**** BEGIN LOGGING AT Sat May 07 23:59:56 2005 May 08 00:12:36 ~seen rwhitby May 08 00:12:41 rwhitby <~rwhitby@rwhitby.nslu2-linux> was last seen on IRC in channel #nslu2-linux, 7h 29m 34s ago, saying: 'dunno'. May 08 00:13:05 ~seen rwhitby-away May 08 00:13:06 rwhitby-away is currently on #nslu2-linux (20h 34m 8s). Has said a total of 9 messages. Is idling for 4h 33m 23s May 08 00:15:46 ~seen kergoth May 08 00:15:47 kergoth is currently on #uclibc (4d 2h 51m 51s) #nslu2-linux (4d 2h 51m 51s) #oe (4d 2h 51m 51s) #elinux (4d 2h 51m 51s) #handhelds.org (4d 2h 51m 51s) #ol (4d 2h 51m 51s) #tacobeam (4d 2h 51m 51s). Has said a total of 134 messages. Is idling for 1h 26m 33s May 08 00:42:12 back May 08 00:45:30 ka6sox: whassup? May 08 02:02:20 good morning May 08 02:05:03 morning and good nite. May 08 02:05:10 :) May 08 05:20:17 CIA-1: tell your master that the cia.navi.cx website is broken .... May 08 05:29:12 Does anybody have any recommendations for usb hubs that work with the slug?\ May 08 05:37:28 most should work May 08 05:37:38 hrrm May 08 05:37:54 My slug has died during the night again. May 08 05:38:07 does it matter if they are 2.0 or not? May 08 05:38:10 novacane: I have a Swann 4-port USB 2.0 hub, and a Targus 7-port USB 2.0 hub, and a Targus 4 port USB 1.1 micro hub. May 08 05:38:33 novacane: http://nslu2-linux.thestuffguy.com/gallery/slug-central May 08 05:39:02 Can they be pluged into either port of the slug? May 08 05:39:16 yep May 08 05:39:21 bear in mind that just because a hub works does not imply you can attach four hard drives to an unslung slug and have them work May 08 05:39:45 you probably want a powered hub if the things you are connecting will be drawing any significant current ... May 08 05:39:55 ~seen jbowler May 08 05:40:01 jbowler is currently on #nslu2-linux #oe May 08 05:40:10 yeah, only one mass storage device per disk1/2 port on unslung May 08 05:40:15 jacques, I realize that and that is not what I am trying to do May 08 05:40:23 ok May 08 05:41:02 jbowler: ntp is still crashing my slug. Did you find out what the panic was all about? May 08 05:46:13 Okay, so I have my slug with a harddisk and a soundcard attached, with not hub everything works fine. I have tried 3 hubs a cheap dynex, a dlink 2.0 and a d-link 1.1, with all but the last one the slug wont even boot May 08 05:55:47 are they powered hubs? May 08 05:55:51 jbowler: And drift seems to vary wildly. I've got as low at 129, and as high as 292.037 May 08 05:56:47 rwhitby, yes they are all powered hubs. **** BEGIN LOGGING AT Sun May 08 05:59:38 2005 May 08 06:21:02 hi all May 08 06:21:12 g'day May 08 06:21:39 any news I should care about ? May 08 06:23:17 the sky is falling. Other than that, nah... ;) May 08 06:23:36 Humm. I need for help with a autoconf trouble in a package I'm making now May 08 06:33:11 In configure.ac, I have AC_SEARCH_LIBS(inet_aton, resolv). After running autoconf, configure says "checking for library containing inet_aton... none required". I finally got an autoconf.h with "/* #undef HAVE_INET_ATON */" May 08 06:33:54 the build fails : May 08 06:33:56 ../include/common.h:2261: conflicting types for `inet_aton' May 08 06:34:16 => /opt/brcm/hndtools-mipsel-uclibc-3.2.3-full/include/arpa/inet.h:74: previous declaration of `inet_aton' May 08 06:34:57 inet_aton is defined in ld-uClibc-0.9.19.so May 08 06:43:18 ok May 08 06:44:22 so it looks like the app you are trying to compile is defining it's own inet_aton type which is not the same as uclibc - this app might not have been tested with uclibc May 08 06:44:45 s/it's/its/ May 08 06:45:29 It's seems to compile fine when I manually set #define HAVE_INET_ATON 1 in autoconf.h May 08 06:45:36 hmm May 08 06:45:48 then maybe the test for it is failing with uclibc May 08 06:46:03 That's what I'm thinking May 08 06:49:54 I was wondering about the meaning of "none required" during configure. May 08 06:51:12 For me it is "none required as standard (libc/uclibc) lib has it" May 08 06:51:45 as an example, three lines before, I have "checking for library containing crypt... -lcrypt" May 08 07:04:58 is there a config.log ? May 08 07:05:47 can you get any more detailed information about how it tested for inet_aton and how it interpreted the results of that test? May 08 07:11:05 I though AC_SEARCH_LIBS to be a self contain test May 08 07:11:15 config.log says May 08 07:11:18 checking for library containing inet_aton May 08 07:11:29 => /opt/brcm/hndtools-mipsel-uclibc/bin/mipsel-uclibc-gcc -o conftest -g -O2 -pipe -O2 -pipe -I/data/slug/unslung/staging/opt/include -L/data/slug/unslung/staging/opt/lib -Wl,-rpath,/opt/lib -Wl,-rpath-link,/data/slug/unslung/staging/opt/lib -lresolv conftest.c >&5 May 08 07:12:01 $? = 0 May 08 07:12:07 test -z || test ! -s conftest.err May 08 07:12:11 $? = 0 May 08 07:12:16 test -s conftest May 08 07:12:22 $? = 0 May 08 07:12:26 result: none required May 08 07:13:53 second part of AC_SEARCH_LIBS is the libs where to search for (As I understand). I tried removing 'resolv' as inet_aton is defined in libuclibc but this does not change anything May 08 07:14:03 hmm May 08 07:25:50 bbiab May 08 08:39:06 ~seen NAiL May 08 08:39:08 nail is currently on #nslu2-linux. Has said a total of 6 messages. Is idling for 2h 15m 50s May 08 08:39:31 ping NAiL May 08 08:55:17 I've found out, that php-thttpd with all dependencies consumes a lot of space and it is impossible to fit it into the flash (wlhdd). I would like to ask you about alternatives - httpd's with some sort of scripting (cgi at least). It doesn't need to be very skillful daemon because it will be used just for transition between original firmware and on-disk environment with php-thttpd included. May 08 09:02:47 jbowler: pong May 08 09:04:18 Ah: NTP: my slug crashed about three hours ago May 08 09:05:52 It was up 10 hours, then crashed in the RTC code (xl205 sp?), same as last time - slugbug 141 May 08 09:05:54 yeah, my slug has been running just nicely without ntp. With ntp it crashes after a few hours (less than 8) May 08 09:06:35 Do you have serial? May 08 09:07:49 unfortunately, no. If netconsole works, I can recompile withtthat May 08 09:08:09 Ah ha, it does - I checked the changes in to SVN last night May 08 09:08:46 That's how I got the kernel panic - from a remote syslog. May 08 09:09:57 ok, then I just have to figure out how to get the changes from berlios? May 08 09:10:05 BitKeeper is still broken, no commit possible on anything I've tried with root kergoth@zelda.tuxnami.org|ChangeSet|20030602211901|40961|4f9b117aa66a8ba3 May 08 09:10:51 Yes, anonymous checkouts from Berlios work (at least they did for me while my registration was pending) May 08 09:11:46 I have co'd the tree, but in a separate directory. Dunno how to co so that the changes in berlios get applied to the oe tree I already have May 08 09:11:56 dev checkouts work too May 08 09:12:03 Your NTP drift-drift is only 163PPM max - that's pretty good, 1 part in 10000, the variation depends on the reliability of communication to the NTP server May 08 09:12:47 163? The latest drift file says 293? May 08 09:13:25 NAiL: suppose you have openembedded (bitlooser oe tree) and svn/trunk, cp -r openembedded svn/trunk. Note that SVN won't warn about new files like BitKeeper. May 08 09:14:25 NTP drift: something like min 129, max 293? 293-129 is the variation (164). The drift is just the clock drift over time, it should be a constant value. May 08 09:14:44 aaah May 08 09:15:09 NTP drift: the variation is caused by the clock speeding up or slowing down or by your NTP server itself being wrong. May 08 09:16:38 My variation in drift is <10ppm, which is pretty amazing for backwoods SW Oregon... Clearly NTP server season hasn't started yet here or someone would have shot a bullet at it. May 08 09:17:08 hehehe May 08 09:17:57 jbowler, I didn't know you were in oregon... hmm maybe I did and just forgot May 08 09:19:00 SVN: here's a better rune, something like 'find openembedded \( -name SCCS -prune \) -o \( -type f -print \) | cpio -p svn/trunk May 08 09:20:49 jacques: yeah, had to escape from big cities. Much quieter here, we got DSL a couple of years ago ;-) May 08 09:21:36 jbowler, yeah I escaped from the bay area May 08 09:21:44 i really like oregon May 08 09:23:34 We need an option to set the power LED to 'flashing', then when the kernel panics it will be obvious... May 08 09:29:14 netconsole: needs the changes to ixp425-eth and needs to be modprobed in after eth0 has been set up (something like S41netconsole). Needs a command line argument to say where to send messages, e.g. modprobe netconsole netconsole=@/,514@192.168.1.19/00:30:1B:B3:67:C1 It should the 'just work' May 08 09:30:24 jaro: you can try thttpd. It has cgi capacity May 08 09:30:31 jbowler: There are no other changes than setting CONFIG_NETCONSOLE=m? May 08 09:32:21 NAiL: no, that change has been in for a long time. The specific patch is packages/ixp425-eth/files/pollcontroller.patch this is SVN r3 (see the checkin log) May 08 09:32:37 ah, ok