| infinity | ogra-cb_: http://paste.ubuntu.com/1380830/ | 00:15 |
|---|---|---|
| infinity | ogra-cb_: Not sure that'll be acceptable. | 00:15 |
| ScottK | infinity: 17 chrootwaits (at least) on heka, so I put it on manual. | 04:30 |
| ogra_ | infinity, i think for 170M i can live with 24min added to the build, lets do it (and probably get a second builder for the flavours), its still way below the timing of a build on teh babbage board, so we had worse | 10:51 |
| infinity | ogra_: And by 24m, you mean 40m? | 11:30 |
| infinity | ogra_: At least, 48-8 is 40 in my world. | 11:30 |
| infinity | ogra_: And probably longer on production hardware, I was testing on a PandaES, not a Panda. | 11:32 |
| infinity | ogra_: So, it'll likely add at least an hour to the build. | 11:32 |
| ogra-cb_ | bah, i kooked at -9 --rsyncable | 11:44 |
| ogra-cb_ | *looked | 11:44 |
| ogra-cb_ | bah plymouth 0.8.8 gets the nexus back into constant reboot loops :( | 11:51 |
| ogra-cb_ | damned | 11:52 |
| infinity | Is that the "trying to attach to a device called 'none'" bug, or something new? | 11:52 |
| ogra-cb_ | not sure yet, i know in 0.8.5 were some fixes for serial console handling, not sure they fixed empty/bogus console= args as well | 11:53 |
| infinity | (I'm not even sure why that was a bug... How hard is it to state the device node before trying to open it?) | 11:53 |
| infinity | s/state/stat/ | 11:53 |
| ogra-cb_ | yeah | 11:53 |
| infinity | Or do something sane when open() returns ENOENT, whatever. | 11:54 |
| ogra-cb_ | well, i havent looked at the 0.8.8 soource yet, the 0.8.4 one was horrid though | 11:55 |
| * ogra-cb_ puts the plymouth override files back in | 12:06 | |
| ogra-cb_ | *sniff* | 12:07 |
| ogra-cb_ | infinity, what about lzma btw ? any chance that runs faster ? | 12:07 |
| infinity | ogra-cb_: Probably slower, but I'll test for fun. | 12:08 |
| infinity | ogra-cb_: Lower xz opt levels may be more acceptable too. | 12:09 |
| * infinity tries xz -4 for kicks. | 12:10 | |
| ogra-cb_ | well, gaining 100M would eb good | 12:10 |
| ogra-cb_ | i mean i have to split the desktop seed anyway into a core part and applications ... there was some WI somewhere | 12:10 |
| * infinity finds --memlimit-compress=limit in the xz manpage, and decides to test that next. | 12:20 | |
| infinity | Although, this doesn't seem to be swapping right now. Just chewing CPU. | 12:21 |
| infinity | So maybe memlimit won't help. | 12:21 |
| infinity | ogra-cb_: I'm going to nap. I'll throw some more random results at you later, and we'll see if something looks like a reasonable size/speed tradeoff. | 12:22 |
| infinity | ogra-cb_: Or, we can re-examine the "build the image on nusakan" idea. :/ | 12:22 |
| infinity | ogra-cb_: Unless make_ext4fs is under pretty seriously heavy development, a single backport for cdimage would probably do us fine, right? | 12:23 |
| ogra-cb_ | yeah | 12:23 |
| infinity | ogra-cb_: (Or, we could just be naughty and keep a privately-built copy in ~cdimage/bin) | 12:24 |
| ogra-cb_ | hehe | 12:24 |
| ogra-cb_ | i bet you could just cp the existing quantal binary in place | 12:24 |
| infinity | Would speed up ac100, nexus7, and core to do the post-tarball zipping and mangling on nusakan. | 12:25 |
| ogra-cb_ | indeed | 12:25 |
| infinity | The Q binary depends on some glibc 2.14 symbols, apparently. | 12:25 |
| ogra-cb_ | but it does involve IS in the end if you want to do it properly at some point | 12:25 |
| ogra-cb_ | ah, crap | 12:25 |
| ogra-cb_ | and nusakan is even lucid i think | 12:26 |
| infinity | Well, some binary in the package does, I didn't look at make_ext4fs itself, just the dpkg deps. | 12:26 |
| ogra-cb_ | iirc xnox did a precise backport too | 12:26 |
| infinity | Well, the precise package would have the same issue. :P | 12:26 |
| infinity | Since precise is 2.15 | 12:26 |
| ogra-cb_ | right | 12:26 |
| ogra-cb_ | just meaning nothing goes back far enough for lucid | 12:27 |
| infinity | (base)adconrad@cthulhu:~$ readelf -a /usr/bin/make_ext4fs | grep 'Name: GLIBC' | 12:28 |
| infinity | 0x0030: Name: GLIBC_2.11 Flags: none Version: 8 | 12:28 |
| infinity | 0x0040: Name: GLIBC_2.14 Flags: none Version: 7 | 12:28 |
| infinity | 0x0050: Name: GLIBC_2.8 Flags: none Version: 6 | 12:28 |
| infinity | Yeah, we'd need to build it for lucid to avoid that. | 12:28 |
| infinity | 0x0060: Name: GLIBC_2.4 Flags: none Version: 5 | 12:28 |
| infinity | 0x0070: Name: GLIBC_2.3.4 Flags: none Version: 3 | 12:28 |
| infinity | 0x0080: Name: GLIBC_2.2.5 Flags: none Version: 2 | 12:28 |
| infinity | Oh, right, shiny new memcpy in 2.14 is versioned. | 12:29 |
| infinity | Anyhow. Nap time. More pondering tomorrow. | 12:32 |
| infinity | Or, more accurately, later today. | 12:32 |
| ogra-cb_ | yeah, sleep well | 12:32 |
| infinity | Oh, that test just finished (-4) | 12:44 |
| infinity | adconrad@shiva:/srv/test$ time xz -4 -c rootfs.tar > rootfs.tar.xz.4 | 12:44 |
| infinity | real33m23.931s | 12:44 |
| infinity | user32m58.602s | 12:44 |
| infinity | sys0m12.578s | 12:44 |
| infinity | -rw-rw-r-- 1 adconrad adconrad 595792596 Nov 23 15:08 rootfs.tar.gz.9.rsync | 12:44 |
| infinity | -rw-rw-r-- 1 adconrad adconrad 424087940 Nov 23 15:57 rootfs.tar.xz | 12:44 |
| infinity | -rw-rw-r-- 1 adconrad adconrad 444655528 Nov 24 05:43 rootfs.tar.xz.4 | 12:44 |
| infinity | That's starting to maybe approach acceptable. Maybe. I dunno. | 12:44 |
| infinity | Doing it on cdimage is likely saner. We'll talk later. | 12:45 |
| cjwatson | stgraber: I missed a debian-cd patch, which probably accounts for your problems. Applied now, so the next daily build should be happier. | 15:46 |
| === yofel_ is now known as yofel | ||
| xnox | ogra ogra_ ogra-cb_ : ppa:ubuntu-nexus7/ubuntu-nexus7-installer now has android-tools for lucid, precise and quantal. | 21:20 |
| * cjwatson fixes the bogus overrides that resulted in ubuntu-zh_CN/raring/{amd64,i386} image build failures | 23:35 | |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!