=== ScriptRipper_ is now known as ScriptRipper | ||
lool | Oh we have two ARM emulators for FSL devices in the archive outside of Qemu | 10:31 |
---|---|---|
lool | softgun can emulate imx21 and can run redboot | 10:31 |
ogra | even the imx51 one ? | 10:37 |
ogra | that would be perfect for setting up the images | 10:38 |
lool | No, just imx21 | 10:51 |
lool | Hmm /me just realized something really stupid :-/ | 13:37 |
ogra | good that you relized it :) | 13:38 |
lool | My perl implementation of fconfig is falling short in front of redboot storing IP addresses and ESA addresses as the bytes of the raw C data structure | 13:38 |
ogra | *realized | 13:38 |
lool | I don't see how I'll match that in perl | 13:38 |
ogra | didnt you have that C tool ? | 13:38 |
ogra | from the source you üpüointed me to | 13:38 |
lool | I had a C tool, but it was crap | 13:39 |
lool | For fconfig | 13:39 |
lool | I started fixing it for a bunch of things, but I gave up when faced with fitting in the addition / changing of values | 13:40 |
lool | ogra: 14:40 < lool> I started fixing it for a bunch of things, but I gave up when faced with fitting in the addition / changing of values | 13:44 |
persia | lool, Can't you use pack() and unpack() to replicate a struct? | 15:07 |
lool | persia: I can, but the struct might change from platform to platform | 15:09 |
lool | it's supposed to be opaque | 15:09 |
persia | Oh, yeah. If the struct isn't constructed in such a way to be independent, then you're out of luck :( | 15:10 |
Martyn | Good morning | 16:01 |
lool | persia: Yes, that's exactly the problem; even the on-flash format for a *bool* can vary between 2 and 4 bytes, depending on sizeof(bool)... | 16:04 |
lool | A bool! | 16:04 |
lool | 4 bytes! | 16:04 |
Martyn | Yep. | 16:08 |
Martyn | lool : Well, depending on the platform, compiler, etc... sure. | 16:08 |
Martyn | lool : These days, where people are no longer thinking in therms of bits and bytes, but rather kilo and mega ... to those developers, why not use an int for a bool? *sigh* | 16:09 |
Martyn | I'm not saying it's right, mind you .. just that it's common. | 16:09 |
Martyn | lool : You'll cry .. but on the PS2, on that cell processor ... a bool takes up a whopping full 128 bits. | 16:10 |
Martyn | There are app notes that say "if you need a boolean, consider allocating an int and using flags..." | 16:10 |
ogra | lool, well, indeed TRUE is four letters ... :) | 16:11 |
Martyn | ogra : *groan* | 16:12 |
Martyn | ogra : So, I'm still stuck without a serial dongle for the board for at least another week. Is there any possibility I could get you to make an image that has the redboot kernel cmdline settings appended with "text"? | 16:13 |
Martyn | ogra : Since the Redboot does hash itself, and thus I can't twiddle the bits manually at the moment with a hex editor... | 16:14 |
ogra | yes, i saw your question, but sorry, i have to much on my plate atm trying to get a new image with a new kernel up | 16:14 |
Martyn | *nod* | 16:14 |
Martyn | So, before I poke into this any deeper.. I thought Freescale had committed to porting u-boot (albeit this was posted on a forum back in '04) ... do you know if they are still responsible for it? | 16:25 |
Martyn | ogra: Are you using a modified Redboot? | 16:40 |
Martyn | Actually, that's a general Q to anyone... | 16:40 |
ogra | i'm using the one from NCommander (which he packaged already but which didnt hit the archive yet) | 16:41 |
* NCommander returns | 16:42 | |
lool | Martyn: The FSL one | 16:46 |
NCommander | wooo, ecosconfig-imx was accepted! | 16:50 |
armin76 | Martyn: what? '04? | 16:50 |
lool | persia: Well concerning IP addresses, I was wrong, the format isn't too crazy as under linux typedef uint32_t in_addr_t | 16:51 |
