**** BEGIN LOGGING AT Sat Jan 15 03:00:01 2011 Jan 15 03:37:08 mirko * r25002 /trunk/target/linux/lantiq/ (6 files in 5 dirs): [lantiq] move irq.c from patch into own file Jan 15 03:37:21 mirko * r25003 /packages/libs/xmlrpc-c/Makefile: [packages/xmlrpc-c] split xmlrpc-c library suite into several packages Jan 15 03:37:24 mirko * r25004 /packages/libs/xmlrpc-c/Makefile: [packages/xmlrpc-c] add package "abyss" - fix dependencies Jan 15 03:45:05 mirko * r25005 /packages/libs/xmlrpc-c/Makefile: [packages/xmlrpc-c] fix dependency xmlrpc-c_abyss-server -> xmlrpc-c_abyss Jan 15 04:25:41 build #68 of at91 is complete: Success [build successful] Build details are at http://tksite.gotdns.org:8010/builders/at91/builds/68 Jan 15 09:27:43 gmorning Jan 15 12:13:56 is there any reason why xmlrpc-c is linking against libxml2? Jan 15 12:14:13 the internal used variant of expat is so much smaller Jan 15 13:00:08 mirko: most likely not Jan 15 13:02:26 mirko * r25006 /packages/libs/xmlrpc-c/Makefile: [packages/xmlrpc-c] use build variants to build xmlrpc-c with libxml2 as well as expat support Jan 15 13:46:00 nbd * r25007 /trunk/package/mac80211/patches/ (551-ath9k_txq_schedule_multi.patch 552-ath9k_fix_bar.patch): ath9k: fix a few more aggregation related issues, should improve stability Jan 15 13:46:04 nbd * r25008 /trunk/package/mac80211/patches/560-mac80211_wds_sta_fix.patch: mac80211: fix frames looping back to a wds station Jan 15 14:08:45 jow_laptop: used build variants now Jan 15 14:09:19 rtorrent is the only app using xmlrpc-c for now which seems to be broken anyway (no builds in snapshots) Jan 15 14:09:28 so i don't care about it for now Jan 15 14:10:51 it is broken due to xmlrpc-c ... Jan 15 14:11:06 can't resolve libxml2 symbol due to missing rpath-link Jan 15 14:14:13 jow_laptop: ah, hehe Jan 15 14:15:00 jow_laptop: i'll take a look Jan 15 15:31:31 nbd * r25009 /branches/backfire/package/mac80211/patches/ (3 files): mac80211: merge fixes from r25007, r25008 Jan 15 15:39:42 build #52 of etrax is complete: Failure [failed compile_3] Build details are at http://tksite.gotdns.org:8010/builders/etrax/builds/52 Jan 15 16:23:05 build #51 of ep93xx is complete: Failure [failed compile_4] Build details are at http://tksite.gotdns.org:8010/builders/ep93xx/builds/51 Jan 15 19:09:52 build #58 of ps3 is complete: Failure [failed compile_3] Build details are at http://tksite.gotdns.org:8010/builders/ps3/builds/58 Jan 15 19:13:48 mirko * r25010 /packages/net/rtorrent/Makefile: [packages/rtorrent] fix build, adjust to new xmlrpc-c structure Jan 15 19:59:39 mirko: I love you! ;) Have been waiting for this, thanks :* Jan 15 20:40:18 jow_laptop: ping Jan 15 20:49:16 das-Iro: lol :) Jan 15 20:49:54 das-Iro: didn't test however it compiles - please report issues, since the xmlrpc-c lib got restructured Jan 15 20:50:36 just started compiling Jan 15 20:52:25 nbd: does kconfig allows to trigger / hook certain functions when selecting options with particular deinfes (such as DEPENDS, etc.)