**** BEGIN LOGGING AT Tue Feb 23 02:59:58 2010 Feb 23 03:29:31 03bzhou * r11345 10optware/trunk/make/transmission.mk: transmission: 1.83 -> 1.91 Feb 23 06:22:36 03bzhou * r11346 10optware/trunk/make/ (bzr.mk bzrtools.mk): bzr: 2.0.4 -> 2.1.0; bzrtools: 2.0.0 -> 2.1.0 Feb 23 07:07:45 03bzhou * r11347 10optware/trunk/ (make/bind.mk platforms/packages-syno-i686.mk): bind: disable epoll on syno-i686 Feb 23 07:18:34 03bzhou * r11348 10optware/trunk/make/plowshare.mk: plowshare: 0.8.1 -> 0.9.1 Feb 23 09:23:51 03jomaster * r11349 10optware/trunk/make/pkgconfig.mk: pkgconfig: using configure.patch for host-build to remove UNDERQUOTING-bug Feb 23 09:24:32 03jomaster * r11350 10optware/trunk/make/gettext.mk: gettext: add host-stage Feb 23 09:25:55 03jomaster * r11351 10optware/trunk/make/ (automake14.mk automake19.mk): automake19 automake14: managing automake-host-stage on demand Feb 23 09:27:03 03jomaster * r11352 10optware/trunk/Makefile: Makefile: new host-tool managing Feb 23 09:27:55 03jomaster * r11353 10optware/trunk/make/gift-opennap.mk: gift-opennap: using automake19 host-tool-system Feb 23 09:29:01 03jomaster * r11354 10optware/trunk/make/libcdio.mk: gettext: using automake19 host-tool-system / add gettext-trick to always find libiconv Feb 23 09:29:31 03jomaster * r11355 10optware/trunk/make/gift-ares.mk: gift-ares: using automake14 host-tool-system Feb 23 15:55:41 jomaster: why not change to "if aclocal-1.9 --version 2>/dev/null; then ..."? Feb 23 15:56:13 the current $(shell ..) construct still takes time at macro expansion time Feb 23 16:01:44 actually, "if !" Feb 23 16:06:09 i'm also not sure the value of conditionally using host tool Feb 23 16:06:22 i like predictability better Feb 23 16:06:41 if we're using our own host tool, then in all cases use our own Feb 23 16:07:20 simplicity and predictability is more important in the long run Feb 23 16:08:57 think about in the future you need to troubleshoot - are we using /usr/bin/aclocal-1.9 or host/staging/opt/bin/aclocal-1.9 ? Feb 23 16:09:31 what if the behavior is slightly different? Feb 23 16:12:54 why should the behavior be different? it is aclocal-1.9 on /usr/bin/ as on host/staging/opt/bin. Feb 23 16:14:47 so why building only a simple tool if it already exists? Feb 23 16:16:11 about your condition: your are the expert. if you say $(shell isn't good, we use your way. Feb 23 16:22:43 aclocal is just an example Feb 23 16:23:10 take erlang for example, do we check erl already exists on build machine? Feb 23 16:25:11 we might need to work on libtool2-host-stage soon Feb 23 16:25:17 you are the maintainer. and i also suppose that building always the host-tool and use it, we are more stable. Feb 23 16:25:37 what about gcc-3.3-host-stage Feb 23 16:26:03 only stage it when needed Feb 23 16:26:36 don't check for gcc-3.3 existence Feb 23 16:27:15 i'm not the maintainer, i'm just feed manager and build master Feb 23 16:27:27 :-D Feb 23 16:29:17 so how to check? checking version of HOSTCC and than checking for gcc-3.3? Feb 23 16:30:21 if you like i remove automake-checks and add host-staging-bin on top of PATH to force using host-tool Feb 23 16:33:50 i said "don't check for gcc-3.3 existence" Feb 23 16:34:29 in the cases the host gcc-3.3 is needed, just go ahead and do host staging, and never look back Feb 23 16:35:25 in the case of automake 1.9 or 1.4, I propse "we don't check existence" either Feb 23 16:36:03 but you said: "only stage it when needed" Feb 23 16:36:14 when we need to use aclocal-1.9, just host staging, and use $(STAGING_PREFIX)/bin/aclocal-1.9 Feb 23 16:36:52 yeah, only fsg3v4 native toolchain needs gcc-3.3 Feb 23 16:37:10 so don't do gcc-3.3-host-stage for other platforms Feb 23 16:37:46 only a handful of packages need automake-1.9, only automake19-host-stage for these few packages Feb 23 16:38:44 and than remove host-stage? Feb 23 16:39:38 we have not solved the unstage problem, neither for staging, nor for host/staging