**** BEGIN LOGGING AT Thu Apr 28 23:59:57 2005 Apr 29 00:31:29 morning all Apr 29 01:06:54 Styno (from Chupa forum) talks to me about http://www.teamasa.com/npwr_lc.shtml Apr 29 01:07:14 did anyone ever take a look at it ? Apr 29 02:07:25 hiu jf-work Apr 29 02:07:38 I've looked at it before. Apr 29 02:07:46 iirc, Lennert has one. Apr 29 02:07:55 hes the one that pointed that out to use earlier. Apr 29 02:08:20 hey Apr 29 02:08:48 Morning. Apr 29 02:09:49 morning Apr 29 02:10:31 If youre thinkinng about it though, be prepared to "drop some coin". I had sticker shock. Apr 29 02:10:43 saw it :( Apr 29 02:11:03 dyoung: nslu2-log isn't in all the channels he should be ... Apr 29 02:11:15 ~lart nslu2-log Apr 29 02:11:16 * jbot takes a rusty axe and swings it violently, taking nslu2-log's head off **** ENDING LOGGING AT Fri Apr 29 02:13:20 2005 **** BEGIN LOGGING AT Fri Apr 29 02:14:05 2005 **** ENDING LOGGING AT Fri Apr 29 02:18:23 2005 **** BEGIN LOGGING AT Fri Apr 29 02:20:06 2005 **** ENDING LOGGING AT Fri Apr 29 02:26:29 2005 **** BEGIN LOGGING AT Fri Apr 29 02:26:42 2005 Apr 29 02:26:57 Okay repaired. Apr 29 02:27:43 hopefully once and for all. Apr 29 03:00:03 dyoung: thx Apr 29 03:02:21 Added bonus for folk too Apr 29 03:02:30 the spca5xx driver works under openslug now Apr 29 03:02:41 I suppose I should push it. Apr 29 03:05:20 cool Apr 29 03:06:08 jp30-work: how do I get your way-cool lean, mean apache going? Apr 29 03:06:39 with the thread-safe php to go with it .... Apr 29 03:08:02 03dyoung 07 * r1.3180.53.1 10openembedded/ (3 files in 2 dirs): Apr 29 03:08:03 Deleted spca5xx_20050319.bb and spca5xx_20050301.bb because they were broken Apr 29 03:08:03 for big endian systems added 20050419 which has basic endian fixes. Apr 29 03:08:59 VoodooZ around/ Apr 29 03:26:31 jf-work: ping Apr 29 03:27:01 pong Apr 29 03:27:12 have you used thttpd much? Apr 29 03:28:03 It seems that when I stream video from php-apache running on Unslung to the Neuston MC-500, it works fine. But when I stream from php-thttpd, it doesn't display. Apr 29 03:28:04 I'm using it a bit. Tried dokuwiki last night. slow... Apr 29 03:28:27 Before I go into deep debugging mode, I was wondering if you'd ever experienced anything weird. Apr 29 03:28:55 not a real stream (I mean, no streaming engine) ? Apr 29 03:29:28 You just have a video on your asus and start read it from somewhere else, taht's right ? Apr 29 03:29:46 sorry, on your slug Apr 29 03:29:51 yeah Apr 29 03:30:31 Do you think I could reproduce the error trying to stream a wmv over http ? Apr 29 03:31:01 actually, I think I might have just found the reason Apr 29 03:31:16 apache returns: "media/movie/Pingu-15.avi|0|0|192.168.1.5:8000/media/movie/Pingu-15.avi|" to the media player Apr 29 03:31:27 and thttpd returns: "media/movie/Pingu-15.avi|0|0|:8080/media/movie/Pingu-15.avi|" Apr 29 03:31:44 Note the missing IP address from the thttpd response. Apr 29 03:31:51 did you try setting the bind host in thttpd.conf ? Apr 29 03:31:52 which means it's something in PHP Apr 29 03:32:01 by default it binds all netcard Apr 29 03:32:05 all ip Apr 29 03:34:53 maybe php calls for such $_SERVER['server_name'] or $_SERVER['server_add'] var and this var cannot be resolved (used to see this recently) Apr 29 03:35:22 I'll look in the code to see what it is doing. Apr 29 03:35:40 how do I set the bind host in thttpd.conf? Apr 29 03:35:53 and maybe in page Apr 29 03:36:34 keyword 'host' Apr 29 03:38:40 what should I look for in the phpinfo page? Apr 29 03:39:12 maybe $_SERVER var Apr 29 03:39:26 thttpd.conf : host=xx.xx.xx.xx Apr 29 03:39:39 it should then bind to this defined IP Apr 29 03:40:10 I think that settings in the thttpd.conf should solve the problem Apr 29 03:41:20 yeah, SERVER_NAME and SERVER_ADDR are present in the apache phpinfo, but not the thttpd phpinfo Apr 29 03:42:23 try setting host in thttpd.conf to see if it solves. If it doesn't, maybe $_SERVER['HTTP_HOST'] could be used instead of SERVER_NAME/ADDR Apr 29 03:43:08 yeah, setting host didn't change it. a lack of SERVER_NAME and SERVER_ADDR is going to affect a lot of PHP scripts running on thttpd .... Apr 29 03:43:26 lets have a look at the thttpd source .... Apr 29 03:54:02 seemes that thttpd calls for a gesthostname to define SERVERn8AME Apr 29 03:54:05 SEREVR_NAME Apr 29 03:54:18 wiley doesn't have it in uclibc Apr 29 03:54:56 but they says that host keyword in thttpd.conf take over Apr 29 03:55:15 see config.h Apr 29 04:29:14 jp30-work: ping Apr 29 04:29:44 building php-apache on my local machine found db-4.3 on the host instead of db-4.2 in staging. Apr 29 04:30:08 tracking it down points to apr-util Apr 29 04:31:35 in particular, running "./staging/opt/bin/apu-config --libs" prints a bad set of libs, based on what is on the host, not what has been cross compiled and staged. Apr 29 04:36:44 ok, looks like it was my local problem - autoconf not being run for apr-util Apr 29 04:39:01 jp30-work: false alarm. my bad. Apr 29 04:42:34 