=== cbrake is now known as cbrake_away | ||
armin76 | uh, emerge is not horrendous | 06:36 |
---|---|---|
armin76 | its better than apt :) | 06:36 |
Martyn | Ugh | 06:37 |
Martyn | Lets not get into the whole package management religious war again | 06:37 |
Martyn | that -practically- counts as a troll. grin | 06:37 |
Martyn | Oh, and I got some of the installer scripts to work finally | 06:38 |
Martyn | so I've got the beginnings of an actual Karmic installer that doesn't involve a buildroot | 06:38 |
kblin | you mean other than rootstock? | 06:44 |
Martyn | yep | 06:51 |
Martyn | Getting there slowly. It's not the actual server installer yet, but I'm grinding away at the problem | 06:51 |
Martyn | my eventual goal is to get an ARM server build that net installs successfully on a couple platforms | 06:52 |
Martyn | PBX-A9, FastModels RealView EB, and the tech my company is working on (Smooth-Stone) | 06:52 |
kblin | none of them ring a bell for me.. anything with a couple of ethernet ports on them? :) | 06:53 |
Martyn | All of them | 06:58 |
Martyn | Well, the RealView EB is not a physical machine .. it runs as a fastmodel software simulation of the entire A9 Core (quad core) at ~150Mhz | 06:59 |
kblin | I'm definetely playing with the wrong hadware right now :) | 06:59 |
Martyn | The PBX-A9 is the Cortex A9 version of ARM's hady PBX platform .. they pretty much recycled it | 06:59 |
Martyn | although the new A9 version has a neat implementation of a dual core Cortex A9 running also at about 150Mhz. Pretty impressive, considering they basically dropped the soft core onto an ASIC | 07:00 |
kblin | if you say so... | 07:00 |
Martyn | (rather than test silicon) | 07:00 |
Martyn | the tech my company is working on is, of course, a real Cortex A9 ... but I can't really share that much of what we're up to. | 07:00 |
Martyn | But I can't wait to play with the real chips :) | 07:00 |
Martyn | kblin : For a couple of ethernet ports, you're probably going to have to wait until TI starts producing non-sample quantities of the OMAP4 | 07:01 |
Martyn | and even then, it's pretty dependent on someone -else- (beagle?) interfacing a chip that doesn't use PCI (like the SMSC 9118) | 07:02 |
Martyn | Since that can be mapped directly onto SRAM | 07:02 |
kblin | so what I presented on the storage developer's conference yesterday was a Samba4 Active Directory domain controller running on a beagle, as that's the hardware I got | 07:02 |
Martyn | Urk | 07:02 |
Martyn | Cortex A8 though, so at least it has some optimizations | 07:02 |
kblin | and as a proof of concept, it's been pretty impressive | 07:02 |
Martyn | What did you use for an ethernet adapter? Phoenix chip off the USB port? | 07:02 |
kblin | moschip, but yeah | 07:03 |
Martyn | I did something similar with a shivaplug | 07:03 |
kblin | yeah, I got one of those yesterday | 07:03 |
kblin | not in time for my presentation, though | 07:03 |
Martyn | multicore is win when doing asynchronous IO | 07:03 |
Martyn | but the arm v5 instruction set (since it does v5 multicore) is limiting | 07:04 |
Martyn | kblin : You'll be impressed with the performance that little thing can pump, when it comes to IO | 07:04 |
kblin | what's killing the beagle is the crappy USB | 07:04 |
Martyn | kblin : The Kirkwood core is interesting, if a bit "yesterday" | 07:04 |
Martyn | yep. | 07:04 |
Martyn | you're going to get equally crappy performance out of the shivaplug, I'm afraid | 07:04 |
kblin | it's going to be more yesterday when ubuntu rops v5 support | 07:04 |
Martyn | I ran a Samba 3 PDC on one, and it worked .. but it wasn't anything stunning | 07:05 |