armin76 | Martyn: i don't think the post i passed to you is from that date | 16:51 |
armin76 | haha | 16:52 |
Martyn | armin76 : Yep, twas. | 16:52 |
armin76 | Martyn: you saw the registration date of the guy who posted it | 16:52 |
Martyn | Ah :) | 16:52 |
armin76 | not the date of the post | 16:52 |
armin76 | http://www.powerdeveloper.org/forums/viewtopic.php?p=12019 | 16:52 |
NCommander | yay for X hanging | 16:53 |
armin76 | Posted: Tue Jan 27, 2009 4:25 pm Post subject: i.MX515 Project Guidelines | 16:53 |
armin76 | Martyn: ^ :) | 16:53 |
ogra | NCommander, did you see my question on the redboot package ? | 16:55 |
NCommander | ogra, no I didn't | 16:55 |
ogra | it looks like ecosconfig and redboot come from the same tarball, is that right ? | 16:55 |
NCommander | ogra, no, separate | 16:56 |
ogra | ok | 16:56 |
ogra | phew | 16:56 |
NCommander | Well ... upstream has them together | 16:56 |
NCommander | freescale has them seperate | 16:56 |
ogra | eeek | 16:56 |
ogra | thats what i asked about | 16:56 |
NCommander | I think the version FSL has is before they were merged | 16:56 |
NCommander | Because when I worked on the NSLU2 RedBoot, they were also separate | 16:56 |
ogra | ok | 16:57 |
NCommander | ogra, its kinda weird :-/. If you want to look for yourself, in the redboot zip, the config tools are in tools/ecos_config_tools, and then I built a tarball out of the redboot code and patches out of the stuff in source | 16:57 |
ogra | no, its fine, i just wanted to know if they were one tarball when you recieved them | 16:58 |
NCommander | ogra, well, it was the giant zip from hell :-) | 16:58 |
Martyn | NCommander : I need to change the redboot config in ogra's image .. would you be so kind as to link or send me the zip? | 16:58 |
ogra | if its a) the tarball for ecosconfig and b) the content of the zip tarred up for redboot, thats fine | 16:58 |
NCommander | Martyn, pull the packages out of my PPA, ecosconfig-imx, and redboot-imx | 16:59 |
NCommander | You need to build on an ARM machine with an arm-gnueabi-linux toolchain. | 16:59 |
Martyn | What's the address for your PPA? | 17:00 |
Martyn | I need to add you to the sources.list | 17:00 |
NCommander | Martyn, https://edge.launchpad.net/~mcasadevall/+archive/ppa | 17:00 |
NCommander | Martyn, its a devirtualized PPA so you can also pull binaries down for it on ARM (since you need to build on ARM if you want to just use debuild) | 17:01 |
NCommander | Martyn, what configuration change to you want to make? | 17:01 |
NCommander | Martyn, http://wiki.ubuntu.com/ARM/BuildingBabbageRedBoot - this will likely help you | 17:01 |
Martyn | got it. | 17:02 |
Martyn | ogra : Is the default installed gcc on your image the eabi toolchain? | 17:03 |
Martyn | installed/installable | 17:03 |
NCommander | Martyn, no, this RedBoot been modified to use a linux toolchain vs. an eabi | 17:03 |
NCommander | It won't compile with an eabi toolchain. | 17:03 |
Martyn | foobie. | 17:03 |
NCommander | foobie? | 17:04 |
NCommander | Martyn, what are you trying to do? | 17:07 |
Martyn | Heh, that's what my young cousin used to say when she tried to say "fubar" | 17:07 |
Martyn | NCommander : My babbage shipped without the JTAG/serial dongle. I want to start ogra's live image in text, rather than full X mode | 17:08 |
NCommander | Oh | 17:08 |
Martyn | The tests I'm running need as much free RAM as possible, and the board only has 512... | 17:08 |
NCommander | The config line isn't in the source code, its in the config partition of the image. | 17:08 |
* NCommander notes that its possible to telnet into RedBoot over ethernet, but those options in ogra's image are off. | 17:08 | |
Martyn | Right, and you can't change the partition of the image by hand (it's hashed, and if changed will prevent booting) | 17:08 |