jf-lunch: yeah, the php code uses SERVER_ADDR Apr 29 04:47:42 xml2-config is being staged to staging/bin, but the directory is not being created first, so it is ending up as a file called "bin" :-( Apr 29 04:49:34 Hmm - for some reason libxml2.mk is moving it from staging/opt/bin to staging/bin. it shouldn't do that .... Apr 29 04:51:19 looks like it might have pre-dated the move to staging/opt/.. .... Apr 29 04:54:57 and php-apache is perpetuating the error by picking it up from staging/bin ... Apr 29 05:00:44 Hmm - CIA-8 is not reporting CVS checkins ... Apr 29 05:04:36 dyoung: is the RV042 yours? Apr 29 05:06:14 Thats [g2]'s latest toy. Apr 29 05:06:25 Its pretty cost effective. Apr 29 05:06:37 I might have to buy one. ;-) Apr 29 05:06:57 Solder on the MiniPCI connector.... Apr 29 05:07:26 <[g2]> dyoung, what would you do with the MiniPCI ? Apr 29 05:07:28 USD$157? Apr 29 05:07:34 <[g2]> nod Apr 29 05:07:38 <[g2]> newegg Apr 29 05:07:45 two slugs Apr 29 05:07:46 Give it USB. Apr 29 05:07:53 <[g2]> how Apr 29 05:08:14 Through one of those cosmotronic or whatever it was called adapters to normal pci Apr 29 05:08:26 <[g2]> ok Apr 29 05:08:31 <[g2]> that makes some sense Apr 29 05:08:39 you'd still have to use intel drivers to use the wan port? Apr 29 05:08:53 ports Apr 29 05:09:29 what's the differentiating feature set ? Apr 29 05:09:31 * [g2] plans on building hw (probably ixp465 based) Apr 29 05:09:49 ixp425 has crypto. Apr 29 05:10:04 ok, so it's the cheapest way to get a 425? Apr 29 05:10:14 pretty much. Apr 29 05:10:16 <[g2]> these are just much cheaper (and more avialable) than the Intel development platform which go for $3K+ Apr 29 05:10:31 ok, that makes sense. Apr 29 05:10:59 if our slugs could run well, less sluggishly at the full rated speed, that would make me happier. Apr 29 05:11:25 [g2]: does it stand vertically? Apr 29 05:11:30 hahahahahaha Apr 29 05:11:35 I was waiting for you to ask that! Apr 29 05:11:54 <[g2]> mechanicals haven't been started yet, but I'll make sure that's an option Apr 29 05:12:06 [g2]: no, I meant the stock RV042 Apr 29 05:12:24 OpenRevy Apr 29 05:12:34 OpenRevvy Apr 29 05:12:42 <[g2]> OpenRV Apr 29 05:12:48 Or is that FreeRV? Apr 29 05:13:01 <[g2]> it'd have to be OpenRV Apr 29 05:13:17 [g2]: will you need to use the intel drivers to access the WAN ports? Apr 29 05:13:28 Okay, so since youre both here. Apr 29 05:13:48 have we completely given up on ever having a binary release? Apr 29 05:13:50 <[g2]> rwhitby, sure Apr 29 05:13:54 FreeRV would be ethernet-less RV then? Apr 29 05:14:05 or should this click-through stuff be pursued? Apr 29 05:14:42 ByronT: you have a letter from Intel that allows you (and/or your company) to do a click-through, right? Apr 29 05:15:05 it is do-able Apr 29 05:15:20 what if we have a "holding company" (i.e. ByronT) that does the binary release ... Apr 29 05:15:36 (mind you it doesnt really matter to me either way because I can always just build it; but I know there are others who may be interested in seeing it to that stage) Apr 29 05:15:53 <[g2]> hey it's open source, just about anyone can do a release Apr 29 05:16:00 dyoung: yeah, that's true for most people in here Apr 29 05:16:47 [g2]: the point is that ByronT has explicit permission from Intel (or at least it's representative) to do so. Apr 29 05:17:19 <[g2]> I applaud ByronT for going the distance, investigating that and seeking resolution Apr 29 05:17:27 <[g2]> proper resolution Apr 29 05:17:45 whereas nslu2-linux doesn't, and may not want to, based on a number of it's members not wanting to go down that path Apr 29 05:18:07 (have explicit permission, that is) Apr 29 05:19:09 so nslu2-linux continues to develop OpenSlug, but never does a binary release. But someone else just happens to make a proper click-through binary release area which is unrelated to nslu2-linux.org Apr 29 05:19:34 ~emulate ka6sox Apr 29 05:19:42 <[g2]> I some point, I'll either a) make hw and release it with the hw, b) deal with Intel on the licensing issue for a forsale product, c) Intel will bless the release and we can just release it Apr 29 05:19:43 that sounds workable to me Apr 29 05:19:44 "Good Plan" Apr 29 05:20:31 well, I don't know if I told you all about the last telephone conversation with Intel Apr 29 05:20:31 could we get RR to host the binary openslug release on his ixp4xx sf site? Apr 29 05:20:35 Okay so that basically answers the basis of my question. Apr 29 05:21:03 (dyoung basis: "So do I need to find a new toy to play with?") Apr 29 05:21:06 <[g2]> I gave RR a chiding in another ml about the issue Apr 29 05:21:16 but I was told that a BSD style license release of the ixp4xx was "immenent" Apr 29 05:21:28 I might have to buy a [g2]-thingee to play with. Apr 29 05:22:05 dyoung: how does the openslug binary release issue affect the decision to buy a Revvy ? Apr 29 05:22:16 <[g2]> IMHO Intel is just plain conflicted and confused on the issue Apr 29 05:23:09 It doesnt affect my desire for more toys. But