**** BEGIN LOGGING AT Thu Apr 17 02:59:57 2008 Apr 17 11:18:43 hi ... anyone in here happens to have any experience with getting "Angstrom" up and running on an NSLU2 ? Apr 17 11:19:16 so far i can flash the slug just fine ... but it seems to simply not boot up afterwards Apr 17 12:58:34 Angstrom requires a serial console on the NSLU2 at this moment. You should consider SlugOS or perhaps Debian instead. Apr 17 13:08:16 mwester: we've actually been using SlugOS so far ... (with great success) ... it's just that from my understanding the ARM EABI provides significantly increased Floating Point performance, and things are starting to look like that would be a very welcome benefit Apr 17 13:08:28 mwester: hence why i started looking into Angstrom Apr 17 13:10:22 We are currently targetting EABI for SlugOS 5, but there is no planned release date. We intend to narrow the gap between Angstrom and SlugOS over time, just to ease support, but Angstrom lacks many SlugOS features that are specific to the type of devices liek the NLSU2. One of those is the total lack of a display/console. Apr 17 13:10:56 Adding a serial port is really not difficult, see the wiki for information. Apr 17 13:11:37 mwester: yeah i know ... we've done so before ... Apr 17 13:12:09 however ... if SlugOS is targeting EABI for SlugOS 5 ... that would simply remove the interest in Angstrom :) Apr 17 13:14:14 Except for the time factor; that'll be a big shift and can't be done quickly (new feeds,etc) Apr 17 13:16:25 honestly at the time being FP performance isn't that big of an issue yet ... for now it's mostly used for Data Acquisition purposes ... once we want to start processing the acquired data as it comes in ... that's when FP performance may become a concern ... though with some twiddling we might just be able to do it all in fixed point ... time will tell Apr 17 14:16:31 03oleo * r8205 10optware/trunk/make/transmission.mk: transmisission: 5611->5633 Apr 17 15:15:16 perhaps i am just being silly but i can't seem to get it to work ... what do i need to put into my slugos/conf/local.conf in order to make it include additional packages into the generated firmware image ? Apr 17 15:16:18 so far it looks like DISTRO_EXTRA_DEPENDS += "openntpd" e.g. causes the openntpd package to be built .... however DISTRO_EXTRA_RDEPENDS += "openntpd" seems to have no effect upon whether or not it is included into the built firmware image Apr 17 15:17:20 This is a bug in bitbake/OE. Apr 17 15:18:04 any way to work around this ? Apr 17 15:18:22 it's a little awkward having to ipkg install the packages afterwards Apr 17 15:18:31 Hmm. actually not - it seems this bug was fixed. Apr 17 15:18:38 So, perhpas it's something else. Apr 17 15:18:44 in that case what am i doing wrong ? Apr 17 15:18:57 Did you change this after you did a sucessful build? Apr 17 15:19:39 i.e. did you take special action to ensure that the build system's dependency tracking mechanism knows that it must rebuild the tasks and the rootfs image? Apr 17 15:20:05 mwester: i did ... however i did rm -Rf anything slugos-image related from slugos/tmp/stamps and slugos/tmp/work/ Apr 17 15:20:53 that should normally be enough to trigger a rebuilt of the rootfs not ? Apr 17 15:21:11 No, it just recreates the image from the existing rootfs. Apr 17 15:21:23 You need to remove the stamps for task-slugos-1.0* Apr 17 15:22:02 The task recipe is the one that does the real work of determining what goes into the rootfs, and is where the EXTRA_RDEPENDS is evaluated. Apr 17 15:22:32 aaah i see Apr 17 15:22:36 ok ... let me try that Apr 17 15:23:43 Just so we know what to consider for SlugOS in the future, can you share what you're changing if it might be generally applicable? Apr 17 15:24:30 sure ... it's basically just openntpd so the system can synchronize its time out of the box ... and the kernel-module-ftdi-sio along with kernel-module-usbserial Apr 17 15:24:38 probably not general purpose enough :) Apr 17 15:24:58 we need the ftdi-sio module for USB<->RS232 / USB<->RS485 Apr 17 15:25:45 we use it for data acquisition purposes with regards to power consumption / power grid monitoring Apr 17 15:27:11 Ah, ok. You're right, that's probably not general enough. But that's a really cool application! Apr 17 15:29:36 yup :) Apr 17 15:31:06 Ok ...that seems to indeed have done the trick according to the log.do_rootfs Apr 17 15:32:00 reflashing the slug Apr 17 15:32:55 perhaps inclusion of openntpd in the default build isn't that bad an idea ... though ftdi-sio and usbserial are definitely not Apr 17 15:33:15 or should i say ... definitely are Apr 17 15:38:18 hmmmm .... openntpd did get installed into the firmware image ... but somehow it can't start since there now is no ntpd user Apr 17 15:38:35 i guess something is a little "oddball" there Apr 17 15:40:38 yup ... did an ipkg remove openntpd followed by an ipkg install openntpd and it looks like during the configuring openntpd stage it automatically adds the ntpd user Apr 17 15:41:13 i guess i'll stick to installing openntpd afterwards for the time being ... Apr 17 15:42:08 would you happen to have any idea how this could be fixed ? Apr 17 16:10:06 Caelian: That should be automagically handled by the package post-install script, which is run at early boot time. Did you reboot after the intiial boot to see if the second boot had things right? Apr 17 16:10:33 In a cross-dev environment, we defer the post-install scripts until the first boot of the target device. Apr 17 16:21:06 mwester: hmmmm ... i didn Apr 17 16:21:17 i'll retry to see if that indeed did the trick Apr 17 16:31:00 That said, there are some pre-defined user ids in the password file; it's not too tough to add some others. Apr 17 16:33:48 nope ... didn't work ... Apr 17 16:34:16 i guess that would work ... but i am not exactly sure how to integrate such a thing into the current build framework (as a local change) Apr 17 16:35:57 with didn't work i meant ... if i install it afterwards the user does get added ... if i include it in the firmware image the user is not added Apr 17 16:40:08 Try something like this: Apr 17 16:40:12 IMAGE_PREPROCESS_COMMAND += "sed -i -es,^root::0,root:BTMzOOAQfESg6:0, ${IMAGE_ROOTFS}/etc/passwd;" Apr 17 16:40:17 in your local.conf file. Apr 17 16:41:00 That sample changes the root password, but you can just as easily change the sed to a something that will append a line to the file. Apr 17 16:41:51 i see what you mean ... i'll fiddle around with that thanks for the suggestion ... i am going to head off for now though ... worked long enough today as is :) Apr 17 18:00:57 03bzhou * r8206 10optware/trunk/ (4 files in 2 dirs): mysql: 4.1.20 -> 4.1.22; set STACK_DIRECTION to 1 if arm or armeb Apr 17 18:16:17 03bzhou * r8207 10optware/trunk/sources/mysql/comp_err.patch: mysql: added missing comp_err.patch Apr 17 18:17:48 03bzhou * r8208 10optware/trunk/sources/mysql/lex_hash.h: mysql: no longer need the pre-generated lex_hash.h Apr 17 19:52:43 Any plans for 2.6.25? Apr 17 19:57:54 Yep. But unspecified as to when. Apr 17 19:58:04 I was going to take a crack at it next week. Apr 17 19:58:19 cool. Apr 17 19:58:25 there's even more RT stuff in there. Apr 17 19:58:44 hi how are you? Apr 17 19:58:45 I started using the HR timers in 2.6.24 and it made a huge diff for my sensor loop Apr 17 19:59:18 I just want to see if the oom-killer behaves on heavily loaded systems on it. Apr 17 19:59:49 somebody can help me please? Apr 17 20:00:00 i see. Apr 17 20:00:04 macpower, just ask your question Apr 17 20:00:07 macpower: ask your question Apr 17 20:00:14 Then we can choose to ignore it. Apr 17 20:00:15 :p Apr 17 20:00:22 haha Apr 17 20:00:23 i use a converter usb serial onnslu2 Apr 17 20:00:37 Ok. Apr 17 20:00:39 and it's mounted on ttyUSB0 Apr 17 20:00:41 it's normal Apr 17 20:01:06 but for me i use a programming software that it use a serial port ttyS0 Apr 17 20:01:20 and i can't change port on software Apr 17 20:01:42 Does it work as roo? Apr 17 20:01:43 ln -s /dev/ttyUSB0 /dev/ttyS0 Apr 17 20:01:46 root? Apr 