**** BEGIN LOGGING AT Thu Jul 14 23:59:57 2005 Jul 15 00:05:36 is postfix building fine in openslug for everyone else? Jul 15 00:06:24 The log file says: "/home/slug/openslug/tmp/staging/i686-linux/bin/postconf: error while loading shared libraries: libpcre.so.0: cannot open shared object file: No such file or directory" Jul 15 00:06:47 Then "./postfix-install: Error: "" should be "no" or an absolute path name." Jul 15 00:07:06 Yes, that was seen yesterday. Jul 15 00:07:11 let's see what it does on nudi .... Jul 15 00:07:19 It succeeds Jul 15 00:07:50 It is because of the make -e problem I believe. Jul 15 00:08:12 what's the make -e problem? Jul 15 00:08:13 (Same as pcre, which almost always fails the first time it is built). Jul 15 00:08:48 The first problem - the missing libpcre.so.0 - is because postconf has been linked without the -Wl,rpath directives. Jul 15 00:09:01 That is the standard SYSLIBS macro from the Makefile. Jul 15 00:09:38 However you should find that temp/run.do_compile contains a SYSLIBS export which should override the value in the Makefile by virtue of make being run with -e. Jul 15 00:10:33 IIRC DaKa2 was trying to track it down, but had to sleep (I think that was what happened...) Jul 15 00:10:51 I cannot reproduce it. nudi doesn't show the problem. Jul 15 00:11:14 I can reproduce something similar in the pcre builds, but haven't had time to try tracking it down. Jul 15 00:12:09 The postfix-install problem is because postconf -h fails (so a local shell variable ends up empty). Jul 15 00:14:59 ok, I can reproduce the postfix failure every time on my home dev box Jul 15 00:15:13 but pcre builds fine Jul 15 00:16:06 Yeah, that's the way it seems to work. Someone made a really gross fix to the pcre problem in the .bb file - I think it doesn't work for me because of make 1s timestamp granularity. Jul 15 00:17:03 DaKa2 deleted the evidence before we could check it. Jul 15 00:17:38 I believe you will find that log.do_compile built postfix without the -Wl (actually I'm sure you will). Jul 15 00:18:28 The question I wanted to know the answer to was whether a hand re-run of temp/run.do_compile* produces the same link (no -Wl) or not. Jul 15 00:26:49 It doesn't help that the .bb file has unset CFLAGS CPPFLAGS CXXFLAGS Jul 15 00:27:34 And no comments. And an explicit load of exports (including a modified SYSLIBS). Jul 15 00:27:54 ok, so what should I be looking for in log.do_compile? Jul 15 00:27:57 I think the maintainer of this package needs to look at it... Jul 15 00:28:12 Look for the step which links 'postconf' Jul 15 00:28:35 Then go in to /src/postconf, rm postconf, and re-run temp/do_compile Jul 15 00:28:57 (hum, put '*' characters before and after the bold bit ;-) Jul 15 00:28:58 "ccache armeb-linux-gcc -march=armv5te -mtune=xscale -DHAS_PCRE -I/home/slug/openslug/tmp/staging/armeb-linux/include -fexpensive-optimizations -fomit-frame-pointer -frename-registers -O2 -g -O -I. -I../../include -DLINUX2 -o postconf postconf.o ../../lib/libglobal.a ../../lib/libutil.a -L/home/slug/openslug/tmp/staging/armeb-linux/lib -lpcre -ldb -lnsl -lresolv" Jul 15 00:29:28 Yep - look in temp/run.do_compile* and you will see that the end of that doesn't match the exported SYSLIBS... Jul 15 00:30:54 Here's what I get by rerunning do_compile (after deleting postconf): Jul 15 00:30:58 ccache armeb-linux-gcc -march=armv5te -mtune=xscale -DHAS_PCRE -I/home/work-tmp/jbowler/nslu2/openslug/staging/armeb-linux/include -Os -g -O -I. -I../../include -DLINUX2 -o postconf postconf.o ../../lib/libglobal.a ../../lib/libutil.a -lpcre -ldb -lnsl -lresolv -L/home/work-tmp/jbowler/nslu2/openslug/staging/armeb-linux/lib -Wl,-rpath-link,/home/work-tmp/jbowler/nslu2/openslug/staging/armeb-linux/lib -Wl,-O1 Jul 15 00:31:55 is postfix an #oe thing, or one of our new additions? Jul 15 00:33:38 "ccache armeb-linux-gcc -march=armv5te -mtune=xscale -DHAS_PCRE -I/home/slug/openslug/tmp/staging/armeb-linux/include -fexpensive-optimizations -fomit-frame-pointer -frename-registers -O2 -g -O -I. -I../../include -DLINUX2 -o postconf postconf.o ../../lib/libglobal.a ../../lib/libutil.a -lpcre -ldb -lnsl -lresolv -L/home/slug/openslug/tmp/staging/armeb-linux/lib -Wl,-rpath-link,/home/slug/openslug Jul 15 00:33:44 Hum... well, if I got back to the bk tree I see kergoth, CoreDump, ggilbert, mickeyl Jul 15 00:34:05 ok, so its an #oe package Jul 15 00:34:31 I suspect the last changes were mainly for bitbake changes, so it's one of the quite large number kergoth ported 2003/2004 Jul 15 00:34:57 The changes I was complaining about (SYSLIBS, the unset) were not in there in the bk tree. Jul 15 00:35:22 hmm - so what does monotone annotate say about it ? Jul 15 00:36:36 I think it gives me the revision ID of a merge. Jul 15 00:37:38 go to /home/slug/openembedded and run "mt annotate packages/postfix/postfix_2.0.20.bb" Jul 15 00:37:50 looks like you have to run annotate from the top of the repo with a full path name Jul 15 00:38:02 Really? Jul 15 00:38:05 ah, sorry - you were already past that Jul 15 00:38:20 ok, so we have hashes for each line ... Jul 15 00:38:45 hmm - not very useful Jul 15 00:39:10 No - it's the same wherever you run if from. It's one of the 'I take a name from the manifest' commands (like log) as opposed to 'I take a file in the working tree'. (And I find these two classes of commands really annoying...) Jul 15 00:40:20 We need to be able to see the full DAG. It's dropping the real revisions by choosing the merges - this is the same as the behaviour which gives unnecessary 3-way merges. Jul 15 00:41:44 I don't know - maybe something happened between the BK tree and monotone... Jul 15 00:44:44 Ah ha. The changes were in 41be7c702d48ccbee9015d2f92589a297f4cb0a2 Jul 15 00:46:21 No, they were in right at the start too. Jul 15 00:47:36 so it's definitely #oe heritage? Jul 15 00:48:27 Yes. I was looking at the -native file in bk. This stuff is kergoth stuff. Jul 15 00:52:38 He puts '-e' into EXTRA_OEMAKE to cause the top level make to be called with -e. This invokes the make magic which should cause all sub-makes to get -e. Jul 15 00:53:00 -e causes anything in the environment to totally cancel settings of the same variable in the makefile. Jul 15 00:53:41 He sets a whole slew (about 4) of environment variables to fake out the make and get the correct link line. Jul 15 00:53:51 For some of us it works, for some of us it doesn't... Jul 15 00:54:20 ok, at home I'm using "GNU Make 3.80" - what's it on Nudi? Jul 15 00:54:20 rwhitby: what version of make do you have? Jul 15 00:54:38 Same version here. Jul 15 00:55:06 Same version on nudi. Jul 15 00:55:37 GNU bash, version 3.00.16(1)-release Jul 15 00:55:55 (that's also /bin/sh) Jul 15 00:55:57 Do you still get the problem if you delete src/postconf/postconf and re-run temp/do_compile* ? Jul 15 00:57:05 checking ... Jul 15 00:57:29 It will work. Jul 15 00:58:08 By which I mean that the gcc link will have -Wl,rpath (etc) on the command line. Jul 15 00:58:30 So, can I get mine to fail ;-) Jul 15 00:59:21 Yes, I can: ccache armeb-linux-gcc -march=armv5te -mtune=xscale -DHAS_PCRE -I/home/work-tmp/jbowler/nslu2/openslug/staging/armeb-linux/include -Os -g -O -I. -I../../include -DLINUX2 -o postconf postconf.o ../../lib/libglobal.a ../../lib/libutil.a -L/home/work-tmp/jbowler/nslu2/openslug/staging/armeb-linux/lib -lpcre -ldb -lnsl -lresolv Jul 15 00:59:32 MFLAGS is your enemy Jul 15 01:00:08 I got mine to fail by constructing a Makefile which simply executed temp/run.do_compile* and using 'make'. Jul 15 01:00:36 If I repeat this excercise using 'make -e' it succeeds. Jul 15 01:01:44 I suggested to DaKa2 (then forgot about it) that he use -e with the top level Makefile - this, of course, causes other things (which work without -e) to fail. Jul 15 01:03:17 unset MFLAGS I think Jul 15 01:04:44 Go jbowler! Jul 15 01:10:11 Well, it's something in the environment. 'env -i' fixes it. Jul 15 01:11:01 MAKEFLAGS Jul 15 01:11:27 The fix is to put unset MFLAGS; unset MAKEFLAGS; bitbake ... in the master makefile. Jul 15 01:11:52 I.e. just put those two unset commands before the bitbake commands. Jul 15 01:12:41 Long term a better fix might be to use the whole env -i thing to precisely control the environment. Jul 15 01:15:07 now I need to sleep... Jul 15 01:20:11 03jbowler 07org.openembedded.nslu2-linux * r55a9cfb8... 10/packages/ (5 files in 2 dirs): (log message trimmed) Jul 15 01:20:11 mpd does not build on osuclibc because libvorbis will not build, that Jul 15 01:20:11 happens because it attempts a static link with the -lm library in the Jul 15 01:20:11 wrong place. Jul 15 01:20:11 monotone split into two versions, one for the netsync-4 versions (up to Jul 15 01:20:13 and including 0.19) one for netsync-5 versions. Both may be installed Jul 15 01:20:15 on one NSLU2, allowing the system to handle both <=0.19 and >=0.20 Jul 15 01:32:10 jbowler-zzz: we could put those things in setup-env Jul 15 01:34:21 testing that now. adding unset MFLAGS and unset MAKEFLAGS to setup-env in openslug dir Jul 15 01:51:42 ok, that seems to have worked. pushing it. Jul 15 01:53:46 03rwhitby 07org.nslu2-linux.dev * r4a6487d7... 10/openslug/Makefile: Added unset MFLAGS and unset MAKEFLAGS to the generation of setup-env in openslug/Makefile Jul 15 02:54:13 03rwhitby 07org.openembedded.nslu2-linux * r068a291c... 10/packages/meta/unslung-packages.bb: Added package-index to unslung-packages Jul 15 03:51:24 bob_tm: g'day Jul 15 03:51:32 bbiab (putting kids to bed) Jul 15 03:53:28 bob_tm :) Jul 15 03:59:48 Ping NAiL Jul 15 04:00:14 rwhitby: Good evening Jul 15 04:22:04 pong Jul 15 04:31:38 bob_tm: g'day. how's things? Jul 15 04:32:03 bob_tm: are you interested in the Unslung release manager job ? Jul 15 04:40:01 [g2]: morning :) Jul 15 04:41:32 <[g2]> hey Jul 15 04:43:46 rwhitby: Out buying coffee for a sec. Making it now - back in 2 mins Jul 15 04:43:54 ok Jul 15 04:54:07 rwhitby: ping? Jul 15 04:57:22 rwhitby-away: ping Jul 15 05:08:32 bob_tm: pong Jul 15 05:08:47 would you believe that I locked myself out of the openslug system that runs miau? Jul 15 05:09:08 No - how? Jul 15 05:09:24 I reinstated an old passwd file from before I migrated to latest 2.2-beta, and it had root's home diretory in a different place, so the ssh keys were no longer found ... Jul 15 05:09:24 * [g2] listens intently Jul 15 05:09:47 just had to reboot with disk unplugged, hotplug disk, edit file, reboot Jul 15 05:09:54 Oh - been there. Jul 15 05:10:23 BTW: I've turboed my slug. Works well. Jul 15 05:10:34 Made serial too. Jul 15 05:10:55 Tomorrow I will borrow a dev slug for two weeks while the owner is on holiday. Jul 15 05:12:09 Anybody know if I can make a 1-wire bus without that USB<->1-wire adapter? Jul 15 05:12:21 http://www.digitec.ch/php/mitte_loaddetail.php?artikel=109273 <--- hmm Plextor ConvertX... hardware DivX over USB2...... does that work on linux ? :) Jul 15 05:12:34 