if the ixp is going to be in the middle of license conflict.... Apr 29 05:23:11 maybe Intel's hiring of the Sun open source advocate will help to smooth over the problems... Apr 29 05:23:56 rwhitby: (dumb question) Did you restart php-thttpd after updating thttpd.conf with 'host' entry ? Apr 29 05:24:09 jf-work: yeah, not a dumb question, but I did. Apr 29 05:24:09 <[g2]> Intel is planning to move in a good direction Apr 29 05:24:14 dyoung is "toy-agnostic"... :) as long as the toy is fun.... Apr 29 05:24:22 <[g2]> It not the direction, it's more the timing Apr 29 05:24:24 exactly. Apr 29 05:24:31 <[g2]> We were ready 2+ months ago Apr 29 05:25:20 <[g2]> it's probably at least months maybe more than a year before Intel will get it's strategy together and dot the i's and cross the t's Apr 29 05:25:30 the use of a "release agent" is a good idea/plan. Apr 29 05:25:42 We just need a volunteer. Apr 29 05:25:45 <[g2]> I'd *LOVE* to be totally wrong, but I'm somewhat a realist on this Apr 29 05:25:59 I can do the click-through Apr 29 05:26:01 [g2]: fully agree Apr 29 05:26:10 on my toddcomp.com site Apr 29 05:26:14 ~praise ByronT Apr 29 05:26:15 <[g2]> I've toyed a little with asking legal at the EFF to look into it or PJ Apr 29 05:26:17 All hail ByronT! Apr 29 05:27:02 <[g2]> ByronT, feel welcome to do the click-through and release a binary Apr 29 05:27:12 I wonder if ka6sox can point openslug.org to a subdomain on my website Apr 29 05:27:29 <[g2]> However, you are taking responsibilty for that by doing so Apr 29 05:27:51 <[g2]> INAL Apr 29 05:28:06 that's right - ka6sox bought openslug.org Apr 29 05:28:24 he could transfer the domain to you if he's not comfortable ... Apr 29 05:28:32 actually, I bought openslug.org... and pointed it to ka6sox's DNS server Apr 29 05:28:44 hah - even better! Apr 29 05:29:05 it's not used for anything else, so you just point it back to your servers .... Apr 29 05:30:57 <- have to run into work... be back later Apr 29 05:31:14 ByronT: if you want to do this, go for it. Apr 29 05:31:37 .... ByronT takes over as OpenSlug release manager ...... Apr 29 05:32:23 <[g2]> Go ByronT, Go ByronT Apr 29 05:32:35 <[g2]> Your the Man! You Apr 29 05:32:35 I'm sure my wife will be happy about this... :) Apr 29 05:32:41 <[g2]> 're the MAN Apr 29 05:33:02 Thanks byront Apr 29 05:33:11 Sorry to have beat the thing to death. Apr 29 05:33:19 * [g2] former OpenSlug release manager Apr 29 05:33:31 * [g2] does the happy dance Apr 29 05:33:52 do you have to do the happy dance whilst standing on me? Apr 29 05:34:13 <[g2]> So ByronT when are you gonna release OpenSlug ? Apr 29 05:34:13 and the best thing is that ByronT already has a cloak .... Apr 29 05:34:32 jf-work: did you get your cloak? Apr 29 05:34:36 "when I have time....." ;) Apr 29 05:35:13 * dyoung hands ByronT a "Round TUIT" Apr 29 05:35:24 ByronT: you got unslung.org too ? Apr 29 05:35:43 yeah, I think I have that one as well Apr 29 05:35:54 let's do the click-through on it too - just in case Apr 29 05:36:03 and then we can track download numbers too ... Apr 29 05:38:16 [g2]: so what does this turn of events due to FreeSlug? Apr 29 05:39:20 <[g2]> rwhitby, there were 32 ppl responding to the poll only 2 so far think it's a great idea Apr 29 05:39:30 I didnt vote. Apr 29 05:39:48 <[g2]> that's exactly why I did the email and the poll, to find out what the community wants Apr 29 05:40:41 32 out of 3300 is not that representative ... Apr 29 05:40:47 I was debating about voting "[g2] is a wimp"; but the per-vote result isnt public. ;-) Apr 29 05:40:59 <[g2]> that was my first vote! Apr 29 05:41:04 heh Apr 29 05:41:08 yeah, looses all it's stir value ... Apr 29 05:41:12 <[g2]> I was wondering who the other two were Apr 29 05:41:18 other 3 now. heh. Apr 29 05:42:03 (No, I still didnt vote) Apr 29 05:42:13 other 4 ... Apr 29 05:42:24 <[g2]> 220, 221 what ever it takes Apr 29 05:42:47 <[g2]> that was probably the new release mgr Apr 29 05:42:49 <[g2]> :) Apr 29 05:55:12 night all Apr 29 05:55:30 night dyoung Apr 29 05:55:38 <[g2]> nite dyoung Apr 29 06:03:46 jf-work: looking at libhttpd.c in thttpd, I can't see how hostname gets set on line 261. Apr 29 06:07:45 dunno. I want to try with a shell CGI to see what env var are defined Apr 29 06:11:19 Looks like the PHP code in OpenShowCenter uses SERVER_ADDR, which is not a defined env var for CGI/1.1, so I can't blame thttpd for not defining it (even though Apache does). They both define HTTP_HOST (also not a defined env var for CGI/1.1) but one defines it to 192.168.1.5, and the other defines it to 192.168.1.5:8080, so I can't just patch the OpenShowCenter code to use that. Apr 29 06:11:19 And looking at the code in libhttpd.c in thttpd, it looks like SERVER_ADDR can only be set to the result of gethostname(), and nothing else affects it. Apr 29 06:11:19 actually, I take that back. the hostname can be passed into that routine. Apr 29 06:11:19 * rwhitby keeps spelunking Apr 29 06:13:25 we still have a pb with server_name which is part of the CGI1.1 spec Apr 29 06:13:55 right. that's what the code should use, and it's not getting set for me Apr 29 06:15:59 hummm... got SERVER_NAME defined when runnning a shell cgi Apr 29 06:17:27 hmm - so it could be a php/thttpd thing? Apr 29 06:18:27 really strange. Apr 29 06:19:03 aha - look in php_thttpd.c, line 295 Apr 29 06:19:16 that's where it should be done, but isn't. Apr 29 06:19:17 #!