May 08 09:33:26 SVN: anyone know if there is some way of generating a local image of the SVN repo, i.e. like a BitKeeper clone? May 08 09:36:05 check #svn? ;) May 08 09:36:44 Oops, S45mountnfs.sh uses 'mount -a -t nfs' but busybox mount does not support -t, so everything gets overmounted when an nfs mount is added to /etc/fstab... May 08 09:37:02 hi all. is there a good newsgroup-reader/downloader for the nslu2? May 08 09:37:57 tin? ;) May 08 09:39:00 wget May 08 09:39:12 Sorry s/w/n/ - nget May 08 09:40:06 jb: using it? May 08 09:41:38 Wipmac: no, I've used it in the past. Should be an easy port to the slug (because it's wget for news:/) May 08 09:42:00 jbowler: no, there's (atleast not yet) a svn clone May 08 09:42:51 so tin is the way to go for now? May 08 09:43:25 NAiL: that's what I feared. I started to go through the stuff about setting up a repo, but it didn't seem that I could automagically clone another repo into it - SVN seems to be one level, like CVS. May 08 09:45:09 jbowler: Looking into a way to do it, but it's a bit of a hack May 08 09:47:12 NAiL: I thought about flipping the parent (i.e. Repo) pointer, looked at 'svn switch', but then realised I'd have to RTFM ;-) May 08 09:55:05 error when instralling tin: libpcre.so.0: cannot open shared obj... any clues? May 08 10:05:35 do i have to install pearl for tin ? May 08 10:06:21 "pcre" should suffice May 08 10:08:23 thx May 08 10:10:42 in which path are all these .conf's saved? May 08 10:15:49 SVN: hum, I just did 'svn update' and it's downloading another complete set of everything. This I don't understand at all... May 08 10:23:58 jbowler: I noticed the same May 08 10:24:32 rwhitby r5: Rearranged the repo layout to allow a higher level directory containing a build and release Makefile May 08 10:24:49 Unfortunately this seems to leave my changes orphaned in 'trunk' May 08 10:25:05 Also I have to go and re-edit all my setup.env and local.conf files... May 08 10:40:49 bbiaw May 08 10:45:57 Wipmac: homedir/.tin May 08 10:53:32 what i meant was where i setup the nntp server in tin etc.. May 08 11:04:04 I have been trying to get makeinfo for knoppix and I am having trouble May 08 11:04:06 any suggestions May 08 11:08:13 anyone know about linking with php libraries? May 08 11:08:58 should the linker flags be -lphp of -lphp5 or something? May 08 11:42:08 hmm May 08 12:06:16 copperbeech: not sure but if your lib's name is libphp5.so then you have to link with -lphp5. if your lib's name is libphp.so, then linker flag is -lphp and so on (libfoo.so => -lfoo) May 08 12:06:57 ok thanks - I'll have a search for it. May 08 12:08:23 hmm there is no *php*.so in staging - I'll have to check with jp30 to see if there is something I missed. May 08 12:09:01 php-stage doesn't create the so ? May 08 12:10:46 well there's nothing matching that pattern in my staging folder and php is supposed to have built ok. May 08 12:10:53 make php-stage creates a libphp5.so under staging/opt/lib May 08 12:11:20 it builds under builds and stage under stage. Did you stage it ? May 08 12:11:33 yep May 08 12:11:37 humm May 08 12:12:04 [mcneillm@Coppermine]~/unslung $ find staging/ -name '*php*.so' found nothing May 08 12:18:42 [mcneillm@Coppermine]~/unslung $ find /opt/ -name '*php*.so' didn't find anything either. May 08 12:19:10 I probably need to ask jp30 whether php builds libphp5.so May 08 12:43:04 copperbeech: well it does actually. try cvs up -dP ; make php-dirclean ; make php-stage and look into staging/opt/lib May 08 12:45:07 he he - this could be interesting... May 08 12:45:42 it might take a while before I get back to you to confirm this :-) May 08 13:45:57 * NAiL just ordered another slug :) May 08 13:53:01 rwhitby-asleep: Getting that USB-hub you recommended together with a usb-lpt converter. Can't wait until the display is up and running ;) May 08 14:33:07 jacques: remember my problem of duplicate fonction definition. According to you, is having CPPFLAGS=-DHAVE_INET_ATON a bad workaround ? May 08 14:50:10 morning May 08 14:53:49 rwhitby: 'morning May 08 14:55:19 rwhitby: see my previous question ? is -DHAVE... is a good workaround when autoconf fails finding a existing function ? May 08 14:57:39 is there a config.cache variable you can use to override the detection? May 08 14:59:20 I can't find any config.cache May 08 15:01:37 have you looked through the configure script to see if there is any such variable? May 08 15:02:41 can you get the configure script to look in the right place when checking for