Martyn | what's REALLY going to kill you is the lack of RAM | 07:05 |
Martyn | no danger of dropping arm v5 for a while. | 07:05 |
kblin | it seems to be a bit faster for AD stuff | 07:05 |
Martyn | Hell, the whole build system is geared to MAKING v5 eabi binaries at the moment, built on QEMU of all things | 07:05 |
kblin | I didn't benchmark file server performance | 07:06 |
Martyn | I'm starting to figure out how to configure launchpad to perform v7 builds now | 07:06 |
Martyn | since that's the key to getting some application performance on the new systems, and I really need to start focusing on that | 07:06 |
Martyn | It's slow going | 07:06 |
kblin | however, Samba4 is just starting to support AD replication, so setting up a small embedded box in a branch office that replicates the main office's AD setup seems like a nice product | 07:07 |
kblin | so fileserver performance isn't too critical for that kind of setup | 07:07 |
Martyn | true | 07:09 |
=== Q_Continuum_ is now known as Q_Continuum | ||
lool | Missed Martyn | 08:18 |
lool | Would have liked to know if it's a d-i bsaed installer | 08:18 |
ogra | lool, wrt usplash, i know that we need USPLASH=y in initramfs.conf (or the .d dir) but i still see no sane way to put it there apart from hacking up the installer and casper probably | 10:21 |
ogra | or use an expensive function that does an arch check in initramfs | 10:23 |
lool | ogra: Just add a file in usplash on build | 10:28 |
ogra | hmm | 10:28 |
siji | hey | 12:18 |
siji | anybody can help me to compile clutter with openGL support | 12:19 |
siji | still am with the same error | 12:20 |
siji | checking EGL/egl.h usability... no | 12:20 |
siji | checking EGL/egl.h presence... no | 12:20 |
siji | checking for EGL/egl.h... no | 12:20 |
siji | configure: error: Unable to locate required GLES headers | 12:20 |
siji | ogra, i got the tutorial abt this for angstrom OS | 12:34 |
siji | anyway to make it for ubuntu | 12:35 |
siji | ? | 12:35 |
ogra | no idea | 12:35 |
siji | ogra,ok | 12:36 |
ogra | did you ask in #beagle (or if there is an angstrom channel, did you ask there) | 12:36 |
siji | ok | 12:40 |
siji | ogra,those header files are in Omap SDK | 12:46 |
siji | and i copied it to unbuntu rootfs | 12:46 |
siji | but no hope | 12:46 |
siji | mike^,u there | 12:47 |
siji | ? | 12:47 |
mike^ | yes | 12:54 |
mike^ | mostly | 12:54 |
siji | mike^, any idea ? | 12:57 |
mike^ | siji, not really, what I did was 'cp -r /path/to/SDK/OGLES/SDKPackage/Builds/OGLES/Include/GLES /path/to/ubuntu/root/usr/include' | 13:02 |
mike^ | 'cp -r /path/to/SDK/OGLES/SDKPackage/Builds/OGLES2/Include/EGL /path/to/ubuntu/root/usr/include' | 13:02 |
mike^ | and than on omap machine: apt-get -s build-dep clutter | 13:02 |
siji | mike^, i tried | 13:03 |
mike^ | apt-get -b source clutter | 13:03 |
siji | mike^ actually i compiled clutter frm Source | 13:03 |
siji | ./configure --with-flavour=eglnative --with-gles=2.0 --with-imagebackend=gdk-pixbuf --with-x=no --with this prefix | 13:03 |
mike^ | siji, I've used the configure options from the .deb itself, I've only changed the --with-flavour to 'eglx' | 13:05 |
siji | mike^, like " apt-get -s build-dep clutter eglx " | 13:07 |
siji | ? | 13:07 |
mike^ | siji, just "apt-get -s build-dep clutter" | 13:10 |
siji | ok | 13:10 |
siji | let me try again | 13:10 |
mike^ | siji, and during the 'apt-get -b clutter' I've interrupted it, and changed debian/rules to use eglx | 13:10 |
siji | ok | 13:11 |
=== cbrake_away is now known as cbrake | ||
lool | rabeeh: http://cdimage.ubuntu.com/ports/daily-live/current/ | 15:07 |
lool | rabeeh: note that it's not always there (might fail to build) | 15:08 |