/bin/shecho "Content-type: text/plain"set | sort Apr 29 06:22:21 btw, php_thttpd.c is a close copy of sapi/thttpd/thttpd.c Apr 29 06:24:52 where is ..../sapi ? Apr 29 06:25:29 ah, php-apache Apr 29 06:25:33 php source Apr 29 06:25:53 it doesn't have it either. Apr 29 06:26:52 so I should patch it instead? Apr 29 06:27:30 I think yes, that was my idea giving you the original path ;) Apr 29 06:28:25 during make install-sapi of php, thttpd.c is copied from sapi/thttpd to /php_thttpd.c Apr 29 06:31:36 rwhitby: Re /initrd busy in 4.x - 3.x has /dev in ramfs, while 4.x has /dev as part of /initrd. Apr 29 06:32:09 aha. Can we fix that? Apr 29 06:32:38 Perhaps by moving /dev back into ramfs. Apr 29 06:33:14 I don't know how/why /dev was changed. Apr 29 06:34:40 3.x had it as ramfs so we wouldnt' be writing to jffs2 for /dev writes Apr 29 06:35:51 lsof on 4.x shows /initrd/dev/console and /dev/console open with different inodes before and after pivot. Apr 29 06:36:16 3.x shows /dev/console as the same inode before and after pivot. Apr 29 06:36:38 3.x doesn't pivot Apr 29 06:36:43 it uses real-root-dev Apr 29 06:38:04 After root change then. Apr 29 06:41:04 jf-work: ok, got SERVER_NAME working for thttpd Apr 29 06:43:40 ok, now I have both php-apache and php-thttpd serving to the Neuston :-) Apr 29 06:43:45 amazing.... I'm still trying to understand the source :( Apr 29 06:43:56 on where to get the var Apr 29 06:44:07 TG ? SG ?... Apr 29 06:44:38 patch is checked in Apr 29 06:44:47 * jf-work needs intensive C lessons Apr 29 06:45:01 no more CIA as I see.. Apr 29 06:46:30 ok, now to make SERVER_ADDR work ... Apr 29 06:46:58 rwhitby: Perhaps leave /dev in ram for initrd and copy to disk when unslinging. Ram loss for orphaned /dev would be less than for orphaned initrd with its /dev. Apr 29 06:47:17 rwhitby: how did you wrote the pacth so quickly ?!! where is the TG struct ? Apr 29 06:47:44 jf-work: I just copied some other code around that spot, looking for patterns Apr 29 06:48:01 and looking in the libthttpd.h file Apr 29 06:48:05 he has a mind to computer interface that allows him to spot problems Apr 29 06:48:36 glc: how does the /dev in ram get initialised? Apr 29 06:49:15 jf-work: for SERVER_ADDR I'm going to need to actually write some new code I think :-) Apr 29 06:49:58 What ? it's not allready done ?! Apr 29 06:50:01 I don't know how it was done in 3.x. I would have to re-flash and go into maint mode to check. Apr 29 06:50:28 glc: in 3.x, we bind mounted it and copied the contents Apr 29 06:55:10 Ever think of booting right into jffs2? Shure would make testing changes easier. Can always re-flash if careless. Apr 29 06:59:40 I understand fallback to maint mode, but who actually does that? Most common jffs2 trash is probably overfill, and that requires a reflash anyway. Apr 29 07:04:41 That would probably be more appropriate for openslug. Apr 29 07:07:04 glc: unfortunately, it's too late here for me to think about switchbox changes. But I do appreciate the analysis you are doing. can you put it all together in an email (including the full proposal for what a replacement switchbox should do) ? Apr 29 07:07:58 I used fallback to recovery mode (is that what you meant) a number of times early in switchbox development, but I doubt any user of 4.x will ever use it. Apr 29 07:09:13 glc: we could (and can) boot directly into jffs2 - it just needs a change to the kernel args. We'd still need to waste the erase block where switchbox is stored (cause redboot copies it) Apr 29 07:09:51 but to make a major change like that would need some non-11:30pm thought .... Apr 29 07:16:53 SERVER_ADDR is specific to apache; sounds strange some php scripts depends on it Apr 29 07:17:14 Yes, I know. Just expressing frustration at how hard it is to do any testing on the slug. Apr 29 07:17:43 jf-work: seems that the neuston doesn't do dns lookups on that string, so it has to be passed an address. Apr 29 07:18:17 hence the usage of SERVER_ADDR (which apache guarantees is always an IP address, whereas SERVER_NAME changes based on vhost name in the request) Apr 29 07:18:21 The only way to make changes to test is to edit on the pc, and recompile the whole thing. takes days to do 10 minutes of troubleshooting. Apr 29 07:19:29 I'll leave you two to your work. I just wanted to let you know that I thought I uncovered the reason for the unable to umount problem. Apr 29 07:19:39 shouldn't openshowcenter be patched to dns lookup the address ? Apr 29 07:19:42 Just too hard to test fixes. Apr 29 07:20:00 glc: that's great