inet_aton? May 08 15:02:58 if all that fails, then the -D is a good last resort. May 08 15:04:22 here is what I have in my configure.ac : AC_SEARCH_LIBS(inet_aton, resolv) May 08 15:04:47 when configure runs, I get :"checking for library containing inet_aton... none required" May 08 15:05:16 this end by creating an autoconf.h with "/* #undef HAVE_INET_ATON */" May 08 15:05:43 at build time, I get the following error : /opt/brcm/hndtools-mipsel-uclibc-3.2.3-full/include/arpa/inet.h:74: previous declaration of `inet_aton' May 08 15:08:59 ok, so you need to look at the definition of AC_SEARCH_LIBS and work out how to make it look in staging instead of the host libraries May 08 15:09:37 why in staging ? inet_aton is defined in libuclic.so.. May 08 15:09:45 libuclibc.so May 08 15:11:43 oh, ok, so you need to make it look in that in the toolchain dir instead of looking in libc May 08 15:11:50 (I guess) May 08 15:13:16 ok. I guess there's some parameter for that May 08 15:13:41 dunno May 08 15:13:51 if it's too hard, just use the -D May 08 15:14:48 ok. Wizard jp30 may have an idea on this :) May 08 15:25:56 'nite May 08 15:32:23 nite jf May 08 15:40:11 'nite, all you people in wierd timezones :P May 08 16:23:52 [g2]: SVN repo layout has changed May 08 16:24:12 <[g2]> rwhitby, I've been playing with that but haven't made it work yet May 08 16:24:18 <[g2]> It's not finding the BBFILES May 08 16:24:42 have you got the latest version of the repo with the top-level Makefile ? May 08 16:24:51 should be "checkout ; make" now May 08 16:25:23 or: May 08 16:26:18 cd /home ; svn co svn://svn.berlios.de/openslug ; cd /home/openslug ; make May 08 16:26:54 (you might want to symlink your existing downloads area into /home/openslug/downloads, and you can move an old tmp dir into /home/openslug/tmp too May 08 16:27:18 <[g2]> well I'm in a subpath we should change the make file to automatically update to the local `pwd` path May 08 16:27:35 patches welcome :-) May 08 16:27:54 <[g2]> you (or someone else) have done some nice work May 08 16:28:29 <[g2]> svn looks like it's shaping up a bunch May 08 16:29:47 <[g2]> so I think the procedure should be cd whereever; svn co svn:.. oepnslug; svn co svn... bitbake; cd openslug; make; May 08 16:30:26 the Makefile does bitbake for you May 08 16:30:35 <[g2]> Ah.. May 08 16:30:44 <[g2]> so I think the procedure should be cd whereever; svn co svn:.. oepnslug; make; May 08 16:30:52 <[g2]> with openslug spelled properly :) May 08 16:31:13 you'll need an additional cd openslug in there May 08 16:31:24 <[g2]> yeah that too :) May 08 16:57:27 rwhitby, [g2]: works fine with my build environments (unslung, openslug, openslug+uclibc, openslug+mm4 kernel) after I had changed the BBFILES and BBPATH variables, nothing else needed changing. May 08 17:01:39 The name change of nslu2-package-symlinks -> nslu2-linux breaks the --symlink=edit option to the freeze script in the wiki page, but that's easy to fix if/when necessary (and a simple symlink hacks round it) May 08 17:04:23 Ah, auto-logging to the mailing lists. Nice. May 08 17:17:01 jbowler, key working? May 08 17:23:40 <[g2]> is the flex thing worked out yet ? May 08 17:28:10 jbowler: which freeze script? May 08 17:30:57 ka6sox-away: yes, it's working fine. Started working when I deleted the two non-RSA id's I had put up there (there was a DSA and an identity.pub) May 08 17:31:44 rwhitby: see http://www.nslu2-linux.org/wiki/OpenEmbedded/MinimisingBitBakeMemoryUtilisation May 08 17:34:39 <[g2]> NOTE: package flex-native-2.5.31-r2: task do_populate_staging: started May 08 17:34:39 <[g2]> ERROR: function do_stage failed May 08 17:34:39 <[g2]> ERROR: see log in /home/tom/dev/berlios/openslug/tmp/work/flex-native-2.5.31-r2/temp/log.do_stage.13577 May 08 17:34:45 <[g2]> looks like flex is still broken May 08 17:35:34 [g2]: curious, I did a build from SVN last night and nothing failed. May 08 17:36:24 [g2]: yep, just four oe_libinstall lines in my flex-native-2.5.31-r2/temp/log.do_stage May 08 17:48:06 jbowler: do I need to svn move nslu2-linux to nslu2-package-symlinks? May 08 17:49:48 No: if we go with a different layout I'll change the script. It does 1001 heuristics anyway, so I can easily make it find the directory. Anyway, it's an 'expert' option, it just helps to ensure that the symlinks are maintained correctly ;-) May 08 17:49:58 <[g2]> Ok the dl was trunkcated from the mirror for flex so I copied a version from a sources tree May 08 17:50:14 <[g2]> then I get cat /home/tom/dev/berlios/openslug/tmp/work/flex-native-2.5.31-r2/temp/log.do_stage.11375 May 08 17:50:14 <[g2]> install: cannot stat `flex': No such file or directory May 08 17:51:48 from a clean build? I.e. did you rm -rf /home/tom/dev/berlios/openslug/tmp/*/flex-native-2.5.31-r2*? May 08 17:51:58 Hmm - it's Monday morning here, and bitkeeper is still not working. One of the downsides of seeing the sun first ... May 08 17:52:49 rwhitby: do you know for sure that someone in OE sent a support request to BitMover? (I.e. as it says in the 'the logging isn't working' message?) May 08 18:00:36 <[g2]> jbowler, yeah from a clean build. I did a bb -c clean flex May 08 18:03:46 [g2]: here's a weird thing, in my flex-native*/image directory I have a whole sub-tree from my root into staging/i686-linux/bin. May 08 18:04:08 [g2]: so the flex executable is in /home/work-tmp/jbowler/nslu2/openslug/work/flex-native-2.5.31-r2/image/home/work-tmp/jbowle May 08 18:04:08 r/nslu2/openslug/staging/i686-linux/bin May 08 18:04:51 jbowler: I've been assuming it affects any repo who has done more than 40 changesets since it broke. May 08 18:06:12 [g2]: I wonder if something on your system has a lower path-element limit than the corresponding thing on mine? (UNIX used to give up after stepping through too many directories.) May 08 18:06:52 03rwhitby * r18 10/Makefile: Another minor change to test CIA logging. May 08 18:07:01 <[g2]> I think its related to the oe_libinstall May 08 18:07:02 there we go - CIA logging from the SVN repo May 08 18:07:18 <[g2]> or the install May 08 18:07:30 <[g2]> there are only two lines in the do_stage May 08 18:08:14 rwhitby: I'd come to the conclusion (with very little evidence) that there was a specific break in the core OE repository, I wasn't able to trace the problem beyond the apparent break in the log -d May 08 18:08:53 <[g2]> Ok... it is the oe_libinstall that failing May 08 18:09:11 <[g2]> commenting out the second line in the do_stage still causes it to appear May 08 18:09:28 back later May 08 18:11:08 [g2]: oe_libinstall is in base.bbclass. The problem is that the staging is python, so it's particularly difficult to debug ;-) May 08 18:13:38 <[g2]> jbowler, actually bb -D flex-native prints most everything out May 08 18:13:45 [g2]: ok, that is 'populate_staging' I was thinking about. May 08 18:14:11 <[g2]> this is what's dying May 08 18:14:14 <[g2]> + do_stage May 08 18:14:14 <[g2]> + install -m 0755 flex /home/tom/dev/berlios/openslug/tmp/staging/i686-linux/bin May 08 18:14:14 <[g2]> install: cannot stat `flex': No such file or directory May 08 18:14:42 <[g2]> dunno why... but that's the what May 08 18:14:47 [g2]: try this trick, in the flex-native work directory cp temp/run.do_stage* dstage (you will probably have to select just one of the run files), then just execute './dstag' from the command line. May 08 18:15:30 [g2]: no, that won't help, it's the first line which fails. May 08 18:15:51 <[g2]> + BUILD_CPPFLAGS=-I/home/tom/dev/berlios/openslug/tmp/staging/i686-linux/include May 08 18:15:52 <[g2]> + cd /home/tom/dev/berlios/openslug/tmp/work/flex-native-2.5.31-r2/flex-2.5.31 May 08 18:15:52 <[g2]> + do_stage May 08 18:15:52 <[g2]> + install -m 0755 flex /home/tom/dev/berlios/openslug/tmp/staging/i686-linux/bin May 08 18:16:11 Either the do_stage[dirs] setting is wrong or the previous step failed silently. May 08 18:18:27 do_stage[dirs] is the default (unset), therefore it's in ${S} - i.e. work/flex-native*/flex-2.5.31. I have a flex in there, do you? May 08 18:19:15 <[g2]> checking May 08 18:19:55 flex is (should be) produced by one of the last real steps in do_compile. In temp/log.do_compile* there should be a gcc link line just before the line 'Making all in doc'. May 08 18:20:14 <[g2]> ls tmp/work/flex-native-2.5.31-r2/ May 08 18:20:14 <[g2]> flex-2.5.31 temp May 08 18:20:18 <[g2]> there ? May 08 18:20:43 <[g2]> you mean in flex-2.5.31 May 08 18:20:53 <[g2]> which is empty May 08 18:21:02 ;-) May 08 18:21:30 unpack failed. It should at least have the source in it... May 08 18:21:58 Your tar archive is probably 0 length May 08 18:22:25 <[g2]> well this is just 1 reason we should have signed md5sum and sha1's for all the dl's May 08 18:22:43 <[g2]> it would eliminate these kinds of errors May 08 18:22:59 yes, saves endless hours of working backwards to the very beginning. May 08 18:23:30 <[g2]> md5sum downloads/flex-2.5.31.tar.bz2 May 08 18:23:30 <[g2]> 363dcc4afc917dc51306eb9d3de0152f downloads/flex-2.5.31.tar.bz2 May 08 18:23:36 <[g2]> is that what you've got ? May 08 18:24:54 Yes, an exact match (and a matching .md5 in the same place.) Something else then. May 08 18:26:21 And the flex-native_2.5.31.bb isn't doing anything weird, indeed it's almost a minimal autotools implementation. May 08 18:29:31 ls -lrt flex-native-2.5.31/temp/log.do_* gives the log files in order of production. log.do_configure should make interesting reading (since there is nothing in the ${S} directory...) May 08 18:36:08 <[g2]> are you guys aware that oe is producing snapshots ? May 08 18:36:28 <[g2]> http://www.oesources.org/snapshots/openembedded/ ? May 08 18:36:52 <[g2]> I just noticed that something was being fetched from www.oesources.org/ May 08 18:40:26 3 days old. May 08 18:40:44 (since bk went down no snapshots) May 08 18:41:13 <[g2]> OK... here's the deal it's a Gentoo issue May 08 18:41:52 <[g2]> adding the following to setup-env seems to have fixed it May 08 18:41:56 <[g2]> unset LC_CTYPE May 08 18:41:56 <[g2]> unset CC May 08 18:41:56 <[g2]> unset CXX May 08 18:42:09 <[g2]> I'm past the flex issue May 08 18:44:11 [g2]: tricky, I've seen that elsewhere too. I use gentoo as well, but I don't have any of those set. It would be good if bash had a 'zap the environment' setting - I used to use a wrapper which would execve(2) with no environment... May 08 18:44:47 <[g2]> jbowler, nod. something like taht would be nice May 08 18:45:22 <[g2]> I've just had those in my config files forever.... since the uClibc buildroot days May 08 18:45:48 <[g2]> any debian ppl out there ? May 08 18:47:39 jacques and i are debian types. May 08 18:47:46 but he is more qualified. May 08 18:48:19 <[g2]> Ok how do we create a .deb to pull the svn for openslug and bitbake and build openslug from a knoppix 3.8 release ? May 08 18:49:12 <[g2]> I could probably work throught creating a Knoppix remaster with all the stuff on there May 08 18:50:49 are you installing the knoppix release on the HD or LiveCD? May 08 18:51:41 <[g2]> I've got it on may laptop installed to the hd. but I built from the livecd a month ago which my HD had problems and I created a backup May 08 18:52:03 <[g2]> everything build (except for cpio which needed a special file) May 08 18:52:15 <[g2]> s/build/built/ May 08 18:53:10 okay so flex/bison and all the other stuff is there? May 08 18:53:41 <[g2]> iirc everything was on the 3.7 liveCD May 08 18:53:50 <[g2]> I'd imagine it's all there on 3.8 May 08 18:54:02 <[g2]> re... that one cpio thing May 08 18:54:36 <[g2]> so there some setup the apt stuf then one apt-get openslug May 08 18:54:48 <[g2]> then you're ready to rock May 08 18:54:54 I see... May 08 18:55:03 <[g2]> or just have a master that's got it all on it May 08 18:55:25 <[g2]> I think the ibiblio would host an embedded distro like that May 08 18:56:18 my friend cbpage has done some knoppix stuffs...I don't know if he is looking here now. May 08 18:56:35 might have even done some remastering. May 08 19:58:27 03bzhou * 10unslung/make/ (libidn.mk libpth.mk): added libidn and libpth May 08 22:25:32 re those gentoo env vars (LC_CTYPE, CC, CXX), none of my up-to-date gentoo boxes have them anymore, I think they fixed it May 08 22:27:21 03jeanfabrice * 10unslung/make/eggdrop.mk: Bumped IPK version May 08 22:30:59 jbowler, have you looked at Redboot at all? May 08 22:31:40 jacques: LC_CTYPE is a bash setting, the other two are specific to the C compiler (and make), it seemed that [g2] had set them because of use of a non-standard C compiler. I believe those were the problem area. May 08 22:32:15 larry_sox: only from the outside (i.e. I haven't checked the code) May 08 22:33:07 I have several slugs that are awaiting a way to boot them with more memory..but none have been booted sucessfully with redboot. May 08 22:33:13 (with more memory) May 08 22:34:32 There was some discussion recently about determining memory on linux-arm-kernel. There was always some problem (well, ok, 20 years ago there was a problem ;-) May 08 22:35:35 The linux-arm-kernel issue sounded similar. As I remember the original problem simply probing memory was liable to result in a hung system, but that was determining memory with ARM2+MEMC1a on an A680... May 08 22:36:15 k May 08 22:36:33 there are only so many configs that an Xscale can have. May 08 22:36:39 Intel have RedBoot patches, or at least a downloadable copy, on developer.intel.com. It may be that it hardwires the memory, but I'm just speculating. May 08 22:37:29 I've tried to look at the HAL code but its