**** BEGIN LOGGING AT Mon Apr 09 02:59:56 2007 Apr 09 07:42:29 this just isn't my day Apr 09 07:44:00 <[florian]> why ? Apr 09 07:45:18 after hours of mussing about, i determined that the serial console flakiness was primarily due to metal fatigue at the solder joints from the wires to the device Apr 09 07:45:49 then, one by one, i resoldered each wire as each other wire broke Apr 09 07:46:59 i don't work well under pressure, and after a few hours of this, i managed to disengage the solder pad itself on pin 1 from its trace Apr 09 07:47:08 that's the Tx pin Apr 09 07:47:33 so, i created a solder bridge to take the place of the trace Apr 09 07:48:13 i now consistently get a response on the serial console, but it's not really readable because it misses bits Apr 09 07:50:20 <[florian]> bad day Apr 09 07:50:24 yes Apr 09 07:50:46 i'm not 100% ready to give up, though Apr 09 07:51:01 <[florian]> why give up ? Apr 09 07:51:28 i meant that it's time for me to ask for ideas Apr 09 07:53:33 <[florian]> I would advise you not to touch a soldering iron :) Apr 09 07:53:59 btw, during it all, at one point i managed to have the serial console working long enough to test that build, and it got farther than 2.6 ever did before, but still not all the way because i made a stupid mistake that, hopefully, is fixed by the current build Apr 09 07:54:25 yeah, that's what i get for trying to solder on the floor Apr 09 07:55:28 i don't think my back will forgive me for some time Apr 09 07:58:22 <[florian]> sn9_: ok, by further, you mean, whereto ? Apr 09 07:59:13 it mounted the root, which it did not before Apr 09 07:59:31 <[florian]> great Apr 09 08:02:02 i've just updated big.patch with what should allow it to boot completely (assuming squashfs; i have not tested the jffs2 images at all yet), but can't really test it. it can't be worse than the last build i tested (the one that mounted the root) Apr 09 08:02:37 <[florian]> ok, I will redownload Apr 09 08:02:42 once you download it, you can probably commit it straight away Apr 09 08:05:14 <[florian]> I see where you want to go Apr 09 08:05:28 your typo in 000-rdc_fixes.patch took me a couple of days to track down Apr 09 08:07:57 any objections? Apr 09 08:08:13 <[florian]> yes, all find commands Apr 09 08:08:43 oh, that was because i simply piped the output of "svn diff" Apr 09 08:08:58 <[florian]> no need to include sysdep.h in the flash map driver, use linux types Apr 09 08:08:59 i had to change those locally Apr 09 08:09:24 there are linux types for the image header format? Apr 09 08:10:29 <[florian]> it uses defined types, whereas linux types should be used Apr 09 08:10:53 lemme refresh my memory bu taking a look at the file... Apr 09 08:10:58 *by Apr 09 08:11:24 <[florian]> I fixed that locally, it is pretty minor Apr 09 08:12:56 ok, now i remember: imghdr.h used those types Apr 09 08:13:17 <[florian]> yes Apr 09 08:14:00 florian * r6911 /trunk/target/linux/rdc-2.6/ (7 files in 4 dirs): Fresh fixes for rdc from sn9, thanks ! Apr 09 08:16:45 your correction to the makefile is incomplete Apr 09 08:17:19 <[florian]> no Apr 09 08:17:34 <[florian]> there is no need to disable initramfs as you do Apr 09 08:17:43 i don't Apr 09 08:17:47 i enable it Apr 09 08:17:51 <[florian]> hum Apr 09 08:18:02 it is an absolute requirement Apr 09 08:18:22 it will not boot otherwise Apr 09 08:18:57 <[florian]> I do not understand why Apr 09 08:19:15 <[florian]> initrams is enabled only if you want to in "Target images" Apr 09 08:19:25 no Apr 09 08:19:33 that is different Apr 09 08:19:47 that is to make the root an initramfs Apr 09 08:19:55 i do not affect that Apr 09 08:21:46 btw, what's the significance of ioremap_nocache()? Apr 09 08:24:14 <[florian]> it prevents the memory from being cached Apr 09 08:25:25 <[florian]> both should work, just ioremap_nocache is safer Apr 09 08:26:06 ok Apr 09 08:26:14 why is 003-rootfstype.patch doubled? Apr 09 08:27:58 <[florian]> what do you mean by doubled ? Apr 09 08:28:18 