information. thanks. Apr 29 07:20:27 glc: we should do something about it - I just don't have the brain power to think about the totality of it right nw Apr 29 07:22:19 I wonder if we could tell redboot to load up the sysconf partition instead of the initrd partition (to save wasting an erase block) and then get the kernel to boot directly into jffs2, where we do all this stuff in the /linuxrc there .... Apr 29 07:22:43 make mtd1 and mtd2 point to the sysconf block Apr 29 07:23:02 ~nefarious Apr 29 07:23:05 nefarious is probably to be infamous by way of nastyness Apr 29 07:23:26 pity dyoung has gone to sleep Apr 29 07:24:12 glc: if we can recoup that block, then I would be all for getting the kernel to boot directly to jffs2 Apr 29 07:24:41 so is it the case that we are 1MB or so short on RAM in 4.x vs 3.x ? Apr 29 07:24:48 (size of initrd) ? Apr 29 07:24:56 Sure would make testing fixes easier too. Apr 29 07:25:07 glc: agreed Apr 29 07:25:40 and would also allow power users to really change the boot behaviour, and all the modules you need would be in the right spot ... Apr 29 07:27:00 It would give us more ram, possibly more jffs2 space, and easier testing. Apr 29 07:27:25 gotta be happy with that. Do you want to try it out? Apr 29 07:27:53 I won't have much time over this weekend (Elise's christening, and all the family is together for the first time in six years) Apr 29 07:28:39 I would be happy to be one of the first alpha testers. I have difficulty developing with this bk stuff. Apr 29 07:29:37 jf-work: looks like supporting SERVER_ADDR would require adding fields to the httpd_server struct in libhttpd.h to store the httpd_sockaddr that the server is listening on, and then using that field in php_thttpd.c to write the SERVER_ADDR separately from SERVER_NAME. Apr 29 07:30:18 is it a good solution to add non standard functionnality to thttpd ? Apr 29 07:30:40 instead of adding 2 lines patch to openshowcenter ? Apr 29 07:30:45 is the CGI/1.1 spec the standard, or is Apache now the standard :-) Apr 29 07:31:07 how many server support SERVER_ADDR ? ;) Apr 29 07:31:15 Once it gets to the stage of bootable, I can help with fixing things if I can do it directly on the slug, and then generate patches against the originals. Apr 29 07:31:29 jf-work: you're probably right - wonder what effect the continual DNS lookups will have on OSC performance Apr 29 07:31:40 or does PHP cache them ... Apr 29 07:33:05 glc: dyoung knows how to change the kernel args, and jacques knows how to change slugimage to overlay the partitions. someone needs to validate that redboot will load a dummy initrd from a valid sysconf partition (which will then be ignored by the kernel args). Apr 29 07:33:35 jf-work: perhaps I will just create an openshowcenter Unslung package, and then just patch it there. Apr 29 07:34:59 on performance : aren't dns lookup only run at first launch of your media ? don't you run /etc/hosts based dns resolution for your local network ? Apr 29 07:35:52 jf-work: yeah, even a slideshow is a number of seconds between DNS lookups. You've convinced me :-) Apr 29 07:37:40 leaving work. have a good and unplugged weekend Apr 29 07:37:47 glc: bummer, do_move in redboot uses hard-coded offsets, not the mtd map Apr 29 07:37:58 jf-work: thx. Apr 29 07:38:16 glc: so we loose an erase block, but the other reasons to do it are still valid. Apr 29 07:38:52 glc: so you could do it by just adding the linuxrc from switchbox into the unslung rootfs, and then changing the kernel args to boot it directly. Apr 29 07:39:34 we would still need to insert a dummy initrd for redboot to do_move(), but slugimage can do that ... Apr 29 07:41:00 glc: I think your idea has legs. Apr 29 07:41:02 night all Apr 29 08:29:11 * jstueve lurks... Apr 29 09:03:07 hello clueless Apr 29 10:19:18 jp30-work: ping Apr 29 10:34:06 hm, CIA-8 doesn't report new checkins Apr 29 10:34:25 at least for me Apr 29 10:38:56 it hasn't been working for a while Apr 29 10:39:16 ah, ok Apr 29 10:39:41 shall we report the checkins be hand in the meantime? Apr 29 10:40:10 feel free to.. ;) Apr 29 10:40:59 postfix now compiles cross and postfix.mk generates control files Apr 29 10:42:06 <[g2]> anyone ever here of a amvet router ? Apr 29 10:49:13 03ptweety * 10unslung/ (make/postfix.mk sources/postfix/postfix-install.patch): Apr 29 10:49:13 postfix.mk now generates control files Apr 29 10:49:13 ready for cross compilation Apr 29 10:49:29 hey! someone woke up CIA Apr 29 10:49:54 <[g2]> ByronT, so what's happening with OpenSlug ? :) Apr 29 10:50:08 loaded question! Apr 29 10:50:27 <[g2]> when the release ? Apr 29 10:51:04 not a set date - *right now*, but within the next week (as in 7 days) Apr 29 10:51:26 I've got to find some bandwidth for one.... Apr 29 10:52:59 re the amvet router, dogpile.com couldn't find anything Apr 29 10:57:40 I'll be polishing the license agreement this weekend, and then shooting for setting the click through up with binary for the first of next week Apr 29 11:54:54 