scientes | [890160.173487] VFS: Busy inodes after unmount of mtd_inodefs. Self-destruct in 5 seconds. Have a nice day... | 05:59 |
---|---|---|
scientes | a little bit of spaceballs eh? | 05:59 |
scientes | in my kernel you say? | 05:59 |
=== gry is now known as gry2 | ||
=== gry2 is now known as gry | ||
sveinse | I'm experiencing with precise debootstrap on precise. However it fails very often with "Couldn't download package..." and it happens for different packages for each invocation of debootstrap. Other with similar experience? | 14:33 |
sveinse | *experimenting | 14:33 |
=== NekoXP is now known as Neko | ||
=== robclark_ is now known as roblark | ||
=== roblark is now known as robclark | ||
sveinse | Where can I find the default compiler settings for armel and armhf? | 16:54 |
NekoXP | sveinse, gcc -dumpmachine | 17:08 |
sveinse | NekoXP: Oh? It returns arm-linux-gnueabi only. | 17:10 |
NekoXP | good :) | 17:11 |
NekoXP | sveinse, gcc -dumpspecs | 17:11 |
NekoXP | that's every default | 17:11 |
sveinse | NekoXP: How do I read it? | 17:13 |
sveinse | I find the specs hard to read. Then perhaps touch foobar.c ; gcc -c -v foorbar.c is easier to understand... | 17:16 |
sveinse | Let me ask it another way: We cross compile against natty with options "-march=armv7-a -mtune=cortex-a8 -ftree-vectorize -mfloat-abi=softfp -ffast-math -mfpu=neon -fno-omit-frame-pointer" | 18:00 |
sveinse | 1) Should this line be changed when using precise? | 18:01 |
sveinse | 2) What should the options look like when cross compiling for armhf | 18:01 |
xranby_ac100 | sveinse: when cros compiling for armhf you need to use -mfloat-abi=hard and make sure you link against a hard compiled userspace like libgcc | 18:56 |
LetoThe2nd | oO( "hard compiled userspace" sounds sweet ) | 18:58 |
sveinse | xranby_ac100: Does the gcc-arm-linux-gnueabi packages include hf libgcc etc? | 19:01 |
xranby_ac100 | sveinse: gcc-arm-linux-gnueabihf for armhf | 19:02 |
sveinse | xranby_ac100: How could I miss that.. Thanks | 19:03 |
xranby_ac100 | sveinse: if you use the ubuntu cross compile toolchains then you do not need to pass any extra tune options, they are set by default like thumb and armv7 optimizations | 19:04 |
xranby_ac100 | sveinse: to clarify if you run gcc -E -v | 19:05 |
xranby_ac100 | then gcc will print all default compiled in options | 19:05 |
sveinse | One option ubuntu gcc does not specify is the -mtune= (iirc). Does it have much effect specifying it compared to not? | 19:06 |
xranby_ac100 | well they usually pass --with-arch=armv7-a | 19:06 |
sveinse | I've always wondered if it really meant something to have the tune as well as arch. | 19:07 |
xranby_ac100 | i cant say off hand without benchmarking if you gain anything by explicitly tune for cortex-a8 | 19:07 |
sveinse | Anyways, do you happen to know if ld in 4.6 is compiled with sysroot support this time? | 19:07 |
sveinse | (If you don't, I'll figure it out) | 19:08 |
xranby_ac100 | if your code are to be run on neon systems then that do make a little difference | 19:08 |
xranby_ac100 | since ubuntu do not enable neon by default to stay compatible with tegra2 cortex-a9 cpus | 19:08 |
Cyberworm | hi | 19:45 |
Cyberworm | I have a problem booting Ubuntu 11.04 on the Pandaboard | 19:46 |
Cyberworm | it freezes at "Enabling oem-config" | 19:46 |
Cyberworm | 11.10 works | 20:05 |
=== Termana is now known as Guest26482 | ||
=== robclark_ is now known as robclark | ||
Cyberworm | but I actually need 11.04 | 20:32 |
Cyberworm | sooo | 20:32 |
Cyberworm | is this a known bug? | 20:32 |
xranby_ac100 | Cyberworm: i think its a known bug. you can remove the /var/lib/oem-config/run file to let the board boot up | 20:36 |
xranby_ac100 | using 11.04 | 20:36 |
Cyberworm | isn't that neccessary for something? | 20:36 |
xranby_ac100 | the oem-config are the setup wizard | 20:36 |
Cyberworm | ah | 20:37 |
Cyberworm | thank you | 20:37 |
xranby_ac100 | youre welcome | 20:38 |
Cyberworm | huh | 20:56 |
Cyberworm | now it doesn't freeze | 20:56 |
Cyberworm | I guess it depends on how you format the sd card | 20:56 |
Cyberworm | on the first try I used to gparted to delete the partitions and then I used the command to transfer the image to the card | 20:56 |
Cyberworm | now I didn't format at all and just transfered the image | 20:57 |
xranby_ac100 | Cyberworm: so.. everything are working now? great, | 21:00 |
Cyberworm | well let's see if the setup works | 21:01 |
xranby_ac100 | what feature do 11.04 offer that is not found in 11.10 and later? | 21:02 |
xranby_ac100 | i am curious why you need to use 11.04 | 21:02 |