could make a slug into a TiVo with that Jul 15 05:14:11 Fuzzel: CHF is not a currency I know - what is the price of this? Jul 15 05:15:19 bob_tm: I got dalsemi to send me a usb ibutton adapter for free - just filled out the sample request on their website Jul 15 05:15:44 rwhitby: Cool - will try. Jul 15 05:16:50 also got a 32Kb eeprom from them, and have stored all my keys on it :-) Jul 15 05:16:57 (ibutton) Jul 15 05:17:15 rwhitby: Unslung Pack Mgr.? Same answer as before - I can't commit to that amount of time/work. Need to continue to focus on money making activities. Jul 15 05:17:44 bob: Swiss Franks, 1.5x EUR (600 CHF = 400 EUR) Jul 15 05:17:47 bob_tm: unslung release manager - firmware, not packages. Jul 15 05:18:12 rwhitby: I stand corrected. Still, same applies. Jul 15 05:18:32 someone knowledgeable, who is running unslug, and can fix bugs. Jul 15 05:18:51 and maybe add a kernel module or two if someone needs it Jul 15 05:18:55 03rwhitby 07org.openembedded.nslu2-linux * rcc91df51... 10/conf/machine/nslu2.conf: Nailed CVSDATE for gnu-config for nslu2-linux Jul 15 05:19:00 rwhitby: add 'time' to the criteria Jul 15 05:19:17 we don't have any great ideas for new features - that's all being done in the packages now Jul 15 05:19:46 rwhitby: I know, but experience tells me that things often take more time than what it looks like. Jul 15 05:19:55 bob_tm: fair enough Jul 15 05:20:22 how about just being someone who has write access and fixes the bugs, but is not responsible for the releases? Jul 15 05:20:38 rwhitby: Modules I can do. Jul 15 05:20:40 (at least just fixes bugs that affect them) Jul 15 05:21:27 bob_tm: are you set up with monotone ? Jul 15 05:21:36 Yes, but I have some minor problems Jul 15 05:22:24 bob_tm: it should build now. Jul 15 05:22:27 (protocol number mismatch and problem with undefined variable "default-collection") Jul 15 05:22:38 ah - you need version 0.19, not 0.20 Jul 15 05:23:00 you will need 0.20 next week after mickeyl upgrades the server :-) Jul 15 05:23:02 I though as much, but portage doesn't have it yet. Maybe there is an unstable release Jul 15 05:23:29 It says 0.20 installed Jul 15 05:23:53 0.20 won't work with our current repo. Jul 15 05:23:59 must be 0.18 or 0.19 Jul 15 05:24:01 So revert back Jul 15 05:24:04 Will do Jul 15 05:24:08 yep, until next week :-) Jul 15 05:24:38 rwhitby: Very nice the new dev setup Jul 15 05:24:48 thx Jul 15 05:25:56 rwhitby: Any plans for multiple sources for the package downloads? Jul 15 05:26:04 (optware) Jul 15 05:26:48 Like FOO_SITE_BACKUP= Jul 15 05:26:50 now that we have feed mirrors set up, we should mirror the sources too Jul 15 05:26:57 then we could do that Jul 15 05:38:37 rwhitby: Was it the DS9490R or B adapter you got? Jul 15 05:39:59 B Jul 15 05:40:07 iButton holder. Jul 15 05:41:06 thanks Jul 15 06:25:23 re all Jul 15 06:33:23 trouble building openslug. think i missed a step. help? Jul 15 06:34:23 <[g2]> kitno455, what happened ? Jul 15 06:34:36 monotone ok, make setup ok Jul 15 06:34:49 is next step make openslug? Jul 15 06:35:18 cause it complains about No such file or directory: '/data/slug/openslug/oe-symlinks/packages/*/*.bb' Jul 15 06:36:19 <[g2]> build-openslug Jul 15 06:36:28 data/slug/openslug/oe-symlinks/packages/* are all files, not dirs Jul 15 06:36:46 thats what i did, sorry Jul 15 06:36:52 build-openslug and openslug are the same Jul 15 06:36:54 it gives those errors Jul 15 06:37:01 yes, they look same in makefile Jul 15 06:37:07 same target Jul 15 06:37:25 i must have missed a step, or my kit is broken Jul 15 06:37:53 check this: Jul 15 06:38:05 cat openslug/oe-symlinks/packages/zlib