NCommander | Martyn, lool was working on a tool to do just that | 17:09 |
Martyn | Yep, it's not ready yet that I know it. | 17:09 |
Martyn | it/of | 17:09 |
armin76 | Martyn: what kind of serial does it need? | 17:11 |
ogra | armin76, a specially developed external board | 17:11 |
ogra | with mxcuart chip | 17:11 |
lool | Hmm problem | 17:11 |
armin76 | oh... | 17:11 |
lool | fconfig includes the MAC address | 17:11 |
lool | Found key :fec_esa_data: of type :6: sensing :fec_esa: with data length 8; value:42:00:00:EA:18:F0: | 17:12 |
ogra | do we care ` | 17:12 |
ogra | ? | 17:12 |
NCommander | mine isn't set right and it works | 17:12 |
NCommander | we're not using the ethernet port during boot. | 17:12 |
ogra | right | 17:12 |
lool | Yes, but if we keep the SD card for booting, it will have broken networking | 17:12 |
lool | (in redboot only of course) | 17:12 |
Martyn | Only temporarily | 17:12 |
Martyn | right | 17:12 |
lool | Is the ethernet address correctly detected on boot by linux? | 17:13 |
ogra | lool, and who cares ? | 17:13 |
* ogra cant check atm | 17:13 | |
ogra | but if not we can generate one randomly | 17:13 |
Martyn | the mac address is stored in the chip, and is read by the driver early on for most adapters... | 17:14 |
Martyn | ogra : Indeed. As long as you don't use the reserved 00:00:00:00:00:00 address ... it should work fine. | 17:15 |
ogra | right | 17:15 |
Martyn | although generating a random one each boot, may confuse an upstream switch :) | 17:15 |
ogra | well, it should have at least the venmdor part set by default | 17:15 |
Martyn | and it is present on the chip anyway. When linux comes up, the driver will set the mac correctly. | 17:16 |
* NCommander wonders if he should strip out some of the ecos source thats unused | 17:16 | |
NCommander | The copyright file is a freaking mess :-/ | 17:16 |
lool | Martyn: Well sometimes the linux driver just relies on the hardware as initialized by the boot loader to know its MAC | 17:18 |
lool | I know some d-i platforms are parsing the MAC address from the bootloader data | 17:18 |
lool | NCommander: yeah for ecosconfig \o/ | 17:18 |
NCommander | lool, RedBoot is done aside from its copyright file (which just broke 130 lines :-/) | 17:19 |
lool | NCommander: You complain about 130 lines of copyright, I'd like to complain about my stupid 600 lines of perl | 17:22 |
lool | Perl seemed like a good choice, but I'm not sure it was | 17:22 |
NCommander | lool, what are you coding in perl? | 17:22 |
ogra | shell FTW ! | 17:22 |
lool | Still on fconfig | 17:22 |
* NCommander would have choosen C and canobolized the existing code. | 17:23 | |
lool | ogra: I'm confident it was a better choice than shell for this task :-) | 17:23 |
ogra | nah | 17:23 |
ogra | :) | 17:23 |
NCommander | autoconf has long providen you can do ANYTHING in sh | 17:23 |
lool | One reason it's useful to have it in C (and not C++ or Perl or anything) is to have it in an initrd rootfs; otherwise the libs suck a lot of space | 17:23 |
lool | NCommander: Yeah, and it has "providen" that it's ugly and nobody gets it right too :-P | 17:24 |
ogra | you think to far in advance | 17:24 |
lool | ogra: That's actually the reason fis.cpp was rewritten in C for NSLU2 :-) | 17:24 |
ogra | we really dont need it in initramfs atm | 17:24 |
ogra | its enough to have it in the rootfs | 17:24 |
lool | ogra: I'm just mentionning it for completeness in this discussion; I do think there could be value in having it in initramfs for d-i | 17:25 |
lool | Anyway | 17:25 |
lool | I can parse the full config now; let's write it from scratch | 17:25 |
ogra | d-i is out of scope as well for jaunty | 17:25 |