Cyberworm | xranby_ac100: there are problems with the player/server robot device server on 11.10 | 21:08 |
xranby_ac100 | Cyberworm: have you checked if this are fixed in 12.04 that gets released next week? (you can get install images today to test) | 21:09 |
Cyberworm | I don't know | 21:09 |
Cyberworm | but we need it for a project so it's better for us to use something where we know it will work | 21:09 |
xranby_ac100 | thats fine, in which language are the robot programmed in? | 21:10 |
Cyberworm | dunno, we just need player/server for simulation | 21:11 |
xranby_ac100 | Cyberworm: ok, good luck with the project.. if you get time left over after its finished.. please come back and i we can check out what the player/server depend on | 21:14 |
Cyberworm | I'll consider it, thanks | 21:14 |
sveinse | What is the state of armhf compared to armel (and precise)? I.e. can I consider migrating my embedded system from natty armel to precise armhf? | 21:30 |
=== Guest26482 is now known as Termana | ||
jhobbs | A/wg 8 | 22:16 |
jeinor | sveinse: as it was explained to me, Ubuntu will officially only support armhf from 12.04. All packages in apt repositories should already exist compiled for armhf. | 22:19 |
jeinor | sveinse: the difference is about the same as the difference between i386 and amd64 in terms of "will my program run or not" | 22:20 |
jeinor | sveinse: it might run using multiarch | 22:20 |
jeinor | sveinse: someone else can probably explain it better and more correct | 22:20 |
sveinse | jeinor: Perhaps my question was a bit, uhm, blunt. We're have been running natty armel on an embedded product for 1.5 years now and its working great. For the next version we're definitely going to upgrade to precise. The question was if we should consider hf (as this would give our apps a huge set of improvements) | 22:24 |
sveinse | Support is not the biggest issue | 22:25 |
jeinor | sveinse: I see. Well, in that case I think there are people better suited to answer your question than me. | 22:25 |
jeinor | :) | 22:25 |
sveinse | thanks, anyways | 22:26 |
mythos | i think, the question is already answered | 22:30 |
mythos | armhf is the supported port for precise. so if your hardware is compatible and your software compiles and is able to pass your testframework (*laughs a little*), go ahead and use precise/armhf | 22:32 |
sveinse | test framework, what is that? :P | 22:33 |
mythos | that's why i laughed =P | 22:33 |
Cyberworm | updating packages takes really long... | 22:34 |
sveinse | So if I should make up my own conclusion it would be: armel and armhf will be supported the same way. There is a larger set of FTBFS in armhf. -- yeah, I think we can do a testrun at least | 22:37 |
sveinse | dist-upgrading armel to armhf on a running system is NOT going to be easy, right | 22:38 |
mythos | easy? i think, thats impossible | 22:38 |
sveinse | Not completely I think | 22:39 |
mythos | nothing ist completely impossible | 22:40 |
sveinse | What makes it hard is because of glibc and libgcc? | 22:42 |
sveinse | E.g. you can't have both armel and armhf on the same system, right? | 22:42 |
mythos | because armel and armhf is not abi-compatible, so it can break _everywhere_ | 22:43 |
mythos | not worth the time to try it | 22:43 |
mythos | with multi-arch, you can | 22:43 |
sveinse | abi incompatible across kernel - userspace as well? | 22:43 |
mythos | the kernel does not care, if it is compiled for armel or armhf | 22:44 |
mythos | and it does not care, what the userspace uses (to be clear) | 22:45 |
sveinse | Yes, so the armel libs and the armhf libs could co-exists | 22:46 |
sveinse | except they clash on file-basis, right | 22:47 |
mythos | with multi-arch, not with editing the sources.list form armel and armhf and do a dist-upgrade | 22:47 |
sveinse | What is multiarch specifically btw | 22:48 |
mythos | the new hot stuff | 22:48 |
sveinse | :D | 22:48 |
mythos | sorry, i'm not sophisticated enough with the basis of multiarch | 22:48 |
Cyberworm | it takes reaaaally long for updates | 23:22 |
Cyberworm | is that a normal thing? | 23:22 |
mythos | sorry, Cyberworm. what are you doing? | 23:34 |
Cyberworm | updating | 23:35 |
mythos | from what to what? | 23:36 |
Cyberworm | just the ubuntu 11.04 packages | 23:36 |
mythos | so a "apt-get update" | 23:38 |
Cyberworm | yes | 23:39 |
mythos | i would say, that it is not the normal behaviour (except your hardware (cpu, persistent memory), internet connection, or anything else is very slow) | 23:49 |
Cyberworm | it's on the pandaboard | 23:53 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!