Jul 15 06:38:12 link ../../openembedded/packages/zlib Jul 15 06:38:37 <[g2]> ls Jul 15 06:38:42 <[g2]> it's a directory Jul 15 06:38:48 no its not Jul 15 06:38:54 <[g2]> that's your problem Jul 15 06:38:55 its a file containing that line Jul 15 06:39:12 check it Jul 15 06:39:20 file openslug/oe-symlinks/packages/zlib Jul 15 06:39:25 openslug/oe-symlinks/packages/zlib: ASCII text, with no line terminators Jul 15 06:39:36 not a dir Jul 15 06:40:08 <[g2]> file openslug/oe-symlinks/packages/zlib/ Jul 15 06:40:08 <[g2]> openslug/oe-symlinks/packages/zlib/: directory Jul 15 06:40:19 <[g2]> I don't know about FCxxxx Jul 15 06:40:20 nice for you :) Jul 15 06:40:40 <[g2]> that's a simple issue on the svn pull Jul 15 06:40:40 ok, off to look at Makefile, find the code that tries to make symlinks... Jul 15 06:40:58 Rename the oe-symlinks dir Jul 15 06:41:00 Do make setup again Jul 15 06:41:05 ok, try that. Jul 15 06:41:29 <[g2]> actually you wanted to go into openslug Jul 15 06:41:40 <[g2]> and just update-symlink iirc Jul 15 06:41:59 go into openslug? Jul 15 06:42:03 it's a symlink in there Jul 15 06:42:11 looking at makefile in openslug/ Jul 15 06:42:19 <[g2]> ok Jul 15 06:43:29 mv oe-symlinks foo Jul 15 06:43:29 Done it yet? Jul 15 06:43:33 make setup Jul 15 06:43:43 same Jul 15 06:43:58 oe-symlinks has packages/ dir Jul 15 06:44:07 but files are all ascii text Jul 15 06:44:13 update svn maybe? Jul 15 06:44:22 svn --version Jul 15 06:44:36 1.0.9 (r11378) Jul 15 06:44:58 It's up to 1.2.1 now Jul 15 06:45:04 right Jul 15 06:45:15 lets update that and try again.... Jul 15 06:45:25 Update it and see if it helps. Something is screwy with your setup. Not seen anybody else with this problem.... Jul 15 06:45:34 what else uses svn? Jul 15 06:45:41 i will need to blast those checkouts too Jul 15 06:45:57 bitbake Jul 15 06:46:00 <[g2]> grep svn Makefile Jul 15 06:46:00 <[g2]> [ -e bitbake/bin/bitbake ] || ( svn co svn://svn.berlios.de/bitbake/trunk/bitbake ) Jul 15 06:46:00 <[g2]> [ -e oe-symlinks/packages ] || ( svn co svn://svn.berlios.de/openslug/trunk/openslug/nslu2-linux oe-symlinks ) Jul 15 06:46:00 <[g2]> ( cd bitbake ; svn update ) Jul 15 06:46:00 <[g2]> ( cd oe-symlinks ; svn update ) Jul 15 06:46:57 You should remove openslug/tmp as well if you've got a potentially broken install Jul 15 06:47:12 that is cache dir? Jul 15 06:47:34 It's the work dir. Contains the stuff it's compiling, the cache etc.... Jul 15 06:47:49 basically you start over from scratch on the build if you remove it Jul 15 06:48:21 dont have JIT compiler on this machine, can i install it after i trash opeslug/tmp? Jul 15 06:49:22 don't know. you'll have to ask one of the others Jul 15 06:51:22 <[g2]> you don't need JIT it just makes things faster Jul 15 06:54:12 hm, managed to reproduce my crash for the third time in the case, but never outside with a fan... so there might be a thermal issue with turboslug.., next step, fan in case.. Jul 15 06:55:04 what crash? Jul 15 06:55:43 http://david.thg.se/saker/dammit-050715 Jul 15 06:55:56 thats one, there has been a few different Jul 15 06:56:14 totaly dead on serial and ssh after that Jul 15 06:57:56 and its damn hot.. Jul 15 07:00:50 how hot is that? Jul 15 07:01:17 Do you even have a heatsink? Jul 15 07:02:40 I have nothing to measure with, but its to hot to keep your finger on the cpu Jul 15 07:02:51 uh? Jul 15 07:02:57 Call me paranoid but I affixed heatsinks to both the core and USB2 chip using special thermal cement/. Jul 15 07:03:03 It's never that hot for me Jul 15 07:03:12 and all of the board and memory+flash is warm Jul 15 07:03:21 The NEC chip is supposed to be hot