**** BEGIN LOGGING AT Tue Oct 16 03:00:02 2012
Oct 16 06:08:33 <russell--> hmm. on another machine r33780 builds x86-generic-net45xx for me
Oct 16 06:08:50 * russell-- will rev up first machine and try again
Oct 16 06:09:07 <russell--> once it finishes what it's doing
Oct 16 07:58:51 <EricJ> I'm making no sense of the menuconfig and how it handles profiles.
Oct 16 07:59:28 <EricJ> When I select a different profile, (e.g. go from profile A to profile B) it should unselect the packages in A and select the packages in B, right?
Oct 16 08:01:47 <markun> EricJ: might be possible with "make defconfig"
Oct 16 08:02:02 <markun> but I'm not totally sure
Oct 16 08:02:32 <EricJ> I'm unfamiliar with that one. What is it?
Oct 16 08:04:24 <markun> EricJ: http://wiki.openwrt.org/doc/howto/build#defconfig (don't know if it also unselects packages from A)
Oct 16 08:05:51 <EricJ> Just tried. It doesn't.
Oct 16 08:05:56 <EricJ> markun: Thanks, though.
Oct 16 08:07:59 <KanjiMonster> EricJ: selecting a profile has only any effect on the package list if you are starting with no config - it changes the defaults, not the actual selections
Oct 16 08:08:32 <markun> btw, can anyone help me with an IPv6 routing problem? With SixXS I set up a tunnel, I got a subnet from them which they route to my tunnel address, assigned an IPv6 address from the subnet to the lan interface and setup radvd to assign IPv6 addresses from the subnet.
Oct 16 08:10:04 <markun> Now my laptop gets an IP address, I can ping the lan's IP address, I can ping the tunnel's IP address but nothing outside the tunnel. I did set net.ipv6.conf.all.forwarding=1 is /etc/sysctl.conf
Oct 16 08:10:25 <markun> from openwrt I can ping outsied (e.g. ipv6.google.com)
Oct 16 08:12:58 <markun> so, that's where I'm stuck. Unfortunately I don't have access to the router from here to test anything right now.
Oct 16 08:35:22 <markun> oops, sorry for posting in the dev channel
Oct 16 10:01:04 <jow_laptop> EricJ: no it shouldn't deselect stuff
Oct 16 10:01:14 <jow_laptop> EricJ: profiles merely predefine defaults
Oct 16 10:01:30 <jow_laptop> if all corresponding symbols are already set they have no effect anymore
Oct 16 11:22:57 <jwendell> hi folks. is there an way to guarantee the order of installation of packages in the build process? if package B depends on package A, then A is installed first...
Oct 16 11:25:12 <jow_laptop> no
Oct 16 11:34:39 <written_tabletpc> i get this http://pastebin.com/SpGSVHXQ build error while building latest trunk for ar71xx platform
Oct 16 11:34:51 <written_tabletpc> any hints?
Oct 16 11:36:54 <blogic> rebuild your toolchain
Oct 16 11:37:39 <written_tabletpc> hmm, so make toolchain/clean: make
Oct 16 11:37:40 <written_tabletpc> ?
Oct 16 11:37:58 <written_tabletpc> s/:/;/
Oct 16 11:43:22 <markun> written_tabletpc: I ended up doing a "make distclean" and compiling everything again, but that's probably overkill.
Oct 16 11:43:51 <written_tabletpc> let's what happen
Oct 16 11:43:56 <written_tabletpc> *see
Oct 16 11:54:15 <jwendell> dude, openwrt makefiles are evil
Oct 16 11:54:16 <jwendell> :)
Oct 16 11:54:51 <jwendell> which file/target is invoked when I run 'make package/install'? couldn't find it so far...