hi Apr 29 11:57:17 hi perlguru Apr 29 11:59:26 much happened last two weeks? Apr 29 12:00:41 not on the slug I don't think Apr 29 12:01:09 some web server development Apr 29 12:01:35 heh - maybe I shouldn't have said that - jp30-work has been rolling right along Apr 29 12:01:48 :) Apr 29 12:02:20 we are trying to get more alpha testers for unslung 4.x Apr 29 12:02:26 maybe i'll add the perl Template-Toolkit Apr 29 12:02:55 what's that? Apr 29 12:03:10 www.template-toolkit.org Apr 29 12:03:33 interesting Apr 29 12:03:41 how about mod_perl? Apr 29 12:03:53 would be nice Apr 29 12:04:16 do we have CGI.pm already? Apr 29 12:04:40 that is a core module Apr 29 12:04:53 ah, shows how well i know perl :( Apr 29 12:04:54 try: perl -MCGI Apr 29 12:05:50 CGI::Application also nice Apr 29 12:06:41 jp30-work, are the perl module makefiles up to standard? Apr 29 12:07:59 i honestly don't know... ptweety might be a good person to ask, he has done a lot of work on unslung's perl modules Apr 29 12:08:32 they look straight forward Apr 29 12:09:09 perhaps i misunderstoof - you mean the .mk files? Apr 29 12:10:00 they look a bit variable Apr 29 12:11:04 yes i meant the makefiles Apr 29 12:42:43 yes i meant thelss Apr 29 12:42:57 oops Apr 29 12:43:24 perlguru, sorry i was absent there for a bit Apr 29 12:44:31 some of the perl-*.mk files are old-style with a seperate control, others are up to date Apr 29 12:44:48 which is up to date? Apr 29 12:45:30 * perlguru setting up a unslung-dev-environment on slug Apr 29 12:52:25 perl-spamassassin seems to be. Apr 29 13:32:42 jp30-work, could you test perl-appconfig? Apr 29 13:33:07 or anyone else using unslung and perl? Apr 29 13:46:37 I added perl-appconfig to native-testing Apr 29 13:55:01 perlguru, testing now Apr 29 13:55:25 Someone said nefarious. Apr 29 13:59:46 perl-appconfig is promoted Apr 29 14:05:28 great Apr 29 14:07:04 perl-template-toolkit almost done Apr 29 14:12:01 well that was scary and annoying Apr 29 14:12:24 a gentoo box I hadn't booted in about six months decided that upgrades were impossible Apr 29 14:13:28 then portage ate itself Apr 29 14:14:47 hehe, --omg-optimized Apr 29 14:19:41 jp30-work, added perl-template-toolkit to native-testing Apr 29 14:30:28 vim: error while loading shared libraries: libXaw.so.7: cannot open shared object file: No such file or directory Apr 29 14:30:42 why? Apr 29 14:31:01 why does vim require X? Apr 29 14:32:16 probably because of the way it was configured/autoconfigured at build-time Apr 29 14:42:35 hi Apr 29 14:43:15 where do i get any linux x86 bitkeeper client? :) Apr 29 14:45:32 bitkeeper.com? Apr 29 14:46:06 there i have to register...oh i just found that there exists a cvs server... :) Apr 29 14:46:23 'night all Apr 29 15:35:52 03jp30 * 10unslung/make/vim.mk: vim: configure --without-x Apr 29 15:59:11 dyoung: did you see glc's idea about booting straight into jffs2? Apr 29 15:59:13 glc, rwhitby: I have the beginnings of a plan. Gotta think about it a bit more. Apr 29 15:59:26 dyoung: spill Apr 29 15:59:52 Can you explain to me why this isnt supposed to work? Apr 29 16:00:45 ... Apr 29 16:01:33 what's "this" ? Apr 29 16:01:52 oh, "This" meaning just making a jffs2 the size of the original (pre-switchbox) ramdisk Apr 29 16:01:59 and telling the kernel to boot it Apr 29 16:02:10 I'm not sure if I'm miunderstanding the problem. Apr 29 16:02:21 jffs2 is different format from initrd Apr 29 16:02:35 Do we care? Apr 29 16:02:47 do_move doesnt seem to check. Apr 29 16:02:59 yeah, cause it has to be on erase block boundaries, whereas ext3 can start after the length header Apr 29 16:03:16 can't start jffs2 in the middle of a block Apr 29 16:03:27 Aha, okay. Apr 29 16:03:48 whats in the first 16bytes of a jffs2 filesystem? Apr 29 16:03:59 I was just thinking the same thing Apr 29 16:04:20 the first two bytes is a magic header - one of a few different values Apr 29 16:05:16 for us it needs to be guaranteed to be less than 10MB (otherwise it will overwrite the kernel which was just moved before it) Apr 29 16:05:37 (before it in time, after it in the memory map) Apr 29 16:06:38 I think that now we are directing people to unsling to an external disk, then we don't need recovery mode any more, and we can do all the switching in the /linuxrc in the jffs2. Apr 29 16:06:59 we waste an erase block, but C'est la vie. Apr 29 16:07:20 Lemme think about it more later. back later. Apr 29 16:07:44 it will also make the linuxrc simpler, cause the modules and exe's will all be there in the right spot, so no copying required. Apr 29 16:09:08 and no need to .altroot, cause you just edit the linuxrc directly in jffs2 Apr 29 16:09:24 Right. Apr 29 16:09:46 also simplify openslug booting. Apr 29 16:11:35 Hmm. what about flash from web interface? Apr 29 16:12:01 Set up ramdisk and pivot? Apr 29 16:12:54 glc: yeah, no change to the logic - it just happens in the linuxrc in jffs2 instead Apr 29 16:13:04 dyoung: yeah. Apr 29 16:13:11 Looks like it's death to