[12:29] <ogra_> cjwatson, hmm, i'm lost in debian-cd ... seemingly the environment doesnt get reset if i run cron.daily-preinstalled ... CONF.sh sets PREINSTALLED_IMAGE_FILESYSTEM to ext3 ... later build_all.sh sets it to rootf.tar.gz if there is an ac100 build, seems that var never gets reset if subsequent subarches are built, any suggestion where i can reset it reliably in that loop ?
[12:32] <cjwatson> ugh, nasty!
[12:32] <ogra_> yes
[12:32] <cjwatson> definitely don't pull stunts like that in build_all.sh
[12:32] <ogra_> hmm, where else could i change it
[12:33] <cjwatson> you only use the value in one make target; just decide its value there instead
[12:33] <ogra_> in the Makefile you mean ?
[12:33] <cjwatson> yes
[12:33] <ogra_> hmm, k
[12:33] <cjwatson> I'm also fairly unconvinced about the existence of a variable which can take values "ext3" and "rootfs.tar.gz", but ...
[12:34] <ogra_> well, its the same image type (preinstalled), the implementation just varies based on HW capabilities
[12:35] <ogra_> we somehow need to represent that
[12:41] <cjwatson> sure, what I mean is that "ext3" and "rootfs.tar.gz" hardly seem to be of the same conceptual type
[12:41] <ogra_> (and it didnt really feel right to add a subarch check to Makefile)
[12:41] <cjwatson> I would add a subarch check to Makefile if I were you
[12:41] <ogra_> they are both preinstalled
[12:42] <ogra_> but ext3 ends up in a two partition image while the tarball is a two file image
[12:42] <ogra_> enforced by the fact that we are targeting an anrdoid device where you need to flash the kernel/bootimg first
[12:43] <cjwatson> I don't want to have a long debate about it, it just feels weird
[12:43] <ogra_> well, i could have invented a completely new image type but we use the same process for both of them
[12:44] <ogra_> i will revisit that in P (and also try to get all that env variable crap a bit cleaner into a config file so you dont need to chack code for it)
[12:44] <ogra_> *hack
[13:04]  * ogra_ tries another mx5 build with the changes applied
[13:26] <ogra_> grrr
[13:27]  * ogra_ hates the fact that shell in Makefiles effectively turns into perl and scatters some semicolons over the place
[15:12] <janimo> Can someone please promote the u-boot-linaro-mx53loco binary to main? It's source is already there
[15:17] <infinity> janimo: I can do it.
[15:17] <janimo> infinity, thanks
[15:42] <ogra_> cjwatson, hmm, exporting in Makefile did definitely not work
[15:43] <cjwatson> ogra_: diff please
[15:45] <ogra_> http://paste.ubuntu.com/688412/
[15:45] <ogra_> results in:
[15:45] <ogra_> + [ ac100 = ac100 ]
[15:45] <ogra_> + export PREINSTALLED_IMAGE_FILESYSTEM=rootfs.tar.gz
[15:45] <ogra_> + export CDIMAGE_COMPRESS=
[15:45] <ogra_> + [ ! -e /srv/cdimage.ubuntu.com/scratch/ubuntu/daily-preinstalled/preinstalled/armel+ac100.ext3 ]
[15:45] <ogra_> + echo No filesystem for armel+ac100!
[15:46] <ogra_> No filesystem for armel+ac100!
[15:46] <ogra_> (the ext3 in the check uses PREINSTALLED_IMAGE_FILESYSTEM for the suffix)
[15:48] <cjwatson> you handle CDIMAGE_COMPRESS in build_all.sh so you still have to set it there
[15:48] <ogra_> exporting doesnt get handed over ?
[15:48] <cjwatson> not to the parent process!
[15:48] <ogra_> well, it fails on the wrong file suffix though, i could even use the compressing code and not gzip in live-build
[15:49] <cjwatson> export means that the environment variable is passed on to child processes
[15:49] <cjwatson> there is no way in Unix to non-consensually modify the environment of a parent process
[15:49] <ogra_> but i need it to recognize the tarball
[15:49] <cjwatson> stop.  wait.
[15:49]  * ogra_ waits
[15:49] <cjwatson> the problem is that $(PREINSTALLED_IMAGE_FILESYSTEM) is a *make* variable not a shell variable
[15:50] <ogra_> so i dont export but just set it ?
[15:51] <cjwatson> you could use $$PREINSTALLED_IMAGE_FILESYSTEM instead and that would avoid the problem, but I think this would be clearer:
[15:51] <cjwatson> http://paste.ubuntu.com/688417/
[15:51] <cjwatson> export/set> not the point
[15:51] <cjwatson> although it is true that you do not need to export it because it isn't needed in child processes; but the point was that you were referring to the variable using the syntax from the wrong language :-)
[15:51] <ogra_> ah, cool, now i get it
[15:52] <ogra_> yeah, mixed language is a pain (see above i *always* forget the semicolons)
[15:56] <cjwatson> that's why I think my version is clearer, because you're doing more in make and less in shell
[15:56] <cjwatson> and don't have to worry about whether you're exporting the variable in the same subshell as where it's referenced
[16:00] <ogra_> yeah, applied that way and running a testbuild already
[16:02] <ogra_> though i find it a bit sad that i have subarch special casing in two places now :)
[16:03] <ogra_> yay, looks good, its downloading the tarball
[16:54] <hggdh> cjwatson: can you please have a peek on bug 849172? Then please tell me what else you might need
[16:55] <cjwatson> hggdh: already fixed
[16:56] <cjwatson> dudped
[16:56] <cjwatson> *duped
[16:56] <hggdh> cjwatson: you, sir, are my hero :-)
[16:57] <cjwatson> mvo did half of it
[16:57] <hggdh> mvo: you are also my hero
[16:57] <cjwatson> eglibc fix just uploaded, so that'll take a while to build
[16:57]  * hggdh is quite free with hero assignments
[17:16] <skaet> cjwatson, infinity -  would like to do some manual test image builds on antimony this afternoon.   Any concerns?
[17:18] <cjwatson> skaet: no.  DEBUG=1 will cause it not to publish anything and to dump all the output to your terminal.  CDIMAGE_NOPUBLISH=1 will simply cause it not to publish anything.
[17:19] <skaet> cjwatson,  thanks.
[17:20] <lamont> skaet: cjwatson: about to cut syncproxy over, btw
[17:22] <skaet> lamont,  approx. timing?
[17:22] <lamont> approximately 20 seconds from now
[17:23] <lamont> or 2 seconds ago, either way
[17:30] <skaet> heh
[17:42] <infinity> Yay new syncproxy!
[18:42] <mvo> hggdh: I always wanted to be a hero (and a lumberjack too) \o/
[18:50] <hggdh> mvo: there you go, my pleasure. And, BTW, thank you :-)
[18:52] <mvo> yw!
[18:53] <micahg> mvo is a hero for uploading the fix for the weird apt bug that was breaking alternate build dependencies more so than normal
[19:16] <hggdh> micahg: indeed :-)