
| _ruben | my current build box has 2 cpus with ht (so 4 logical cpus), when compiling a kernel, what would be the "optimal" concurrency setting? | 13:01 |
| rtg | _ruben, by default the kernel build determines the number of CPU and does a 'make -jX' accordingly | 13:02 |
| _ruben | rtg: ah ok, didnt know that | 13:02 |
| rtg | _ruben, you can override it by setting CONCURRENCY_LEVEL as an environment variable | 13:03 |
| _ruben | but i suppose the default should be "best" in general... | 13:05 |
| _ruben | i had my doubts between N and N+1 .. based on very vague memories from long long ago | 13:05 |
| smb | _ruben, You might always try out what is best in your case, but generally the build is io bound in most cases | 13:06 |
| rtg | _ruben, it hardly makes much difference. you're usually I/O bound anyway | 13:06 |
| rtg | smb, stop reading my mind | 13:06 |
| * smb tries | 13:06 | |
| smb | For i7 systems there have been suggesd | 13:07 |
| smb | stions to limit the number of cores and so allow the rest to go to boos mode | 13:07 |
| rtg | smb, what is boos mode? | 13:08 |
| smb | (fingers still a bit numb after doing a walk outside) | 13:08 |
| smb | On those if you do not use all cores the remaining go faster (the die temp is the same in that case) | 13:08 |
| rtg | smb, not many i7's in the real world yet, are there? | 13:09 |
| smb | rtg, one in my room :) | 13:09 |
| rtg | smb, and 2 in my shop | 13:09 |
| smb | So at least 3. And there was some coverage on those in one magazine I read. But yeah, probably a bit costly for normal usage. | 13:12 |
| _ruben | default build gives me 67% user, 8% sys, 25% idle .. disk util around 5% .. according to iostat | 13:48 |
| smb | I believe to remember the iowait statistics were the interesting ones. But must admit I have not really looked closely lately. | 13:52 |
| _ruben | smb: was looking at those just now .. between 0.00% and 0.30% .. software raid1 over 2 sata disks | 13:55 |
| smb | Sounds a bit like there would be room for trying a bigger -j value to see whether this would get your cpus more saturated. I think in the end for us it was an ok approach to go just for the numbers of cpu's. You still can do more than one build in parallel... | 13:58 |
| _ruben | heh .. increase concurrency to 6 and iowait sky rockets :) | 14:02 |
| smb | hehe, ok, so the default is not too bad. :) | 14:04 |
| _ruben | apparently so :) | 14:05 |
| hashimi | Hi everyone | 14:08 |
| hashimi | I have a problem | 14:09 |
| hashimi | I want to change my boot splash image from my own setup linux. | 14:09 |
| hashimi | But i don't know how to do it. | 14:09 |
| hashimi | could anyone help me? | 14:09 |
| smb | hashimi, Hopefully they did not send you here from there, but I think in #ubuntu-devel there might be more chance to find one with that knowledge | 14:11 |
| hashimi | Ok really thanks for reply. I will go out to there. | 14:11 |
| hashimi | Thanks smb | 14:11 |
| _ruben | btw .. currently i build kernel by calling debian/rules directly, is there a recommended/supported/whatever method that'd yield proper .dsc files and the like, so i can dput it to my own repo | 14:12 |
| smb | _ruben, debuild -B should yield the same results as the buildd's | 14:13 |
| smb | possibly needing -uc -us for not signing the files by default or overriding the signers key | 14:14 |
| _ruben | smb: that'd probably build all flavours right? as im only interested in my own custom flavour | 14:14 |
| smb | Right in that case you would either need to modify things to remove flavours or do as you do now. But that created the deb files only | 14:16 |
| smb | For the source package you can call "debuild -S -i -I" after a "fakeroot debian/rules clean" | 14:16 |
| smb | err | 14:17 |
| smb | "dpkg-buildpackage -S -i -I" or "debuild -S" I think, though I usually use the former | 14:18 |
| _ruben | smb: ah ok, thanks for the tips | 14:25 |
| _ruben | crap .. used the wrong buildroot .. used a karmic one, while i need a kernel for hardy :p | 14:44 |
| _ruben | gives me a libc6 dependency error on install :) | 14:45 |
| smb | Doh, and oh btw. I think in Hardy the default -j might not (yet) be related to the number of CPU's- | 14:50 |
| _ruben | good call | 14:52 |
| * _ruben hits ^C | 14:52 | |
| _ruben | noticed a -j1 :) | 14:52 |
| _ruben | lets try concurrency level of 5 | 14:53 |
| === BenC2 is now known as BenC | ||
| === yofel_ is now known as yofel | ||
| _ruben | weird .. my i386 backport of karmic's kernel (with custom patch) does work as expected .. but the amd64 backport fails to detect my lvm at boot :/ | 16:55 |
| _ruben | oh well, that's something for next week to figure out | 16:55 |
| rackerhacker | jjohansen: were you able to remember where that extra kernel memory came from in linux-ec2 on amd64? | 17:09 |
| jjohansen | rackerhacker: no, I need to spend some time looking at the code and I just haven't managed to get to it yet | 17:12 |
| rackerhacker | jjohansen: that's okay, i figured i'd just check in ;) | 17:13 |
| === asac_ is now known as asac | ||
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!