lool | I didnt test this image either so it might not even boot | 15:08 |
ogra | the imx51 one boots fine | 15:08 |
lool | odd my rsync just failed with rsync: link_stat "/ports/daily-live/current/karmic-desktop-armel+dove.img" (in cdimage) failed: No such file or directory (2) | 15:08 |
ogra | so i'd expect the dove one to do that too | 15:08 |
ogra | the link was just moved :) | 15:09 |
lool | ogra: You kicked a build? | 15:09 |
* ogra knows that prob and stopped using current/ long ago | 15:09 | |
ogra | no | 15:09 |
ogra | i'm still installing | 15:09 |
lool | I never had an issue with that | 15:09 |
lool | and the web index has /current working fine | 15:09 |
ogra | well, i know it breaks if the link moves while you are actually in the sync process | 15:10 |
lool | I just started my rsync | 15:10 |
rabeeh | which u-boot configuration are you using? | 15:10 |
rabeeh | (bootcmd etc...) | 15:11 |
lool | rabeeh: That's the good question to ask | 15:11 |
lool | the image is meant to be started with the new bootcmd discussed in $bug | 15:11 |
ogra | there should be a wikipage | 15:11 |
lool | yeah | 15:11 |
lool | https://wiki.ubuntu.com/ARM/MarvellDoveKarmicInstall | 15:11 |
lool | rabeeh: ^ that's semi current | 15:12 |
lool | rabeeh: Ideally you'd just the loop and it would autostart the USB key install | 15:12 |
ogra | https://wiki.ubuntu.com/ARM/MarvellDoveKarmicInstall | 15:12 |
ogra | bah | 15:12 |
ogra | lool beats me | 15:12 |
lool | ogra: Young padawan | 15:12 |
* ogra buys http://www.frankelcostume.com/Beard-18-Grey-Oldtimer-P16867C93.aspx :P | 15:12 | |
lool | rabeeh: I pushed support for dove-z0 netboot images if anybody cares; will be published with next debian-installer upload; untested though | 15:14 |
lool | The dove netboots were incomplete (were relying on uimage kernels which we dont have anymore) so these will be fixed at this time as well | 15:14 |
lool | rabeeh: http://ports.ubuntu.com/ubuntu-ports/dists/karmic/main/installer-armel/current/images/ is the base dir for the netboot images | 15:15 |
rabeeh | which one will have UNR? | 15:15 |
lool | ogra: Oh you were the male model below the beard? | 15:16 |
lool | rabeeh: We might add it if we get 3d working | 15:16 |
lool | I mean opengl | 15:16 |
lool | rabeeh: But for now we cancelled that support in karmic | 15:16 |
lool | rabeeh: it's not very hard to add, perhaps some hours + testing but we need a way to test first | 15:17 |
rabeeh | right | 15:17 |
rabeeh | what's the best way to do it - | 15:17 |
rabeeh | 1. clutter + opengl-es | 15:17 |
rabeeh | 2. clutter + opengl + translation to opengl-es + opengl-es | 15:18 |
rabeeh | ? | 15:18 |
rabeeh | lool: translation library like the one you previously mentioned to me? | 15:18 |
ogra | lool, heh, we still have usplash on shutdown on the live image :) | 15:24 |
lool | rabeeh: In theory clutter can work on either IIRC | 15:26 |
lool | rabeeh: But there might be some issues with clutter 1.0 | 15:26 |
lool | rabeeh: The point of clutter is to abstract that away | 15:26 |
ogra | sniff, no gdm for me after install :/ | 15:26 |
rabeeh | if the performance of 2 is similar to 1; personally i would go with 2, since any other application that uses opengl (which are the majority for netbooks) would use the hardware engine | 15:30 |
ogra | hmm, it starts if i satrt dbus manually | 15:31 |
ogra | sadly i forgot to start hal too :P | 15:31 |
ogra | no xinput | 15:31 |
lool | rabeeh: Sure that's certainly doable; the question is how much time to fix clutter versus how much time to get that bridge working | 15:33 |
lool | I didnt look into the latter | 15:33 |
ogra | sigh, so on second boot dbus comes up fine | 15:33 |