? Jan 15 20:53:19 i'm not quite happy with the current PROVIDES situation - it requires to pre-select sth. to show up sth. totally different Jan 15 20:54:53 mirko: afaik not Jan 15 20:55:21 in this partciular case a libxmlrpc-impl. needs to be selected to show up rtorrent Jan 15 20:55:35 that's not quite intentional Jan 15 20:55:43 cannot be solved with a "default y" ? Jan 15 20:56:02 xMff: hm? Jan 15 20:56:15 DEFAULT:=y in the Package/foo define Jan 15 20:56:28 xMff: what?! Jan 15 20:56:33 i wasn't aware of that feature Jan 15 20:56:42 you can even pass whole expressions: DEFAULT:=y if ! CONFIG_PACKAGE_foo Jan 15 20:57:32 jesus Jan 15 20:57:32 :) Jan 15 20:58:15 so, if A and B provide Z, A is marked as default - and X depends on Z, selecting Z selects A? Jan 15 20:59:15 oh, nice Jan 15 20:59:32 well you need to specify it as: DEFAULT:=y if PACKAGE_X Jan 15 21:00:03 otherwise (just DEFAULT:=y) it will be always enabled unless a user explicitely disables it Jan 15 21:00:37 you need to inspect the generated kconfig source, I don't recall the exact semantics Jan 15 21:02:53 or even shorter: DEFAULT:=PACKAGE_foo then its selected along with "foo", so if foo=m it is m, if foo=y it is y Jan 15 21:03:29 xMff: hey Jo-Philip, can you have a look at the patch I submitted? it works for 2.6.33 and later, but it fails for earlier. not sure why. I suspect it's a KCONFIG issue or conflict with config-default... Jan 15 21:03:39 so in this particuar example the small xmlrpc-c impl needs DEFAULT:=PACKAGE_rtorrent Jan 15 21:04:46 xMff: i see - thanks Jan 15 21:05:55 but well, we really need provides and conflicts Jan 15 21:06:03 xMff: yes Jan 15 21:06:19 xMff: what you mentioned is quite nice, however I guess it does not fit the case here.. Jan 15 21:06:35 CONFLICTS would help pretty much Jan 15 21:07:01 since within xmlrpc-c rtorrent shouldn't be mentioned Jan 15 21:07:18 we could extend the dep syntax to allow !foo as reverse of +foo Jan 15 21:07:18 and i don't wanna create a feature flag for every particular build variant Jan 15 21:07:41 for flags it works already @!FOO Jan 15 21:07:44 since both varaints are not binary compatible Jan 15 21:07:48 indeed Jan 15 21:08:11 the conflicts can be propagated to pkg metadata as well, it supports them Jan 15 21:08:17 *opkg Jan 15 21:08:20 yes, that would help a lot i guess Jan 15 21:08:54 package x: depends:=!y, provide:=z package y: depends:=!x, provide:=z Jan 15 21:08:56 it supports provides as well, but expressing provides in kconfig is not trivial I think Jan 15 21:09:17 that's why i asked initially about hooks/trigger in kconfig Jan 15 21:10:08 well, i think it shouldn't be _too_ hard to implement PROVIDES functionality Jan 15 21:10:24 there's already a hook when selecting a pkg which selects all dependencies Jan 15 21:10:28 xMff: the linux-x.x.x/.config doesn't have CONFIG_CS5535_GPIO=m set... not sure why. Jan 15 21:10:48 philipp64|laptop: I will look tomorrow Jan 15 21:11:00 if mentioned pgk is exported as PROVIDES is should select the default-marked one Jan 15 21:11:06 *it Jan 15 21:11:35 or even _ask_ Jan 15 21:11:53 present a Config.in-styled interface listing packages which provide this Jan 15 21:12:05 well internally provides would get aggregated to a select/choice kconf section Jan 15 21:12:10 but that's the advanced version i guess Jan 15 21:12:16 xMff Jan 15 21:12:17 yep Jan 15 21:12:20 so all standard ways (menuconfig, oldconfig, ...) would ask Jan 15 21:12:42 no, I take it back... CONFIG_CS5535_GPIO=m was set after all. Jan 15 21:13:51 I have no idea how to express preference though Jan 15 21:14:08 as provides could be spread across multiple source packages Jan 15 21:16:25 it could be recursive indeed Jan 15 21:16:33 I could imagine