ogra | all we wnat is to write the initial config and the initramfs and kernel binaries to SD for now | 17:26 |
lool | Yes | 17:26 |
lool | ogra: Would have I thought of that though, I would have pursued the C version, as ugly as it was, or rewritten one | 17:27 |
ogra | thats what i thought you were doing until this morning | 17:27 |
* ogra curses ... no input possibility in the latest kernel | 17:28 | |
lool | ogra: A C version? | 17:31 |
ogra | lool, the tool you pointed me to the other day | 17:33 |
lool | ogra: Well the fconfig one was awful in the implementation | 17:35 |
ogra | but it worked ? | 17:35 |
lool | I fixed it up for the read code path, but the write code path was too awful | 17:35 |
lool | ogra: It didn't have any write support and that required a lot of things | 17:35 |
ogra | hmm, i thought the site you pointed me to talked about write support | 17:37 |
lool | ogra: For fis.c probably | 17:37 |
ogra | right | 17:37 |
lool | ogra: Two different upstreams | 17:37 |
lool | fconfig is wholy different | 17:37 |
lool | It's a separate project | 17:37 |
lool | I found 3 fis implementations and one fconfig one; the fconfig one was the problematic one | 17:38 |
lool | Well TBH I'm not perfectly happy with the fis ones, but it's ok, we can use them as is | 17:38 |
ogra | fputs("Read, write or list Redboot configuration\n", stdout); | 17:38 |
ogra | fputs("usage: fconfig [-r|-w|-l]" | 17:38 |
ogra | " -d dev -o offset -n nickname -x value\n", stdout); | 17:38 |
ogra | fputs("'dev' may be a char device, block device or a file\n", stdout); | 17:38 |
ogra | fputs("'offset' is the offset in bytes on the device\n", stdout); | 17:38 |
ogra | fputs("Supported types: \n", stdout); | 17:38 |
ogra | for (i = 0; i < NUM_TYPES; i++) { | 17:38 |
ogra | printf(" - %s\n", TYPE_NAME(i)); | 17:38 |
ogra | } | 17:38 |
ogra | fputs("Additional switches: \n", stdout); | 17:38 |
ogra | fputs(" -v:\tVerbose mode, use more to increase verbosity\n", stdout); | 17:38 |
ogra | fputs(" -s:\tSilent mode, absolutely no messages printed\n", stdout); | 17:38 |
ogra | thats what i see in the source | 17:38 |
ogra | i would have expected -w to be "write" | 17:39 |
lool | Yes, but it overwrites an existing config with the same config | 17:39 |
ogra | ugh | 17:39 |
lool | Exactly what I said: "ugh" :) | 17:39 |
NCommander | ugh | 17:41 |
* NCommander prays he's wrong | 17:41 | |
ogra | gramattically your ugh looks right :) | 17:42 |
NCommander | lool, ogra, we've got a problem :-/ | 17:42 |
NCommander | The RedBoot/ecos tree has 4 clause BSD code intermixed with GPLv2 code. | 17:42 |
* ogra has lots of problems toiday ... no news | 17:42 | |
ogra | fis create initramfs | 17:43 |
ogra | whoops ... wrong focus | 17:43 |
* NCommander sighs, and takes a look upstream | 17:44 | |
lool | NCommander: Hmpf | 17:44 |
lool | NCommander: Where exactly is the 4-clause BSD? | 17:44 |
NCommander | lool, its the ecos network stack, I don't think I can rip it out. | 17:45 |
NCommander | (it looks like someone copy and pasted an older version of the OpenBSD stack ...) | 17:45 |
NCommander | Before they made the change to a newer BSD license (the copyright notices are 1992-1993) | 17:45 |
lool | NCommander: Is this an upstream issue? | 17:45 |
NCommander | Looks like, I'm poking ecos's CVS repo to see if it was resolved in a later version | 17:46 |
lool | Even RedBoot bits are 4-clause, such as packages/redboot/current/include/net/tftp.h | 17:46 |
lool | NCommander: it was not | 17:46 |
NCommander | so its still an issue upstream? | 17:46 |
lool | NCommander: So 1) bring it up on the upstream list 2) mention the issue on debian/copyright and the fact that you contacted upstream about it | 17:46 |
lool | NCommander: It's still an issue upstream | 17:46 |