Jul 15 07:03:24 Im pushing it really hard Jul 15 07:03:28 well, Im not using usb Jul 15 07:03:30 DaKa2, what are you doing with it Jul 15 07:03:38 <[g2]> Tiersten, mount via NFS rootfs and do large builds Jul 15 07:03:44 exactly Jul 15 07:03:45 A heatsink will help spread the heat away from the hot spots on the chip. Jul 15 07:04:15 [g2]: Your one is that hot as well? Jul 15 07:04:18 Hmm, Ill see if I can find some thermal glue.. or something Jul 15 07:04:31 Most PC shops sell some. Jul 15 07:04:36 if it is lying on its side, you can use nearly anything Jul 15 07:04:37 <[g2]> Tiersten, I don't think so Jul 15 07:04:53 <[g2]> But I'm cooler here by 10 F Jul 15 07:05:11 <[g2]> and I'm running slower due to disk I/O Jul 15 07:06:06 <[g2]> His built time was 10 minutes faster than my FatSlug build time Jul 15 07:06:15 <[g2]> for perl Jul 15 07:06:29 python... Jul 15 07:06:37 <[g2]> ok python Jul 15 07:07:09 building over NFS is faster than local disk? insane Jul 15 07:07:10 <[g2]> ok 3 minutes Jul 15 07:07:44 local disk goes thru usb port Jul 15 07:08:09 <[g2]> it's not insane at all Jul 15 07:08:11 usb high over head, and controlled thru main processor Jul 15 07:08:19 DaKa2: If your one a FWIXP420BB or a FWIXP420BL? Jul 15 07:08:23 network gets some help from coprocessor Jul 15 07:08:26 BL Jul 15 07:08:31 hrm Jul 15 07:08:33 all my slugs are BL Jul 15 07:08:42 will need to get some more data on this Jul 15 07:08:55 All your slugs are belong to us. Jul 15 07:08:58 do we know aproximate serial number break on slugs Jul 15 07:09:05 between two processors? Jul 15 07:09:12 No Jul 15 07:09:22 It appears the very early ones were BB Jul 15 07:09:24 <[g2]> more likely the model number would change Jul 15 07:10:09 this could be something completly else, maybe not at all because of heat, but thats the only thing Ive managed to come up with Jul 15 07:10:12 <[g2]> Linksys tweaked the model numbers on the boxes and devices slightly for most wrt54g changes Jul 15 07:10:29 what if BL is lower speed bin of same part. Jul 15 07:10:36 it would likely have more heat issues Jul 15 07:11:20 <[g2]> DaKa2, is it always dying at the same spot ? Jul 15 07:11:32 [g2]: no Jul 15 07:11:37 its really random Jul 15 07:11:48 <[g2]> I'm mean in the ooops Jul 15 07:12:05 usually with an kernel oops or five, but they are completly random too Jul 15 07:12:36 <[g2]> an oops collection would be interesting Jul 15 07:13:37 Ill try making it crash some more then :-) Jul 15 07:16:49 <[g2]> well the most intersting one will be with the usb-ethernet adapter Jul 15 07:17:03 <[g2]> cause that's one you can email to lak Jul 15 07:18:20 well, if its heat you get kernel oops on random things so that shouldn't be interesting anyway, but if I get them with a fan on then that would be interesting Jul 15 07:20:26 <[g2]> well just spreading out the heat on the chip with a heat sink might just be the ticket Jul 15 07:20:55 just make sure you apply the thermal paste liberally to fill the gaps. Jul 15 07:21:20 You might want to just test it out with normal thermal compound before actually gluing it. Jul 15 07:21:34 I know there is an ir thermometor somewhere around here.. Jul 15 07:21:42 uhm, i dont think liberally is good.... Jul 15 07:21:54 you want the bare minimum to fill the gaps Jul 15 07:22:11 I've just attached one with normal stuff right now for testing Jul 15 07:23:07 well, not too much obviously, but you have to spread it evenly around. Jul 15 07:23:15 thermal goo should be spread as thinly as possible. it's just to fill out any tiny ridges in the surface of the IC or the heatsink. Jul 15 07:23:34 yup Jul 15 07:23:38 direct