Oct 16 11:55:11 <jwendell> I just found '$(curdir)/install:=$(curdir)/install-cleanup' in package/Makefile
Oct 16 12:20:57 <written_tabletpc> blogic: rebuilding the toolchain isn't enough :-(
Oct 16 12:25:30 <luka> written_tabletpc: i had the same problem
Oct 16 12:25:34 <luka> #
Oct 16 12:25:34 <luka> home/openwrt/openwrt/wndr3700/staging_dir/toolchain-mips_r2_
Oct 16 12:25:34 <luka> #
Oct 16 12:25:40 <luka> ups
Oct 16 12:25:51 <luka> rm /home/openwrt/openwrt/wndr3700/staging_dir/toolchain-mips_r2_* -rf
Oct 16 12:26:05 <written_tabletpc> so, at least it's not pebkac
Oct 16 12:26:06 <luka> and then run make
Oct 16 12:28:15 <blogic> jwendell: evil is relative
Oct 16 12:29:03 <jwendell> :)
Oct 16 12:30:22 <jwendell> I'm trying to understand it to know if it's possible to guarantee an order of installation of packages
Oct 16 12:30:29 <jow_laptop> it is not
Oct 16 12:30:49 <blogic> jwendell: you cant
Oct 16 12:31:04 <jwendell> how is that done currently? random order?
Oct 16 12:31:07 <jow_laptop> yes
Oct 16 12:31:09 <blogic> if you rely on an order of installation, then i assume one of your packages overwrites stuff from another
Oct 16 12:31:13 <blogic> and that is evil
Oct 16 12:31:18 <jwendell> yep
Oct 16 12:31:19 <jow_laptop> its undefined and implementation specific
Oct 16 12:31:32 <jow_laptop> and the behaviour can change without further notice
Oct 16 12:31:33 <blogic> and that evil is not relative but pure ;)
Oct 16 12:31:41 <blogic> jwendell: why you need install order ?
Oct 16 12:32:13 <jwendell> I'm creating packages that overwrite other packages stuff (config files)
Oct 16 12:32:41 <blogic> hrr hrr
Oct 16 12:32:44 <jwendell> for instance: I'm creating a complementar package for openvpn which overwrites /etc/config/openvpn
Oct 16 12:32:57 <jow_laptop> sigh, didn't we had this discussion already?
Oct 16 12:32:58 <jwendell> I did that through uci-defaults
Oct 16 12:32:59 <blogic> use a build variant inside openvpn
Oct 16 12:33:00 <jow_laptop> use uci-defaults
Oct 16 12:33:10 <blogic> or use uci.defaults
Oct 16 12:33:13 <jwendell> uci-defaults is bad for package updates
Oct 16 12:33:15 <jow_laptop> uci-defaults do not care about install order
Oct 16 12:33:19 <jow_laptop> they run on first boot
Oct 16 12:33:25 <jow_laptop> only if you implement it wrong
Oct 16 12:33:35 <jwendell> what happens if I update openvpn?
Oct 16 12:33:52 <jow_laptop>  /etc/config/openvpn will not change because its modified
Oct 16 12:35:00 <jwendell> but if I update my complementar openvpn package, it will put a file in /etc/uci-defaults that will run next boot and will mess things up
Oct 16 12:35:20 <jow_laptop> not if implemented correctly
Oct 16 12:35:21 <written_tabletpc> let's see if make toolchain/clean tools/clean
Oct 16 12:35:24 <written_tabletpc> is enough
Oct 16 12:36:10 <jow_laptop> anyhow, your solution becomes overcomplicated (we had this discussion too already)
Oct 16 12:36:14 <jwendell> I could write a script that will check for thousands of uci options and set them if unset, but this is insane
Oct 16 12:36:25 <jow_laptop> if you go as far as hacking the install phase to gurantee some install order
Oct 16 12:36:31 <jow_laptop> you can as well just modify the openvpn package
Oct 16 12:37:05 <jwendell> if I did this hack, I'd submit it for official inclusion
Oct 16 12:37:22 <jow_laptop> also having the same files in multiple packages makes opkg complain about conflicts
Oct 16 12:37:31 <jow_laptop> however you spin it, it is *wrong*
Oct 16 12:41:05 <mazilo> ping nbd
Oct 16 12:43:56 <written_tabletpc> still rebuilds tools & toolchain :-)
Oct 16 12:45:30 <written_tabletpc> let's see how far the build goes
Oct 16 12:45:34 <written_tabletpc> this time
Oct 16 12:46:31 <mazilo> nbd: should the $(TOOLCHAIN_DIR)/include/*.h files be included/packaged for uClibc in the package/toolchain/Makefile ?