switchbox! Apr 29 16:22:33 long live ... ? Apr 29 16:25:32 dyoung: I'm checking an initial cut at the linuxrc now ... Apr 29 16:25:39 checking in Apr 29 16:28:20 jacques: shall we make slugimage put a dummy length at the start of a single block ramdisk area, and put the payload in there? Apr 29 16:29:07 03rwhitby 07 * r1.3196 10openembedded/packages/nslu2-binary-only/ (unslung-rootfs_2.3r25.bb unslung-rootfs-2.3r25/linuxrc): First cut of new linuxrc for Unslung Apr 29 16:29:35 dyoung: what should the kernel args be? Apr 29 16:30:28 CMDLINE_ROOT = "root=/dev/mtdblock4 rootfstype=jffs2 rw init=/linuxrc mem=32M@0x00000000" Apr 29 16:30:29 ? Apr 29 16:31:12 we can test today, still using the current switchbox as the dummy ramdisk (which will not get used) Apr 29 16:31:31 rwhitby, we could, but why put it there instead of the last block? we still need the partition table and trailer in the last block so we can't use it for anything else Apr 29 16:32:05 yeah, you're right. Apr 29 16:32:21 so, what can we use 256Kb-16b for ? Apr 29 16:33:49 i guess there's no reason why we couldn't have multiple payload bays Apr 29 16:34:08 yeah, but we're not even using the first one now ... ;-) Apr 29 16:35:07 heh :-) Apr 29 16:35:25 Hmm - I can't really test this until my wl530g arrives, cause I don't have a spare LAN port for devslug ... Apr 29 16:37:15 which reminds me I need another switch Apr 29 16:37:28 I hear the RV042 is a good one ... Apr 29 16:39:09 dyoung, jacques: does that kernel args line look right? Apr 29 16:40:41 looks OK to me Apr 29 16:40:54 need anything about the console in there? Apr 29 16:42:05 that's done elsewhere Apr 29 16:42:12 in OE Apr 29 16:42:29 ah Apr 29 16:44:33 dyoung, glc: are you able to test that linuxrc? Apr 29 16:47:47 If you have a download somewhere or if it is checked in for a normal build. Apr 29 16:49:21 <[g2]> rwhitby, can you join me in openjtag ? Apr 29 16:50:22 glc: here it comes .... Apr 29 16:50:29 03rwhitby 07 * r1.3197 10openembedded/packages/linux/unslung-kernel_2.3r25.bb: New Unslung kernel args to boot straight from jffs2 linuxrc Apr 29 16:50:43 Should I duck? Apr 29 16:51:01 should be able to build unslung-image Apr 29 16:51:15 O.K. Off I go. Apr 29 16:52:08 rwhitby: This will be possible to do with openslug as well? And that would fre up (most of) the initrd partition? Apr 29 16:52:15 Or am I way off now? :) Apr 29 16:53:46 NAiL: yes, all except 1 block Apr 29 16:53:56 Nice :) Apr 29 16:54:17 (which is only one block, cause switchbox is two blocks) Apr 29 17:06:33 sounds good. Apr 29 17:06:39 and we can recover 128kb Apr 29 17:06:55 at least. Apr 29 17:07:06 plus payload2 Apr 29 17:07:25 I'm not sure sure why using the whole "original ramfs" cant be written as a jffs2. is there somehting hard coded in the kernel? no right? Apr 29 17:07:47 I must be missing something obvious. Apr 29 17:09:51 redboot will try and use the first 16 bytes as a length header, and do_move will overwrite the just-loaded-into-memory kernel at 0x1a00000 if the first two bytes are a length greater than 10MB Apr 29 17:10:01 (maybe four bytes) Apr 29 17:10:23 it's a redboot thing, not a kernel thing Apr 29 17:10:30 Okay, I must have misinspected do_move when I looked at it this morning. lemme look again. Apr 29 17:13:05 Oh. Apr 29 17:13:07 My bad. Apr 29 17:13:19 I just kinda glossed over that filesize bit. ;-) Apr 29 17:18:09 dyoung: it's worth an investigation - we might be lucky Apr 29 17:18:31 as long as that block does not get erased by the jffs2 in it's garbarge collection Apr 29 17:19:11 so we need to basically find out what the first 4 bytes are of a jffs2 fs Apr 29 17:19:25 "I deleted /random-file and now my slug doesn't boot any more ...." Apr 29 17:19:35 and verify its "stuff" and not important stuff. Apr 29 17:19:45 "same here, but it was a completely different file" Apr 29 17:20:05 dyoung: it's a magic number - I looked at it a while ago Apr 29 17:20:14 it's in the slugimage sources ... Apr 29 17:21:04 jacques knows it, cause he wrote that bit ... Apr 29 17:21:13 0x19852003, 0x0000000c, 0xf060dc98 Apr 29 17:21:35 wheres my hex calculator.... Apr 29 17:22:05 well, thats a big nubmer Apr 29 17:22:14 428MB Apr 29 17:22:19 bummer Apr 29 17:22:33 428154883 to be exact. Apr 29 17:22:49 I believe 0x1985 is the jffs2 magic Apr 29 17:24:10 That still puts in in the realm of "Large Number". Oh well; my plan sucked. Apr 29 17:24:20 so instead of thinking of this as loosing 128Kb of jffs2, we should think of it as gaining 128Kb of jffs2 and 1MB or so of RAM .... Apr 29 17:24:43 ya. Apr 29 17:25:04 its a feature, not a bug. ;-) Apr 29 17:26:00 anyone recommend a media server to use with twonky? Apr 29 17:26:43 So what nefarious use will the payload have... Apr 29 17:27:04 music... some dixv torrents (tv not movie) Apr 29 17:27:20 photos Apr 29 17:27:56 jstueve: sorry, that wasnt direct at you; I was speaking of the payload in the firmware which we just changed around. Apr 29 17:28:04 lol.. okay.. sorry Apr 