* ogra hates race conditions | 15:33 | |
ogra | but but | 15:33 |
ogra | but | 15:33 |
ogra | wow, the babbage boots *faaaaast* ! | 15:33 |
lool | ogra: with the new stuff? | 15:34 |
ogra | yeah | 15:34 |
lool | That's really good news | 15:34 |
ogra | incredible | 15:34 |
lool | That was the biggest issue | 15:34 |
ogra | but see above | 15:34 |
lool | Yeah | 15:34 |
ogra | dbus crached on first boot after install | 15:34 |
ogra | *crashed | 15:34 |
ogra | second one is really quick and everything is up | 15:35 |
* ogra digs for his stopwatch | 15:35 | |
lool | ogra: is this from SD? | 15:37 |
ogra | 15sec from first screen output to X, pls ten more to see the panels | 15:37 |
lool | or after install? | 15:37 |
ogra | nope, USB key | 15:37 |
lool | wow | 15:37 |
ogra | after install | 15:37 |
ogra | adding the applets takes a bit | 15:38 |
lool | it's comparable to my x301 with jaunty | 15:38 |
ogra | but its a massively different experience | 15:38 |
lool | I think I'll like the boot stuff when it lands | 15:38 |
ogra | i do already ! | 15:38 |
lool | I'd love to try that on dove | 15:39 |
ogra | spinning disks are said to be much slower | 15:39 |
lool | too bad we have no live image for z0 | 15:39 |
* ogra installs bootchart on the babbage | 15:40 | |
ogra | bah, wrong keymap | 15:41 |
=== playya__ is now known as playya | ||
rabeeh | lool: i'm trying to use your netboot initrd | 16:30 |
rabeeh | lool: i get the following (with my kernel) - | 16:30 |
rabeeh | RAMDISK: gzip image found at block 0 | 16:30 |
rabeeh | RAMDISK: incomplete write (30549 != 32768) | 16:30 |
rabeeh | write error | 16:30 |
rabeeh | which addresses do you tftp load uImage and the initrd? | 16:30 |
bjf | rabeeh, how many "users" of Y* boards exist outside of Marvell or Canonical? | 16:34 |
rabeeh | bjf: i don't know; i don't have the numbers | 16:35 |
bjf | rabeeh, I'm not really looking for specifics, I'm thinking about number impacted if a bad kernel went out for some reason | 16:37 |
bjf | rabeeh, we are thinking about how flexible we should be in taking kernel changes | 16:38 |
bjf | rabeeh, if the number of users that would be impacted is small we can be more flexible | 16:39 |
rabeeh | bjf: hmmm.... i think you should be very flexible and never send a bad kernel for any reason :) | 16:39 |
bjf | rabeeh, lol | 16:39 |
rabeeh | bjf: seriously i really don't have the numbers; but i can say that the demand is huge. | 16:39 |
rabeeh | bjf: i think it's ok for the next few weeks to have bad kernels (for some reason) until people get used to karmic images | 16:40 |
rabeeh | bjf: but as i talked to saeed today; he told me that there are minor changes left for the git tree to be 100% in sync with marvell's | 16:40 |
rabeeh | bjf: so the risk is really very low on having bad kernels due to dove support. | 16:41 |
bjf | rabeeh, I really wouldn't expect a bad kernel to go out, however if we allow lots of changes to the karmic kernel after it has been released, then there is a greater possibility | 16:41 |
bjf | rabeeh, the marvell changes are carried on a topic branch so only that would be impacted | 16:42 |
lool | rabeeh: Didn't try the netboot images yet | 16:47 |
lool | rabeeh: I'd use the same addresses as the boot script | 16:47 |
lool | rabeeh: 0x00200000 for kernel and 0x01100000 for ramdisk | 16:48 |
lool | rabeeh: note that kernel is in zimage format instead of uimage; will be fixed with next debian-installer upload | 16:48 |
armin76 | rabeeh: did you receive my mail? | 17:14 |
ali1234 | m when compiling | 17:53 |
=== bjf is now known as bjf-afk |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!