it as follows: if a package depends on something that is a provides, then it gains a submenu in menuconfig stating Use "something" implementation: (x) something-foo ( ) something-bar ( ) something else Jan 15 21:16:54 that's what i meant called the "advanced version" :) Jan 15 21:17:21 if only one package is able to satisfy the provides it becomes a simple "select ..." Jan 15 21:17:39 and no submenu is shown Jan 15 21:19:41 well, no package should provide sth. no other package provides - but yeah Jan 15 21:21:17 xMff: by the way, any news regarding the iconv / intl stuff we talked about? Jan 15 21:21:32 didn't look into the log yet Jan 15 21:21:43 xMff: no, I take it back. it is building 2.6.32 correctly. my bad. I wasn't cleaning up correctly between changing versions. Jan 15 21:21:48 so I'm good for a commit. Jan 15 21:33:04 mirko: no, I'll see if I can tackle it tomorrow Jan 15 21:43:21 mirko: It compiles, lets ship it ;) Jan 15 21:44:07 das-Iro: which xmlrpc-c variant did you use? Jan 15 21:44:12 (internal) or (libxml2)? Jan 15 21:45:04 both? :D Jan 15 21:45:10 das-Iro: eeks Jan 15 21:45:15 http://mirror.das-iro.de/openwrt/kirkwood/packages/ Jan 15 21:45:27 ok, we _definitely_ need CONFLICTS Jan 15 21:45:48 das-Iro: only one should be selected Jan 15 21:46:18 yeah Jan 15 21:47:02 when i select xmlrpc-c_internal the subsection of "xmlrpc-c_libxml2" appears and shows --- xmlrpc-c_server Jan 15 21:47:45 yep Jan 15 21:47:57 either xmlrpc-c_internal or xmlrpc-c_libxml2 should be selected Jan 15 21:48:20 using interneal ow Jan 15 21:48:23 *now Jan 15 21:48:29 internal is smaller Jan 15 21:48:32 yep Jan 15 21:48:39 you do not need abyss_server btew Jan 15 21:48:41 *btw Jan 15 21:48:44 libxml2 400k :o Jan 15 21:48:48 neither the client Jan 15 21:49:02 das-Iro: 1,1M is libxml2 afaik Jan 15 21:49:31 yep, package allready is 400k Jan 15 21:50:11 so it would be even smaller on fs, since jffs2 uses lzma Jan 15 21:50:34 neath Jan 15 21:50:39 anyway, only makeds sense using xmlrpc-c_libxml2 when it co-exists with another package using libxml2, too Jan 15 21:50:59 true Jan 15 21:59:40 das-Iro: i'm curious, why you do not use transmission? Jan 15 22:00:07 rtorrent is better ;) Jan 15 22:00:34 well I could not setup transmission in 10s so i got bored Jan 15 22:00:55 and i already have a fancy config for rtorrent Jan 15 22:01:01 i see Jan 15 22:02:35 btw i think rtorrent and libtorrent can be build in parallel Jan 15 22:03:06 I compiled them an desktops, and never got problems Jan 15 22:03:31 (parallel as in make -j4) Jan 15 22:03:57 what version of 2.6.37 are people building? I want to try it on Geos. Jan 15 22:04:22 xMff: resent the diffs in case anyone wanted to eyeball them. Jan 15 22:15:22 mirko * r25011 /packages/libs/xmlrpc-c/ (.Makefile.swp Makefile): Jan 15 22:15:22 [packages/xmlrpc-c] disable build variants for now, as: Jan 15 22:15:22 - we do not have a CONFLICTS yet, so both versions could be built and Jan 15 22:15:22 would intefere Jan 15 22:15:22 - variants are not binary-compatible so applications using this lib can Jan 15 22:15:22 not depend on the meta-package (PROVIDES) exported by both Jan 15 22:15:23 - the libxml2 variant isn't going to be used soon anyway Jan 15 22:15:24 *sigh* Jan 15 22:15:26 anyone looking at why e2fsprogs fails to parallel build? Jan 15 22:16:32 philipp64|laptop: do they build fine in non-parallel