**** BEGIN LOGGING AT Sat Jun 18 02:59:56 2011 Jun 18 07:35:08 mirko * r27213 /trunk/toolchain/eglibc/Config.in: [toolchain/eglibc] when selecting eglibc, use version 2.12 by default Jun 18 07:36:57 mirko Hi, so if I now build from the latest openwrt trunk everything, libs from /usr/lib are now found without ldconfig and LD_LIBRARY_PATH? Jun 18 07:38:48 no, it's not yet committed Jun 18 07:39:59 lamikr: http://pastebin.com/EKYTgWqH Jun 18 07:48:35 Ok, btw. About my ldconfig patch. It's also done in a way that it did not affect to uclibc builds as the ldconfig and the linkbinding script, etc. would only be included to glibc and eglibc builds. Jun 18 07:51:57 lamikr: there's no benefit of that - no package will include rules just for glibc/eglibc Jun 18 07:52:32 when there's gonna be use-case - maybe - but right now it's just overload.. Jun 18 07:52:37 *be a use-case Jun 18 07:57:19 mirko Ok, I understand your point and anyway in the end I am happy if things just start working by adding /usr/lib for ld search path. There were anyway already a sector in base-files Makefile for copying certain files only in glibc/eglibc cases. So I just added these ldconfig stuff to that section. Jun 18 07:58:11 so in that way they were not included when base-package was packaged for uclibc. Jun 18 08:58:37 anybody cares about elibc versions < 2.11 ? Jun 18 09:14:27 mirko: why are some people {e,g}libc instaed of uclibc? are there any programs not working with uclibc? Jun 18 09:47:09 Hauke: uclibc is still buggy when using complex applications, especially while using threads Jun 18 09:47:23 i got deadlocks, race-conditions, etc. Jun 18 09:48:16 additionally the uClibc afaik didn't get audited yet quite well and i was told by people it has some security flaws Jun 18 09:49:32 and sometimes size doesn't matter but performance Jun 18 09:50:07 Hauke: i switched to eglibc from uclibc in one project because of stated issues Jun 18 09:50:21 mirko: ok then we should support one version of eglibc Jun 18 09:50:57 Hauke: i'm getting 2.13 in right now and would like to purge out everything below 2.11 Jun 18 09:51:11 yes makes sense Jun 18 09:51:28 and what about glibc our version is ~4 years old Jun 18 09:51:42 Hauke: i'd like to to the same with glibc Jun 18 09:51:50 but eglibc is my main focus right now Jun 18 09:52:00 ok nice Jun 18 10:59:10 hi Jun 18 10:59:22 can someone help me hack bootloader sources? Jun 18 11:15:41 mirko * r27214 /trunk/toolchain/eglibc/Makefile: [toolchain/eglibc} eglibc in fact can be built with -Os Jun 18 11:15:44 mirko * r27215 /trunk/toolchain/eglibc/ (Config.in Config.version Makefile): [toolchain/eglibc] add support for eglibc version 2.13 Jun 18 11:15:47 mirko * r27216 /trunk/toolchain/eglibc/patches/ (4 files in 2 dirs): [toolchain/eglibc] patch: add /usr/lib to default search path of dynamic linker Jun 18 11:26:39 shit - latest supported version of glibc is 2.9? current is 2.13 - really would like to get rid of glibc completly.. Jun 18 11:34:47 mirko: glibc 2.13 does not work with openwrt? Jun 18 12:29:57 <_trine> larsc, was there any resolution to crypto_hash not compiling in kirkwood ? Jun 18 12:55:05 nope Jun 18 12:56:10 but i don't think that it is kirkwood specific Jun 18 13:00:07 <_trine> yes I kind of thought that Jun 18 13:00:29 <_trine> at the moment kirkwood will not compile with it Jun 18 13:14:32 i might have time to take a closer look today or tomorrow evening Jun 18 14:40:50 <_trine> larsc, thanks Jun 18 14:47:27 ping nbd Jun 18 14:47:32 hi Jun 18 14:49:31 pong Jun 18 14:49:32 hi Jun 18 15:04:53 nbd did you read the mail i sent to ml? Jun 18 15:05:04 it seems the b44 driver has a bug Jun 18 15:05:25 where should i report this error? in the lkml? Jun 18 15:17:47 tripolar: seen it, but didn't look at the code yet Jun 18 15:26:26 hauke * r27217 /trunk/target/linux/brcm47xx/ (22 files in 2 dirs): brcm47xx: add support for kernel 3.0 Jun 18 15:48:52 nbd but would you report it at lkml? Jun 18 15:52:20 tripolar: well, first it should be reproducable; having a small program triggering it would be quite helpful Jun 18 15:53:51 mirko: see updates to #9483 and bugs it references/fixes Jun 18 15:53:56 hauke * r27218 /trunk/ (12 files in 6 dirs): kernel: update to kernel 2.6.38.8 Jun 18 15:54:55 hauke * r27219 /trunk/target/linux/brcm47xx/patches-2.6.39/ (3 files): brcm47xx: refresh patches Jun 18 15:59:02 KanjiMonster: i have one here its just a short c program Jun 18 15:59:07 i sent it to the ml Jun 18 16:01:23 KanjiMonster the prog i used to test befor was written in c++ and a thread was sending - now in c i can procude the error in ~10sec - have 162 lines of b44 ssb0:0: eth0: powering down PHY after 10sec Jun 18 16:02:00 but the port is open and no firewall is dropping the packages Jun 18 16:02:32 the icmp unreachable message wont come trough under such a high load but the eth shouldnt go down Jun 18 16:07:34 KanjiMonster sent to the ml Jun 18 16:08:03 nbd: do you have a device ussing a b44 eth? if yes- could you try to reproduce the error? Jun 18 16:15:33 KanjiMonster: do oyu have a device with a b44 network card? Jun 18 16:21:32 when i drop all udp packets the error doesnt accure Jun 18 16:25:30 occur Jun 18 16:25:49 seems the mass of icmp unreachable packets that would be sent, are triggering the error Jun 18 16:37:48 i got disconnected Jun 18 16:42:16 nbd: should i remove just the transmit fifo error? the receive fifo doesnt seem to be included as it works when the packets are dropped or am i understanding something wrong here? Jun 18 16:42:38 #define ISTAT_RFO 0x00004000 /* Receive FIFO Overflow */ Jun 18 16:42:38 #define ISTAT_TFU 0x00008000 /* Transmit FIFO Underflow */ Jun 18 16:44:17 oh it an underflow Jun 18 16:54:58 moo Jun 18 16:57:15 tripolar: what do you mean by "receive fifo doesnt seem to be included as it works when the packets are dropped"? Jun 18 16:57:19 i don't understand Jun 18 16:58:52 * Chocky offers magic decoder ring Jun 18 17:00:05 nbd i ment, why doesnt the error occure when i drop the packets Jun 18 17:00:12 meant Jun 18 17:00:30 Chocky: are you a teacher? Jun 18 17:00:55 I'm not, and I fail to see the relevance. Are you trying to excuse your bad spelling when there's clearly some problem understanding you? Jun 18 17:01:34 Chocky okay i will write slower and correct Jun 18 17:21:39 nbd i sent you a mail. if i remove ISTAT_RFO, the eth doesnt come up anymore and i get a connection refused when i try to connect to the device Jun 18 17:25:11 did you double-check this? Jun 18 17:25:15 it sounds rather weird Jun 18 17:25:36 i don't see why disabling the interrupt for the receive overflow check would cause the ethernet to not come up anymore Jun 18 17:26:33 i only added this - -#define ISTAT_ERRORS (ISTAT_DSCE|ISTAT_DATAE|ISTAT_DPE|ISTAT_RDU|ISTAT_RFO| Jun 18 17:26:33 ISTAT_TFU) Jun 18 17:26:33 +#define ISTAT_ERRORS (ISTAT_DSCE|ISTAT_DATAE|ISTAT_DPE|ISTAT_RDU|ISTAT_TFU) patch and afterwards i get the connection refused Jun 18 17:27:00 try reverting it, then putting it back Jun 18 17:27:04 to check if that's really causing it Jun 18 17:27:20 yeah i just compile an image without the patch applied Jun 18 17:28:32 nbd: but at normal bootup i also have this message in dmesg but only one time Jun 18 17:28:32 b44 ssb0:0: eth0: powering down PHY Jun 18 17:33:35 okay you where right - the ssh server doesnt come up anymore - maybe there is somewhere another bug Jun 18 17:34:01 @nbd Jun 18 17:55:23 mirko * r27220 /trunk/toolchain/eglibc/Makefile: [toolchain/eglibc] explicitly disable the use of ldconfig Jun 18 19:09:59 is there a way to install a package after base-files during the makeprocess? Jun 18 19:10:16 it seems the make order has changed and the last package that gets installed is base-files Jun 18 19:10:32 where is this defined? Jun 18 19:11:11 ahh found it .. changeset 27138 Jun 18 19:16:14 the problem is only when someone installs a package, that adds a group/user in its image, all changes to files that base-files include will be lost ... Jun 18 19:16:49 group/user - etc/{passd/group} is only an example ... Jun 18 19:16:55 passwd Jun 18 19:19:28 ping jow_laptop Jun 18 19:22:15 base-files show be the first thing that get installed Jun 18 20:38:42 crap Jun 18 21:06:42 Hauke One good example from app requiring eglibc or glibc is omniorb that I have tried to get running with openwrt. With uclibc I can get it running, but it starts segfaulting in places where the code run in some methods at threads throws exceptions. With eglibc it works fine. Jun 18 21:06:58 yes, I had that too Jun 18 21:07:04 with my app. pthreads is bust **** ENDING LOGGING AT Sun Jun 19 02:59:57 2011