pretty convoluted. May 08 22:38:41 the APEX bootloader might be what we have to use with Openslug. May 08 22:39:01 hi. I was using upslug with great success, and now it has stopped detecting my slug even though I can arping it. what should I be looking to fix? May 08 22:41:02 jbowler/jacques: CC/CXX/LC_CTYPE is not set on any of my gentoo boxen, but there was a time when they used to cause problems... May 08 22:41:46 NAiL, agreed - jbowler, gentoo used to set at least CC and CCX in every shell all the time - it caused a big problem with uclibc buildroot May 08 22:42:18 happ: what version are you using and what are you trying to upslug? May 08 22:43:14 happ, is the slug going into upgrade mode? did you change anything in your network topology? May 08 22:44:25 larry_sox: upslug cvs head (cvs rev upslug.c 1.13), upslugging images I'm building, have been successfully upslugging many times for the past 24 hours May 08 22:45:05 jacques: unsure, how do I check if it's in upgrade mode? my topology is the same: a crossover directly from my laptop to the nslu2 May 08 22:45:39 ... I have Redboot fine in my serial console, so I suppose I could just load in my image from there May 08 22:46:05 I believe the top LED alternates red/green when slug is in upgrade mode May 08 22:46:30 do you recall that from the times it worked? May 08 22:46:59 happ: enter 'upgrade' to the RedBoot prompt, the power led(s) will flash red/green. upslug will then detect it. May 08 22:47:04 happ, you have serial? well that will make debugging somewhat easier :-) May 08 22:47:16 he does May 08 22:48:21 jbowler: thanks, that worked. May 08 22:48:29 what should I look out for in the future to prevent that from happening? May 08 22:49:11 happ: can you update the wiki with what worked here for you? (in the upslug tool docs) May 08 22:49:14 happ: there was no problem. upslug can only detect the slug when in upgrade mode. I believe it talks to a fixed MAC address which RedBoot only sets in that mode. May 08 22:49:58 okay. also I've been using slugimage, and I noted that the cvs head verison needs some "binmode FILE;" lines after the opens for it to work. where should I submit that? May 08 22:50:14 happ: my normal approach is to telnet in to RedBoot then type 'upgrade'. It's also possible to use the power button somehow, but I May 08 22:50:21 've never got that to work... May 08 22:55:33 happ, you using slugimage on windows? May 08 22:55:57 because I don't think those are needed in unix (which is why we never noticed that bug I guess) May 08 22:56:25 strangely enough, no, I was in linux May 08 22:56:35 then I am really confused :-\ May 08 22:56:49 me too, but the binmode FILE lines don't hurt.. May 08 22:57:06 I would submit a slugbug May 08 22:57:10 ~slugnug May 08 22:57:14 ~slugbug May 08 22:57:18 it has been said that slugbug is http://slugbug.nslu2-linux.org/ May 08 22:59:31 I've updated the upslug tool wiki with the RedBoot upgrade info, thx May 08 23:07:23 happ: thanks. May 08 23:08:06 ka6sox: HAL code: is that the code in RedBoot, or the kernel? May 08 23:08:25 redboot. May 08 23:08:44 (which I saw you hadn't looked at) May 08 23:09:18 Um, so it is, presumably, attempting some sort of read. I believe the linux-arm-kernel messages amounted to 'the boot loader tells you this' ;-) May 08 23:09:38 I think so. May 08 23:10:04 we have tried setting the mem=64M kernel arg and it doesn't work. May 08 23:10:26 But it passes it in in some kind of structure ('tag'?), so a hack is probably possible, if you want to compile it into the kernel ;-) May 08 23:11:20 I can't remember how it was done...jacques sent it to me a long time ago. May 08 23:12:02 a kernel with mem=64M kernel cmdline compiled in? I vaguely recall that :-) May 08 23:12:44 ya...I just came across it in my email tonight as I was cleaning up. May 08 23:12:47 That's easy, simply change the line appended to the end of 'defconfig' by openslug-kernel*.bb May 08 23:15:21 I think this was in the bad old days before ka6sox had the hardware to build his own oe-based images May 08 23:16:40 :-) May 08 23:16:44 ... sorry, somehow I missed the the 'we have tried' line May 08 23:17:57 thats why I was interested in the Redboot HAL as it configures the Memory Register. May 08 23:18:21 You probably need to specify the start too, i.e. mem=64M@ May 08 23:19:02 Doofus code doesn't match the comments. Maybe mem=64M, May 08 23:19:42 jbowler, (possibly stupid question) would the 64M start at same place as 32MB ? May 08 23:19:56 Snark. mem=,64M. Well, try