mode? Jan 15 22:16:51 mirko * r25012 /packages/libs/xmlrpc-c/.Makefile.swp: delete accidentally committed *.swp file Jan 15 22:21:48 mirko: yes Jan 15 22:26:24 philipp64|laptop: then I think whoever wants to take a look is more than welcome to do so :) - there are other (in my opinion more important) fields which need some love as well Jan 15 22:27:09 well, sure... but if you're doing automated builds then it means you have to run the entire build -j1.... which drastically slows things down. Jan 15 22:27:32 that's totally correct Jan 15 22:28:07 isn't there a PARALLEL_BUILD:=0 or something that could be thrown into the Makefile to temporarily "fix" it? Jan 15 22:32:03 philipp64|laptop: try adding PKG_BUILD_PARALLEL:=0 to e2fsprogs' makefile Jan 15 22:33:26 a bit confused since, there are PKG_BUILD_PARALLEL:=1 and PKG_BUILD_PARALLEL:=0 Jan 15 22:35:11 guess it depends on what's the default, defined in PKG_BUILD_PARALLEL Jan 15 22:35:22 (PKG_BUILD_PARALLEL in .config I mean) Jan 15 22:47:24 Is a RaLink based platform router supported under OpenWRT? Jan 15 22:52:07 mazilo: there is some support for rt288x and rt305x, but none for rt3662/rt3883 (afaik) Jan 15 23:10:13 what's the command that updates the various config-default files with new symbols following a version bump? Jan 15 23:27:47 philipp64|laptop: do you mean "make oldconfig"? Jan 15 23:31:20 I thought there was a script that scribbled on target/linux/generic/config-default, etc. Jan 15 23:44:18 no such script then... Jan 15 23:47:51 KanjiMonster: added it and it made no different. build still fails. Jan 15 23:48:21 philipp64|laptop: hm, okay, then it doesn't do what I thought it does Jan 15 23:59:42 it looks to me like e2fsprogs is trying to build before uClibc has been built. Jan 16 00:03:02 actually no... it's being built before gcc has been compiled. Jan 16 00:43:35 man, that's a slooooow build. Jan 16 00:43:53 how do you build the toolchain manually? Jan 16 01:03:52 any committers around? Jan 16 01:09:32 build #60 of pxcab is complete: Failure [failed compile_3] Build details are at http://tksite.gotdns.org:8010/builders/pxcab/builds/60 Jan 16 01:29:41 nbd * r25013 /trunk/toolchain/uClibc/patches-0.9.32/460-powerpc_libm_fixes.patch: uClibc 0.9.32: fix libm compile for powerpc Jan 16 01:29:45 nbd * r25014 /trunk/package/mac80211/Makefile: mac80211: fix compile errors on powerpc Jan 16 01:29:51 nbd * r25015 /trunk/target/linux/mpc85xx/ (config-default patches/110-fix_mpc8548_cds.patch): mpc85xx: fix mpc8548 support (reorder serial ports, add default stdout path to device tree), enable pcie Jan 16 01:29:56 nbd * r25016 /trunk/target/linux/mpc85xx/config-default: mpc85xx: clean up bogus kernel config overrides Jan 16 01:30:00 nbd * r25017 /trunk/target/linux/generic/ (config-2.6.36 config-2.6.37): add missing kernel kernel config symbol Jan 16 01:30:04 nbd * r25018 /trunk/ (4 files in 4 dirs): Jan 16 01:30:04 mpc85xx: add a new 'spe_fpu' feature flag for the FPU used on freescale powerpc cpus Jan 16 01:30:04 The SPE FPU is ABI-incompatible with the regular powerpc FPU, Jan 16 01:30:04 this needs to be reflected in the toolchain target name. Jan 16 01:30:04 Fixes floating point crashes in user space Jan 16 01:30:57 nbd * r25019 /packages/net/iperf/Makefile: iperf: use build variants, fix uclibc compile issues **** ENDING LOGGING AT Sun Jan 16 02:59:58 2011