contact from the heatsink to the CPU is best. then it's thermal goo and finally air Jul 15 07:23:39 ok, my choice of word was wrong! :) Jul 15 07:24:10 I assume the IXP is fairly flat. Lapping the package would be insane Jul 15 07:24:20 hehehe I was just thinking about that! Jul 15 07:24:39 The heatsink might not be though. Jul 15 07:24:52 anything is better than nothing! Jul 15 07:24:56 If it's good enough by just leaving the case open then I doubt you'd need to Jul 15 07:25:22 doesn't hurt though Jul 15 07:25:41 uh huh.... Jul 15 07:25:58 No point doing it if there isn't any use though Jul 15 07:27:49 well, he was talking about overheating so perhaps he does have a problem. Jul 15 07:28:35 Taking it out of the case with a fan works fine Jul 15 07:29:40 a fan is noisier, more expensive so I would go passive first. Jul 15 07:29:58 but then again, whatever the user wants to do is fine. Jul 15 07:30:07 :) Jul 15 07:30:14 Most of the time you don't need to lap anything Jul 15 07:30:20 It's just good enough Jul 15 07:30:35 If it was that terrible then the heatsink manufacturers would go out of business in 5 minutes Jul 15 07:30:39 hard to lap it when the battery, and speaker, etc are nearby Jul 15 07:30:57 nah. the heatsink Jul 15 07:31:00 Of course! people only lap in rare cases. Jul 15 07:31:15 lap? Jul 15 07:31:20 no point in lapping th sink if the cpu is bumpy Jul 15 07:31:21 grind it smooth Jul 15 07:31:29 like valves or lenses Jul 15 07:31:32 oh.. Jul 15 07:31:37 is it bumpy? I thought it was flat? Jul 15 07:31:53 look at the plastic with a lense Jul 15 07:32:26 not flat like the metal heatspreader of a P4 or athlon 64 Jul 15 07:32:38 you mean smooth? Jul 15 07:32:39 or the ceramic of an athlon/p3 Jul 15 07:32:56 Because I've just looked at mine and put a ruler up against it and it is flat Jul 15 07:33:10 different order of magnitude Jul 15 07:33:17 ceramic? Jul 15 07:33:25 what ceramic? Jul 15 07:33:40 athlons and some p3 have a ceramic top Jul 15 07:33:48 Some thermal paste would be fine with the flatness of the IXP Jul 15 07:34:03 I've never seen a P3 with a ceramic top Jul 15 07:34:23 that pink or green rectangle? Jul 15 07:34:27 That's not ceramic Jul 15 07:34:32 that's the bottom of the substrate Jul 15 07:34:52 thats raw silicon wafer? Jul 15 07:34:59 It's the actual die with some sort of protective goo Jul 15 07:35:01 yes Jul 15 07:35:04 stupid Jul 15 07:35:19 no wonder the sligtest touch ruins the cpu Jul 15 07:35:40 that's why places started selling shims like crazy Jul 15 07:35:56 dumb athlon with their rubber pads Jul 15 07:36:35 got svn 1.2.1 installed Jul 15 07:36:41 openslug building now Jul 15 07:36:42 you get good thermal transfer because you're directly touching the substrate but the area is tiny so overall it's not too great Jul 15 07:37:01 same amount of contact a heat spreader would have Jul 15 07:37:09 but one less layer that way Jul 15 07:37:23 good idea for hermetically sealed equipment Jul 15 07:37:33 but end users are clumsy Jul 15 07:37:43 the heatspreaders stop people cracking it because they're too clumsy when putting on the heatsink Jul 15 07:37:48 right Jul 15 07:37:51 or they've bought a crap design of heatsink Jul 15 07:38:07 I've seen some where you can pretty much keep clamping it down until the core cracks Jul 15 07:38:10 oh, dont get me started on that :) Jul 15 07:38:14 :) Jul 15 07:38:32 I've never managed to break a cpu, and I've built over 200 computers.. Jul 15 07:38:45 some of the real hardcore overclockers have removed the heatspreader Jul 15 07:38:48 on newer chips Jul 15 07:38:54 hell, i had a K6 that i cut the heatspreader off of Jul 15 