29 17:28:10 I was looking for an excuse to say "nefarious". Apr 29 17:28:19 * jstueve loves the word nefarious Apr 29 17:28:30 is it the word of the day? Apr 29 17:29:25 I dunno, but it is one of my favorite words. Apr 29 17:29:28 ~nefarious Apr 29 17:29:35 nefarious is, like, to be infamous by way of nastyness Apr 29 17:29:35 I read on Yahoo that the OpenSlug binary problem might have a solution... Apr 29 17:29:41 /linuxrc: 50: Syntax error: "fi" unexpectedKernel panic: Attempted to kill init! Apr 29 17:29:52 jstueve: Neuston MC-500, served from OpenShowCenter on Unslung Apr 29 17:30:02 hmm, that *is* a problem now isnt it... Apr 29 17:30:12 if you fsck up your linuxrc, you'll be rendered unbootable. Apr 29 17:30:41 glc: apologies - as I said I couldn't test it .. Apr 29 17:31:32 glc: at least upslug will work to recover Apr 29 17:31:50 (if you use the reset button trick, or catch it in serial) Apr 29 17:31:57 I have serial. Apr 29 17:32:04 yeah, I figured so Apr 29 17:32:30 jstueve: nefarious is the nslu2-linux mascot word .... Apr 29 17:32:53 so is this caveat emptor for linuxrc modifications performed by Joe Enduser ? Apr 29 17:33:00 sounds mean.. lean... Apr 29 17:33:25 dyoung-web: yep. joe user will have unslung and will never see jffs2 again Apr 29 17:33:46 rwhitby, did you see that i had trouble building libdvb? Apr 29 17:33:56 nope Apr 29 17:34:13 "../include/devices.hh:14:32: linux/dvb/frontend.h: No such file or directory" Apr 29 17:34:15 hehe I was giggling when I saw jp30 demote it. Apr 29 17:34:29 Flag files would still be safer for Joe user. Mess one up it will still boot. Apr 29 17:34:34 jstueve: dyoung is going to buy the US version of the Neuston Apr 29 17:34:36 looks like some of the linux headers are missing from the toolchain Apr 29 17:34:46 glc: yeah, we keep the flag files for unslung to use Apr 29 17:35:22 jp30-work: the DVB package is downloaded for those Apr 29 17:35:27 inside libdvb Apr 29 17:35:44 so I must have missed an include somewhere Apr 29 17:35:55 was just going to say... a missing -I maybe? Apr 29 17:36:26 yeah - i've got it there for the configure at leat I think Apr 29 17:36:32 let me try building it here from scratch Apr 29 17:36:59 yeah, it builds for me and is using -I../builds/libdvb/DVB/include ... Apr 29 17:37:30 and ipk is done. Apr 29 17:37:31 weird Apr 29 17:37:38 is the log on nudi? Apr 29 17:38:24 let me try again and create a log this time Apr 29 17:39:10 dyoung: yeah, jp30 now trumps me on packages stuff :-) Apr 29 17:39:40 in meritocracy terms, he surpassed me quite a while ago .... Apr 29 17:40:04 look at libdvb.log in the packages dir Apr 29 17:40:49 ...or i can cut and paste more bits of it to irc if that would be easier Apr 29 17:42:20 jp30-work: it's not downloading DVB properly Apr 29 17:42:39 see the zcat error ... Apr 29 17:42:52 ah, i see Apr 29 17:45:56 can you add $(DL_DIR)/$(LIBDVB_HEADERS_SOURCE) to the .configured target dependencies? Apr 29 17:46:07 no probs Apr 29 17:46:11 actually, I can do it here Apr 29 17:46:47 03rwhitby * 10unslung/make/libdvb.mk: Added the dependency for the DVB headers download. Apr 29 17:46:55 that's why jp30 is the package release manager, and not me :-) Apr 29 17:49:05 ok, back later. glc: good luck with the linuxrc Apr 29 17:49:56 Kinda hard to edit when it won't boot. Apr 29 17:50:32 I'll try editing the downloaded copy and rebuild. Apr 29 17:50:34 yeah, you'll have to edit it in BK and rebuild unfortunately Apr 29 17:50:51 you've got BK write access, right? Apr 29 17:51:00 you can check in your fixes :-) Apr 29 17:51:03 I could never get it to work. Apr 29 17:51:30 I'll try again. Apr 29 17:51:50 03jp30 * 10unslung/Makefile: promote libdvb Apr 29 17:54:55 'night all Apr 29 17:58:15 jp30-work: mod-python just failed for me doing a rebuild of all packages from scratch on nudi in my area Apr 29 17:58:31 sh: line 1: /home/rwhitby/unslung/packages/staging/opt/bin/build/libtool: No such file or directory Apr 29 17:59:16 that looks like a strange path to be using ,,,, staging/opt/bin/build/libtool ... Apr 29 17:59:35 rwhitby-away, add a wiley? Apr 29 17:59:54 ka6sox-zzzz: pardon? Apr 29 18:00:04 oh, farm Apr 29 18:00:10 Add a wiley to the compile farm? Apr 29 18:00:15 not until we have native compilation working Apr 29 18:00:24 and we'll need to raise some more donations to do it Apr 29 18:00:39 k Apr 29 18:00:42 we don't know if native compilation is possible yet ... Apr 29 18:01:12 ka6sox ping Apr 29 18:01:20 pong Apr 29 18:16:34 rwhitby-away, `./staging/opt/bin/apr-config --installbuilddir`? Apr 29 18:17:47 could be the case that mod-python was missing apr-stage Apr 29 20:23:13 03bzhou * 10unslung/ (Makefile make/samba.mk sources/samba/configure.in.patch): made samba cross compilable, generate control file, strip binaries Apr 29 20:41:35 03bzhou * 10unslung/make/samba.mk: upstream upgrade from 3.0.11 to 3.0.14a Apr 29 23:33:18 * dyoung bah humbugs. Apr 29 23:49:06 heh **** ENDING LOGGING AT Fri Apr 29 23:59:56 2005