Feb 23 16:40:14 but we survived for so long for staging, hope we can survive for a little bit for host/staging without removal Feb 23 16:42:07 if debian/redhat can make aclocal-1.9 to co-exist with aclocal-1.4 under /usr, we probably can make it work in host/opt Feb 23 16:42:33 for me it works Feb 23 16:42:45 on debian and also in host-stage Feb 23 16:43:38 i summarized: i remove checking for automake/aclocal and we build it everywhere we need. and we don't expand PATH but using automake/aclocal-host-stage-paths Feb 23 16:44:29 everywhere we need means 1.9 or 1.4 Feb 23 16:44:54 ah, not both together? Feb 23 16:45:00 use $(HOST_STAGING_PREFIX)/bin/aclocal-1.9 Feb 23 16:45:37 or $(HOST_STAGING_PREFIX)/bin/aclocal-1.4 Feb 23 16:45:44 think no problem. we can making automake19-host-stage depending on automake14-host-UNstage Feb 23 16:46:19 can automake19-host-stage and automake14-host-stage co-exist? Feb 23 16:46:24 yes Feb 23 16:46:30 works here Feb 23 16:46:38 let's not unstage Feb 23 16:47:08 think about the case when we have automake13 Feb 23 16:47:27 unstage is difficult Feb 23 16:47:35 until now i change name from automake to automake-1.9 but should be better with programm-suffix, isn't it? Feb 23 16:48:25 most packages work with /usr/bin/aclocal fine (don't require specific version) Feb 23 16:48:32 we'll leave it the way it is Feb 23 16:48:50 just use aclocal (or autoreconf) Feb 23 16:49:23 only use host tool for aclocal-1.9 or 1.4 Feb 23 16:49:55 okay, I gotta go to work Feb 23 16:49:57 but for the 3 packages a specific version is required we call: ACLOCAL="aclocal-1.9 -I m4" AUTOMAKE=automake-1.9 autoreconf -i -v; Feb 23 16:50:28 yes, we mean both the same Feb 23 16:50:59 only for the 3 packages, use ACLOCAL="$(HOST_STAGING_PREFIX)/bin/aclocal-1.9" or 1.4 Feb 23 16:51:06 yes Feb 23 16:51:11 as i said Feb 23 16:51:28 only for the 3 packages, we need automake1x-host-stage Feb 23 17:59:28 03bzhou * r11356 10optware/trunk/ (make/x264.mk sources/x264/encoder-analyse.c.patch): x264: fixed uclibc build by adding log2f macro Feb 23 18:39:05 @eno: we should TARBALLS_DIR RESULT_TOP from crosstool-scripts and specify them in crosstool{-native}.mk Feb 23 18:39:59 and we should set TARBALLS_DIR to $(DL_DIR). i realy don't like "TARBALLS_DIR=$HOME/downloads" Feb 23 18:40:39 to my mind optware should be a system with all in optware-dir Feb 23 18:41:08 (of course not all tools, but donwloads,...) Feb 23 18:49:30 i also don't like that crosstool-native uses /opt/ of build-machine, but don't know a solution. Feb 23 18:51:18 03jomaster * r11357 10optware/trunk/Makefile: Makefile: host-tool-stage without tests on build-machine Feb 23 18:51:57 03jomaster * r11358 10optware/trunk/make/ (automake14.mk automake19.mk): automake14 automake 19: host-tool-stage without tests on build-machine Feb 23 18:52:52 03jomaster * r11359 10optware/trunk/make/gift-ares.mk: gift-ares: set path for automake-1.4 host-tool-stage Feb 23 18:53:14 03jomaster * r11360 10optware/trunk/make/gift-opennap.mk: gift-opennap: set path for automake-1.9 host-tool-stage Feb 23 18:53:37 03jomaster * r11361 10optware/trunk/make/libcdio.mk: libcdio: set path for automake-1.9 host-tool-stage Feb 23 19:22:52 03jomaster * r11362 10optware/trunk/make/gcc.mk: gcc: add host-tool-stage Feb 23 19:24:29 03jomaster * r11363 10optware/trunk/make/crosstool-native.mk: crosstool-native: add gcc-host-tool-stage / set TARBALLS_DIR Feb 23 19:47:48 03robham * r11364 10optware/trunk/make/samba34.mk: samba34: SAMBA_cv* changed to samba_cv* - package now runs with shares working (although still some run time errors loged during share access), package stable enough for testing Feb 24 00:35:30 03bzhou * r11365 10optware/trunk/platforms/packages-ts101.mk: packages-ts101: demoted x264 Feb 24 00:35:31 03bzhou * r11366 10optware/trunk/ (3 files in 3 dirs): x264: allow ts101 to use earlier snapshot that builds **** ENDING LOGGING AT Wed Feb 24 03:00:02 2010