Oct 16 12:49:21 <jow_laptop> we never package include files
Oct 16 12:51:17 <mazilo> jow_laptop: IC.
Oct 16 13:07:27 <written_tabletpc> okay. the toolchain is definatly broken
Oct 16 13:07:51 <written_tabletpc> i tried with tools/clean & toolchain/clean
Oct 16 13:08:07 <written_tabletpc> blogic: ^
Oct 16 13:09:00 <written_tabletpc> what's now?
Oct 16 13:14:06 <written_tabletpc> i'd like to avoid a distclean
Oct 16 13:16:56 <written_tabletpc> blogic: ?
Oct 16 13:17:11 <blogic> yes
Oct 16 13:17:13 <blogic> hi
Oct 16 13:17:57 <written_tabletpc> ?
Oct 16 13:18:24 <blogic> you dont want to distclen
Oct 16 13:18:53 <written_tabletpc> right
Oct 16 16:28:42 <jwendell> guys, in uci.sh, uci_load(), you run 'uci export $PACKAGE' and then runs 'eval' on its output
Oct 16 16:28:51 <jwendell> what's the logic behind this?
Oct 16 16:29:27 <jow_laptop> the logic behind it is to run uci export and run eval on the output
Oct 16 16:29:52 <jwendell> hehe
Oct 16 16:29:53 <blogic> :-)
Oct 16 16:30:58 <jwendell> it doesn't make sense to me, once it will evaluate things like 'package system config system 'cfg02e48a' option timezone 'UTC'....
Oct 16 16:31:10 <jow_laptop> well
Oct 16 16:31:35 <jow_laptop> before it does that it declares a procedure package(), a procedure config() and a procedure option()
Oct 16 16:31:47 <jwendell> ahhhhhhhhhhhhhhhhhh
Oct 16 16:31:47 <jow_laptop> the pattern used is callback based parsing
Oct 16 16:46:58 <russell--> dev.openwrt.org is timing out: "504 Gateway Time-out
Oct 16 16:47:06 <russell--> nginx"
Oct 16 16:51:41 <russell--> lol, just reading scrollback, /me likes blogic's "jwendell: you cant", assumes he meant in response to "I'm trying to understand [openwrt makefiles]"
Oct 16 16:52:57 <blogic> no i meant "14:30 < jwendell> I'm trying to understand it to know if it's possible to guarantee an order of installation of packages
Oct 16 16:53:16 <blogic> i related to the last bit
Oct 16 16:53:17 <russell--> my interpretation if funnier! :-)
Oct 16 16:53:22 <russell--> is*
Oct 16 16:53:26 <blogic> well
Oct 16 16:58:19 <jwendell> :)
Oct 16 17:02:22 <jwendell> it's so good to see the patch queue moving on... :)
Oct 16 17:02:44 <jwendell> I miss the cia-vc robot here
Oct 16 17:17:57 <jwendell> how to add an value as the first option in a list?
Oct 16 17:18:15 <jwendell> add_list appends, I'd like to prepend
Oct 16 17:21:03 <jwendell> in uci I mean hehe
Oct 16 17:41:16 <russell--> x86-generic (no select-all-packages) builds, x86-generic with select-all-packages dies in target/linux/compile (cleaning doesn't help) ... not sure if it's specific to net45xx, but that's what i'm building
Oct 16 17:59:13 <russell--> yeah, all x86-generic
Oct 16 17:59:41 <russell--> (or at least x86-generic-generic does the same thing as x86-generic-net45xx)
Oct 16 22:02:17 <russell--> i would register a ticket about x86-generic, but ...
Oct 16 23:04:54 <written_tabletpc> again 502 -bad gateway on dev.openwrt.Orff :-(
Oct 16 23:06:03 <written_tabletpc> i think you never get it Fixer...
Oct 16 23:06:40 <written_tabletpc> s/Fixer/fixed/
**** ENDING LOGGING AT Wed Oct 17 03:00:01 2012