**** BEGIN LOGGING AT Sun Feb 14 02:59:57 2010 Feb 14 07:49:35 eno: are you there? Feb 14 07:57:13 jomaster: fsg3v4 is rebuilding Feb 14 07:57:27 very nice. Feb 14 08:00:41 eno: what do you think about **-stage-uninstall targets which will called after **-stage isn't required anymore like: configure-target(stage/configure) / build-target(build/stage-uninstall) Feb 14 08:01:32 this could solve the problem to change from mysql to mysql5, cause both can be used in different packages. Feb 14 09:07:52 @eno: svn-commits aborts since yesterday, but changes will accepted (using rapidsvn) Feb 14 16:49:23 jomaster: both host-tools and stage-uninstall are good ideas Feb 14 16:49:52 thats good Feb 14 16:49:53 other package systems even uses ipkg/opkg to manage staging Feb 14 16:50:13 so -stage == ipkg install Feb 14 16:50:27 -stage-uninstall == ipkg uninstall Feb 14 16:50:55 there're of course some tweaks needed that make staging ipk different from normal ipk Feb 14 16:51:16 but thats a lot of work to change this all Feb 14 16:51:26 yes exactly Feb 14 16:52:08 and most of us only do it in our spare time Feb 14 16:52:21 :-D Feb 14 16:52:43 would be nice if the change can be introduced incrementally Feb 14 16:52:48 i am ill since some day. so i could do a little work Feb 14 16:53:50 both our builder and developer test machines are still ubuntu 8.04 Feb 14 16:54:11 so staging means something like: ipkg install foo --prefix staging-dir Feb 14 16:54:16 there will be some breakage if we move to newer distribution Feb 14 16:54:45 right, also pkgconfig tweak if necessary Feb 14 16:55:57 that's to hard for my freetime. i always only have a few hours a month :-D Feb 14 16:56:11 but to change the whole system takes a little more Feb 14 16:56:32 yes, same here Feb 14 16:57:09 but good chance for a branch, isn't it? Feb 14 16:57:27 i believe debian packaging system uses a diff approach, each deb builds dependencies from scratch Feb 14 16:57:48 probably in chroot Feb 14 16:58:13 this is much more work for build-server Feb 14 16:58:42 true, but it's very clean Feb 14 16:58:43 same as "diskclean" before building a new package Feb 14 16:59:20 so also no staging-collusions happen Feb 14 16:59:34 right Feb 14 17:00:39 i will play around with building host-tools. thats also important for selfhosting, isn't it? Feb 14 17:00:55 and not such a great project Feb 14 17:01:18 yes, important for moving to newer host distro too Feb 14 17:03:51 last question: what happens to irc-boot reporting svn-changes? Feb 14 17:04:55 the problem might be in svn, irc-bot timed out too, eventually causing it to quit Feb 14 17:05:17 we've seen this before, rebooting svn should solve it Feb 14 17:05:31 i'll ask rwhitby or ka6sox for that Feb 14 17:06:04 and last note: somthing like 'diskclean' before building new package also reduce hidden build-errors. e.g. forgotten staging, but staging was created by a package before. Feb 14 17:07:00 yes, it's a big tradeoff of using machine resource vs. human resource Feb 14 17:08:29 wish we have a builder cluster, infinitely scalable by adding machines Feb 14 17:08:38 s/have/had/ Feb 14 17:08:38 eno meant: wish we had a builder cluster, infinitely scalable by adding machines Feb 14 17:10:24 i never will learn all about irc :-D Feb 14 19:06:35 @eno: is it possible to change a variable inside of a target (like command 'WGET=foo' inside of a target like 'check_wget)? Feb 14 19:31:31 jomaster: don't understand the question Feb 14 19:32:50 jomaster: BTW, don't add $(_HOST_BUILD_DIR) to -dirclean Feb 14 19:33:12 ok, why? Feb 14 19:33:27 host builds are supposed to be shared among different targets Feb 14 19:34:04 don't know if it possible. i have a makefile-rule and like to change the value of variable WGET Feb 14 19:35:10 what's the scope that WGET value should apply? Feb 14 19:36:44 optware .mk files, even though separated by files, are pretty flat in terms of scope Feb 14 19:40:06 i like to change the global variable WGET defined in /Makefile. Feb 14 19:40:50 for one particular platform (target), or for all platforms? Feb 14 19:42:44 what's the reason behind the change you want to make? use host built version? Feb 14 19:43:30 i like to change this variable during Makefile run inside of a rule. is this possible? (don't effort you. it is not so important) Feb 14 19:45:53 is that an explicit rule or implicit rule? Feb 14 19:47:19 can you use a WGET2 for example? Feb 14 19:49:36 *hihi* Feb 14 19:51:04 i explain my idea i had: i have a rule which check if wget exists and if not wget-host-stage is called and global variable WGET gets the path to host-stage wget. Feb 14 19:53:47 that seems just too complicated Feb 14 19:53:55 for wget, it's not worth the trouble Feb 14 19:54:22 we just want make to do simple things Feb 14 19:55:02 for tools like aclocal, we test the host version to be ready, then switch to use the host version Feb 14 19:55:39 for anything that busybox supports, let's not spend effort on it Feb 14 19:58:25 i know, that was a senseless sample. i only wanted to explain the variable-problem. Feb 14 20:00:19 you might be able to get it working with $(SHELL) and conditional setting of WGET Feb 14 20:00:35 but IMHO I'd avoid it Feb 14 20:01:16 we don't want to support the two cases at the same time Feb 14 20:12:39 jomaster: fsg3v4 rebuilt done Feb 14 20:19:55 that is great. i test it Feb 14 20:26:50 works great. many thanks **** ENDING LOGGING AT Mon Feb 15 02:59:57 2010