look in it Apr 09 08:28:44 <[florian]> ok, did not see that Apr 09 08:29:22 florian * r6912 /trunk/target/linux/rdc-2.6/patches/003-rootfstype.patch: Fix patch doubling Apr 09 08:29:27 also, you did not add the rootfs file, and did not fix airlink.c Apr 09 08:30:07 florian * r6913 /trunk/tools/firmware-utils/src/airlink.c: Fix airlink image generation tool Apr 09 08:30:25 <[florian]> I am not at all fond of the rootfs fixes, so I will try finding something better Apr 09 08:31:40 why? Apr 09 08:32:02 <[florian]> because you are registering statically devices Apr 09 08:32:23 only for boot; they disappear as soon as root is mounted Apr 09 08:33:33 er, when the root is switched, rather Apr 09 08:39:03 let me explain the rationale a little: Apr 09 08:40:57 openwrt's rootfs patch assumes that mtdblock0 is root, but here, that's the kernel, and mtdblock1 is root. for this reason, the kernel cannot automatically mount the root unless the /dev/mtdblock1 node is present Apr 09 08:42:02 that means it has to go into a tiny initramfs Apr 09 08:42:52 but the normal openwrt make rules explicitly disable the use of an initramfs except in cases where it's root Apr 09 08:43:30 the makefile addition you didn't include overrides that Apr 09 08:45:14 this is the simplest and most straightforward way to handle the rdc weirdness Apr 09 08:45:19 <[florian]> would it be easier to change the openwrt rootfs patch to use mtdblock1 instead of jumping into an initramfs then mount the root ? Apr 09 08:45:38 that would break all existing installations Apr 09 08:46:28 besides, i do not want to enable that patch for rdc, as it somewhat conflicts with the flash mapping Apr 09 08:47:30 <[florian]> since you do not override the kernel command line, root=/dev/mtdblock1 anyway Apr 09 08:47:37 exactly Apr 09 08:47:38 <[florian]> is used anyway, sorry Apr 09 08:48:10 <[florian]> I think I see what you want to do Apr 09 08:48:26 it is used, but without the initramfs, it is not found Apr 09 08:48:49 <[mbm]> you could change the mtd order Apr 09 08:49:13 [mbm]: you haven't seen the crazy bootloader code Apr 09 08:53:19 <[mbm]> I mean how you assign the mtd partitions Apr 09 08:53:24 yes Apr 09 08:53:35 <[mbm]> hrm Apr 09 08:53:47 it has to play nice with the bootloader Apr 09 08:53:54 that reminds me... Apr 09 08:54:03 * [mbm] should take a look at it.. pretty sure I have a few rdc boards that I never got around to reflashing Apr 09 08:54:30 currently, the kernel is bigger than the bootloader allows without horrendous workarounds Apr 09 08:54:50 an lzma loader is very badly needed Apr 09 08:55:52 [florian]: the initramfs is only 196 _bytes_, anyway Apr 09 08:57:47 sn9_: there's lzma for x86 just never got around adding it :) Apr 09 08:58:03 sn9_: I was too busy with other stuff Apr 09 08:58:13 this is x86, too Apr 09 08:58:29 <[mbm]> think that was the point Apr 09 08:58:37 ah Apr 09 08:58:38 ;) Apr 09 08:59:10 want a link so you can fix it up? (but add it to both x86 generic and rdc then) Apr 09 08:59:43 i should see a successful boot with the workarounds, first Apr 09 08:59:57 http://www.linuxfromscratch.org/patches/downloads/linux/linux-2.6-lzma-1.patch Apr 09 09:00:46 <[mbm]> hmm not a standlone wrapper? Apr 09 09:01:56 i think that the initramfs and makefile addition are worth committing for now Apr 09 11:57:34 florian * r6914 /packages/net/ntpd/Makefile: Fix postinstallation script errors Apr 09 12:46:35 right...so now I have several devices with OpenWrt on it... Apr 09 12:46:40 need to find a good use for it :) Apr 09 12:47:17 nagios would be great... Apr 09 12:54:05 <[florian]> still working cross-compiling it Apr 09 12:55:52 ok Apr 09 20:36:56 <[mbm]> hmm.. we should change the mtd utils error message Apr 09 23:25:43 pavlov * r6915 /packages/libs/postgresql/ (7 files in 3 dirs): postgresql upgrade to 8.2.3, bringing in the server as well... this needs some data dir decisions to be made. to be discussed. **** ENDING LOGGING AT Tue Apr 10 02:59:57 2007