17 20:01:46 ? Apr 17 20:02:00 so for use software with serial converter, i have add a symlink : /dev/tty/USB0 /dev/ttyS0 Apr 17 20:02:14 but this is only a virtual link Apr 17 20:02:31 You might be able to configure udev to do what you want. Apr 17 20:02:37 ok Apr 17 20:02:54 but what i configure udev for use ttyS0 native Apr 17 20:02:55 please Apr 17 20:03:30 A) find the developer of that software, and kick them in the ***** for such stupid design, then B) use "ls -l" to find the major and minor device numbers given to /dev/tty/USB0, and C) use "mknod" to create /dev/ttyS0 with the matching major and minor numbers. Apr 17 20:03:47 Use google to find info on mknod, if you are not familiar with that command. Apr 17 20:03:59 i have try mknod Apr 17 20:04:30 Then you'll have to provide more information on what doesn't work with the symbolic link, and with the mknod solutions. Apr 17 20:04:55 and how i use mknod please, i remove ttyS0 Apr 17 20:05:08 mknod ttyS0 -c 8 0 Apr 17 20:05:42 but after ? please Apr 17 20:06:21 http://www.google.com/url?sa=t&ct=res&cd=1&url=http%3A%2F%2Fwww.faqs.org%2Fdocs%2Flinux_admin%2Fx797.html&ei=rq0HSKmQGqOKiAGSzuGtCQ&usg=AFQjCNGb7p7MIplLU79_PYn1MZAVvTlNzA&sig2=iPk_5aOlHwQkDgfXyBQ0bA Apr 17 20:06:27 First hit from google give an example. Apr 17 20:06:38 Find the numbers you need from the output from "ls -l /dev/tty/USB0" Apr 17 20:07:28 I don't have one of those devices plugged in, so I don't know what the numbers should be. Apr 17 20:07:35 I see that mwester has had similar frustrations with devs hardcoding device names in their code Apr 17 20:08:13 lol, you do wonder if people find it so hard to use variables.. Apr 17 20:08:24 Yep. And also bad experiences with devs who assume that the device name is always /dev/ttyS... Apr 17 20:08:44 They get the variable right, but they put in too much assumption. Apr 17 20:09:33 But it isn't just Linux devs; the windows apps that only know about four serial ports are also a big problem. All my bluetooth serial ports show up >3. Apr 17 20:09:55 ok Apr 17 20:10:00 if i create a tyS0 Apr 17 20:10:12 ttyS0, it's not according to ttyUSB0 Apr 17 20:10:12 ? Apr 17 20:10:30 I don't understand your question? Apr 17 20:11:05 after i have create a ttyS0 bu Apr 17 20:11:09 by mknod Apr 17 20:11:32 i have a ttyS0 and ttyUSB0 Apr 17 20:11:44 by i don't use ttyS0 for ttyUSB0 Apr 17 20:11:55 but a symlink don't work Apr 17 20:12:46 Linux doesn't go by the name -- it goes by the major and minor numbers that are associated. By creating ttyS0 with the same numbers as ttyUSB0, as far as Linux is concerned they are one and the same device. Apr 17 20:13:27 A symlink may not work, depending on how the application is written, and apparently your application doesn't like sym links. Apr 17 20:13:52 so i can't create automaticly on boot a device that associed to ttyS0 from ttyUSB0, in udev? Apr 17 20:14:55 That's what udev is for. But you'll have to figure out the correct changes to make to the rules. Apr 17 20:15:45 this is can work : KERNEL=="ttyUSB[0-9]*" SYMLINK+=""ttyS0"s ? Apr 17 20:16:34 ? don't know - you'll have to try it. Apr 17 20:17:00 no i can't try, i don't have my plateforme Apr 17 20:17:07 i try tommorow Apr 17 20:18:25 but i need to remove a line on udev when i add this line? Apr 17 20:28:51 thanks for all Apr 17 20:28:56 bye good night Apr 17 22:36:18 haha Apr 17 22:36:31 googling "using cowsay" brings up the NSLU2-linux svn and the cowsay mk file Apr 17 22:37:21 mooo? Apr 17 22:37:47 joshin, what firmware do you want in your slug? Apr 17 22:38:01 I'm currently testing with 4.9Alpha Apr 17 22:38:35 slugos-be 4.9a Apr 18 00:12:52 ka6sox-laptop: doesn't matter. I'm going to be trying to build a 2.6.25 kernel for it. Apr 18 00:13:30 See if they finally figured out how to leash the oom-killer Apr 18 00:15:25 I'mjust happy that you've been able to fatten my slug! :) Apr 18 00:29:51 ooh **** BEGIN LOGGING AT Fri Apr 18 01:21:56 2008 Apr 18 01:29:59 #ubuntu-us-ut **** ENDING LOGGING AT Fri Apr 18 02:59:56 2008