07:38:58 heh Jul 15 07:39:06 like 1998 or so Jul 15 07:39:17 learned real quick not to do that again Jul 15 07:39:30 it broke? Jul 15 07:40:02 no, but then i had to build a new clip for heatsink Jul 15 07:40:07 ah Jul 15 07:40:19 which broke tabs off board socket Jul 15 07:41:11 hell, i used to run a 386-40 at 50mhz with water cooling in 1995 or so Jul 15 07:41:17 I remember when these flipchips first came out. Some people didn't understand and started to lap the die even though it is flat Jul 15 07:41:48 i gave up on overclocking a long time ago. i got a job :) Jul 15 07:41:50 Surprisingly, the CPU doesn't appreciate you grinding the die Jul 15 07:41:56 yeah Jul 15 07:42:25 do you know what size process is used on IXP? Jul 15 07:42:47 No. I doubt it's anything exotic Jul 15 07:43:24 i am just amazed to see 266mhz that wont burn you :) Jul 15 07:44:42 It's very low power Jul 15 07:44:55 0.18 micron Jul 15 07:45:03 It's in the product brief for the IXP$20 Jul 15 08:03:57 rwhitby-asleep: MFLAGS/MAKEFLAGS - yes, I didn't actually look at the master Makefile when I said that. All that needs to happen is that the variables are removed somewhere between the top level make and the one bitbake runs. Jul 15 08:59:42 who figured out the hotspot on the IXP? Somewhere near the "i", wasn't it? Jul 15 09:03:09 dyoung, Jul 15 09:03:13 figured it out Jul 15 09:05:20 aha Jul 15 09:07:13 Considering sticking a very very small heatsink on that spot Jul 15 09:07:37 just glue an aluminum roofing nail to it Jul 15 09:08:44 roofing nail? Jul 15 09:09:05 hmm, you dont know anything about nails :) Jul 15 09:09:18 :-P Jul 15 09:09:26 used for asphalt or fiber shingles Jul 15 09:09:36 aaah Jul 15 09:09:37 ok Jul 15 09:09:38 big, flat head Jul 15 09:09:40 yeah Jul 15 09:09:59 i have used that before on small fets when i had no sink Jul 15 09:10:06 lol NAiL doesn't know about nails Jul 15 09:10:11 I've got three tiny (1cm*0.5cm) aluminum heatsinks that I could use Jul 15 09:10:27 jacques: not in english, no :) Jul 15 09:10:50 NAiL, i am a dumb american using ascii pine Jul 15 09:11:03 what does the first character of your first name look like? Jul 15 09:11:19 An O with a / through it Jul 15 09:11:28 Ø Jul 15 09:11:36 cool. its just a blank for me Jul 15 09:11:44 this slash / not this one \ Jul 15 09:11:48 Also known as the "oe" ligature Jul 15 09:11:56 forward-slash, yes Jul 15 09:12:15 Only used in Norway and Denmark, afaik Jul 15 09:12:24 And my parents had to stick one in my name Jul 15 09:12:24 world gets a little smaller every day Jul 15 09:12:37 what is your default charset there? Jul 15 09:13:06 cp865, iso8859-1, utf-8, Jul 15 09:13:19 oh, its in 8859-1? Jul 15 09:13:37 yeah, 8859-1 does all the western chars Jul 15 09:13:59 eastern europe uses a different one IIRC Jul 15 09:14:10 i figured they were 'northern' chars :) Jul 15 09:14:35 well, yeah. But we're lucky enough to be north-*west* :D Jul 15 09:15:46 btw- thanks for taking the openslug maintainer mantle Jul 15 09:15:56 It's a real pain in the a** getting the � (OE) right. � (AA) or � (AE) is more common Jul 15 09:16:27 np. Haven't been able to do much the last week though. Hope to be putting in more work this weekend and next week :) Jul 15 09:17:10 building openslug now Jul 15 09:17:23 i see that it could not get upslug via http Jul 15 09:17:36 so it does cvs instead? Jul 15 09:18:24 I thought it did cvs in any case? Jul 15 09:18:42 it tried to get http://www.oesources.org/source/current/upslug_cvs.sourceforge.net_UPSLUG_1_0_.tar.gz Jul 15 09:18:46 but got a 404 Jul 15 09:18:54 so then it tried cvs and got it Jul 15 09:18:58 ah Jul 15 09:21:55