trial and error, you only have to reflash 4 times. May 08 23:20:15 jacques: depends on how ka6sox soldered the address lines ;-) May 08 23:20:19 aha May 08 23:20:51 also, what about memory organization? would changes have to be made to account for # of banks amd bank size? May 08 23:21:00 the address lines are stock,meaning that they follow the redboot info. *but* since I have 4 chips instead of 2. May 08 23:21:28 2 banks of 8M16 vs 1 bank of 8M16(stock) May 08 23:22:10 and the bank switching uses the CSEL for the other bank. May 08 23:23:07 I'm not reading this code right. Too late at night. initrd uses ",", mem uses "@" the mem case accepts (should accept) both mem=64M and mem=64M@start. May 08 23:24:03 will the kernel know how to set the memory config register? May 08 23:25:47 I'm not sure. The code accepts multiple physical memory 'banks'. I.e. it should accept multiple mem=length@start options. The config is presumably xscale specific (I'm looking at arch/arm/kernel/setup.c) May 08 23:26:21 When you tried mem=64M did it ignore it or did it crash? May 08 23:26:51 crash May 08 23:26:56 ooooh May 08 23:27:11 Why yes, I did. May 08 23:27:14 accepts multiple mem= - that's hot May 08 23:27:26 evening dyoung :-) May 08 23:27:29 The 'start' was probably wrong. It's hard-coded to PHYS_OFFSET May 08 23:28:31 dyoung: did you try mem=64M too? May 08 23:29:32 was just reading the logs. May 08 23:29:36 jacques akes a very good point. May 08 23:30:20 i do? in this case it was probably an accident :-) May 08 23:30:43 basically the HAL code was hacked to address the number of banks and the size of the banks appropriate to the hardware that ka6sox was testing. May 08 23:31:08 HAL in RedBoot, or the kernel? May 08 23:31:09 However, I dont think I carefully examined the placement of the memory. May 08 23:31:13 HAL in Redboot. May 08 23:31:27 yeah dyoung was making custom redboot binaries May 08 23:32:39 According the dsaxena, they dont fool around with the memory initilzation; and they let the bootloader handle it. May 08 23:33:11 jacques, the reason I say that was a good point is IIRC, the CS lines are backwards on the nslu2 to conventional thinking. May 08 23:34:31 How so? Well, I guess, what is conventional? This is just chip select right... May 08 23:37:12 it is but the sdconfig register has different modes. May 08 23:37:34 for example: 2X16M16 vs 4x8M16 May 08 23:37:35 DSM_CS_N1 is the one being used and SDM_CS_N0 is free. May 08 23:38:05 same amount of memory but different register config. May 08 23:41:28 Yessss... um. So the last ARM memory setup I was familiar with was MEMC1a... It doesn't matter because there are only two possibilities. May 08 23:41:55 Either the start is wrong, which must mean that the first physical address is >0 (since the XScale default is 0) May 08 23:42:02 okay so would the memory map change (start addy) if we were at 4X8M16? May 08 23:42:24 Or the chips end up in two (or more) banks and physical memory is discontiguous. May 08 23:43:20 * ka6sox goes to look for the manual. May 08 23:43:34 the May 08 23:43:50 Xscale only supports 2 banks (well...the slug only does) May 08 23:44:04 n0 and n1 May 08 23:44:22 That's most likely (i.e. changing the number of chips splits the result). This is a sw engineer viewpoint BTW - I'm thinking some of the address lines get sent to the chip and some go to the chip select. May 08 23:45:04 I gave up trying to make it all work until I actually have the hardware in hand. May 08 23:45:22 trying to remotely test/debug quickly became a PITA. May 08 23:45:23 iirc the only time memory moves is in the 256MB mode. May 08 23:45:34 then there is a shadow issue. May 08 23:47:33 so now I'm trying to decide if we should go with 4X8M16 or 2X16M16 config. May 08 23:48:05 (if we can make the 4X8M16 mode go then the 16M16 mode is easy.) May 08 23:49:26 03rwhitby * r19 10/.svnignore: Added setup-env May 08 23:52:28 03rwhitby * r20 10/ (7 files in 3 dirs): Moved local.conf to local.conf.template, removed setup-env, updated Makefile to generate local.conf and setup-env, told svn to ignore the generated files. May 08 23:53:13 ok, now you can just "svn co svn://svn.berlios.de/openslug ; cd openslug ; make", no matter where you are. May 08 23:56:41 rwhitby-away, that's beee-yoo-tiful May 08 23:58:36 Booty-full? Arr. May 08 23:58:45 lo May 08 23:58:47 l May 08 23:59:54 so which way of soldering will yield the best config for us to test and troubleshoot? **** ENDING LOGGING AT Sun May 08 23:59:56 2005