lool | Well from a checkout some days old, but it doesn't change much | 17:47 |
ogra | worst case it forces the packages to restricted | 17:47 |
NCommander | Er, lool, GPLv2 and BSD 4-clause are incompatible; you can't license code together. | 17:47 |
NCommander | The two licenses conflict with each other ... | 17:47 |
lool | NCommander: But ECos folks did and RedBoot has been shipping in millions of devices | 17:47 |
ogra | right | 17:47 |
* NCommander remembers he was asked this question on his P&P for NM ... | 17:48 | |
NCommander | I thought the adversing clause constitued an additional restriction which breaks the GPL. | 17:48 |
ogra | debian ... | 17:48 |
lool | NCommander: It does | 17:48 |
NCommander | I'm lost then ... | 17:48 |
lool | ogra: Did you start looking into cdimage bits for SD card images for babbage? | 18:41 |
lool | NCommander: Where is the redboot source you're using ATM? | 18:56 |
NCommander | lool, on my HDD. | 18:57 |
NCommander | lool, I'm uploading a copy to REVU in a few minutes | 18:57 |
lool | NCommander: I just need to check values of redboot configs; any older version is fine | 18:57 |
lool | I'm grabbing redboot-imx from your ppa | 18:57 |
NCommander | lool, that's fine. | 18:58 |
* NCommander is just determining if this package is going to universe or to multiverse | 18:58 | |
NCommander | I don't think we have anything in main with four-clause BSD (which could be problematic for anyone adversing), but on the flip side, debian-legal says 4-clause BSD can go in main | 18:59 |
NCommander | (debian main) | 18:59 |
* lool curses Xorg crashes when doing development | 19:08 | |
NCommander | lool, I'm suffering from the random USB device insertions taking X down. | 19:18 |
Martyn | Is there a full ubuntu 9.04 live image available for the OMAP3 evm board built by mistral? | 19:22 |
* Martyn reads through the scrollback ... | 19:22 | |
* Martyn is with ogra -- shell FTW | 19:23 | |
Martyn | NCommander : You have it right. 4-clause and GPL are mutually incompatible | 19:25 |
Martyn | NCommander : LGPL and 4-clause may be compatible, depending. | 19:26 |
NCommander | Martyn, there is an exception in the GPL code in this package to allow it to be linked. | 19:26 |
Martyn | NCommander: Sounds like the lesser GPL then | 19:26 |
NCommander | Martyn, no, its GPL with an exception clause | 19:27 |
NCommander | (there is a difference) | 19:27 |
Martyn | Weird. | 19:27 |
Martyn | Cool but weird :) | 19:27 |
* Martyn doesn't care about the licenses much when developing .. I just Want It To Work(tm) | 19:28 | |
Martyn | If it means I have a rewrite a section of code afterwards, the development goes faster. | 19:28 |
Martyn | ogra : I may have a network driver bug to report. Should I post it to Launchpad? | 19:30 |
Martyn | Rx stops completely, and in a way I can reproduce, when fragmented packets are present. | 19:31 |
Martyn | (I've been doing some lighttpd load testing. It occurs on the Babbage, but not on the EVM or Beagle) | 19:32 |
Martyn | Tx still sends packets, Rx does not receive any, and ifconfig reports no further packets being counted. | 19:33 |
faso | Hi All...I am trying to get jaunty on my N810 | 22:17 |
faso | I am trying to follow instructions http://www.internettablettalk.com/forums/showthread.php?t=25975... | 22:18 |
faso | I am unable to do debootstrap | 22:18 |
faso | I wonder why | 22:18 |
faso | appreciate any help? | 22:19 |
NCommander | faso, what was the error? | 22:32 |
faso | actually I am unable to locate debootstrap in any of the repositories | 22:35 |
faso | Do u know where I can install it from | 22:35 |
faso | ? | 22:35 |
lool | faso: It's in all Ubuntus | 22:57 |
lool | I mean available to them | 22:57 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!