**** BEGIN LOGGING AT Fri May 08 02:59:59 2015
May 08 07:36:02 <mckoan> good morning
May 08 08:29:22 <bluelightning> morning all
May 08 08:41:13 <mckoan> hi bluelightning 
May 08 08:43:08 <bluelightning> hi mckoan
May 08 09:58:59 <pompomJuice> koen: I need your help to solve my sftp problem.
May 08 09:59:29 <pompomJuice> for some reason if I build a debug image it works, but a release build causes the connection to be dropped
May 08 10:00:10 <pompomJuice> when i check the diff in the logs, all I can see is a massive spam by systemd that complains about a dosen or so dependency jobs that it cannot add because they are masked...
May 08 10:01:07 <pompomJuice> have you seen this?
May 08 10:02:04 <pompomJuice> actually... ignore the masked deps.
May 08 10:02:30 <pompomJuice> What I see is on release build I get a : pam_unix(sshd:session): session closed for user root
May 08 10:02:52 <pompomJuice> I have sshd_config configured to accept root etc.
May 08 10:15:32 <pompomJuice> I have diffed pam.d between debug and release
May 08 10:15:35 <pompomJuice> they are the same
May 08 10:26:28 <koen> and ssh*.conf?
May 08 10:26:47 <pompomJuice> ssh_config is the same
May 08 10:26:56 <pompomJuice> sshd_config differs with one line
May 08 10:27:31 <pompomJuice> the one that accepts root: PermitRootLogin yes
May 08 10:27:34 <pompomJuice> however
May 08 10:27:54 <pompomJuice> I have done a diff between the logs between debug and release
May 08 10:28:09 <pompomJuice> and what I see is a difference that I cannot understand
May 08 10:28:41 <pompomJuice> on the working side the log states:  Accepted keyboard-interactive
May 08 10:29:23 <pompomJuice> on the broken side I get 2 lines before that line: Connection from 10.0.0.83 port 54135 (AND) Postponed keyboard-interactive fo
May 08 10:29:37 <pompomJuice> for some reason there is a postpone 
May 08 10:31:50 <pompomJuice> actually ignore everything I just said... the one was in DEBUG output mode
May 08 10:33:55 <pompomJuice> new log diff:
May 08 10:34:01 <pompomJuice> with debug level on both sides
May 08 10:34:19 <pompomJuice> They are both the same, but on the broken side I get the following extra output: http://hastebin.com/acotefoyud.css
May 08 10:34:32 <pompomJuice> basically it states that the remote host closed the connection
May 08 10:34:49 <pompomJuice> filezilla again reports: Received unexpected end-of-file from SFTP server
May 08 10:34:52 <bluelightning> pompomJuice: are you logging in as root with this connection?
May 08 10:34:57 <pompomJuice> yes
May 08 10:35:07 <pompomJuice> I have tried creating a user and trying that
May 08 10:35:09 <pompomJuice> same issue
May 08 10:35:55 <pompomJuice> The only difference between the builds is a RELEASE_MODE=release instead of debug
May 08 10:36:08 <pompomJuice> and this RELEASE_MODE is a angstrom thing if I understand correctly
May 08 10:36:25 <bluelightning> it would be yes
May 08 10:36:37 <pompomJuice> Also: I have installed packagegroup-core-full-cmdline
May 08 10:36:49 <pompomJuice> but it was happening without that packagegroup as well
May 08 10:38:26 <pompomJuice> I thought the problem might have been this: http://blog.droidzone.in/2013/03/07/cannot-sftp-can-ssh-and-scp/
May 08 10:38:36 <pompomJuice> but from what I can see it is not that
May 08 10:39:23 <pompomJuice> so now I am stuck
May 08 10:40:15 <pompomJuice> and I have to go ti a meeting now... thanks for listening... any ideas would be apreciated
May 08 11:22:59 <Noor> hi all
May 08 11:23:56 <Noor> we are seeing that pkg-config package is not behaving correctly. it is not adding space in include search path for one of the packages
May 08 11:24:03 <Noor> did anyone see this behaviour
May 08 14:54:44 <Noor> pkg-config new version have some issues I think
May 08 14:55:07 <kergoth> what does 'not adding space' mean?
May 08 14:55:21 <Noor> while buildin gst-fsl-plugin it does not give space while searching for the path for a library
May 08 14:55:43 <kergoth> ?
May 08 14:55:49 <kergoth> what does "give space" mean?
May 08 14:55:51 <kergoth> space where?
May 08 14:55:53 <kergoth> for what purpose?
May 08 14:55:55 <Noor> kergoth: give me a sec
May 08 14:56:01 <kergoth> did you manually run the pkg-config command in devshell?
May 08 14:56:25 <Noor> kergoth: yeah
May 08 14:56:27 <Noor> FSL_MM_CORE_CFLAGS1=`pkg-config --cflags libfslaudiocodec 2>/dev/null`
May 08 14:56:38 <Noor> and 
May 08 14:56:39 <Noor> FSL_MM_CORE_CFLAGS1+=`pkg-config --cflags libfslvideocodec 2>/dev/null`
May 08 14:57:06 <kergoth> why would it both set and add the same flags twice?
May 08 14:57:12 <Noor> results a space in between two paths if I use winter release pkg-config
May 08 14:57:32 <kergoth> what do you mean by two paths?
May 08 14:57:34 <Noor> its not the same flag
May 08 14:57:36 <kergoth> —cflags returns cflags, not a path
May 08 14:57:42 <Noor> in make file its different
May 08 14:57:43 <kergoth> oh, right, misread
May 08 14:58:00 <kergoth> it's up to make to add the separator when += is used, last i checked
May 08 14:58:05 <Noor> yeah its different here as well
May 08 14:58:09 <kergoth> i don't see what pkg-config has to do with that at all
May 08 14:58:43 <Noor> when I run with new pkg-config here is the result of echo FSL_MM_CORE_CFLAGS1
May 08 15:00:00 <Noor> -I/home/noor/SB/release_2015.12/build_mx6q/tmp/sysroots/mx6q/usr/include/imx-mm/audio-codec-I/home/noor/SB/release_2015.12/build_mx6q/tmp/sysroots/mx6q/usr/include/imx-mm/video-codec
May 08 15:00:06 <Noor> this is in devshell
May 08 15:00:16 <Noor> if I change the pkg-config in sysroot
May 08 15:00:19 <Noor> i get this
May 08 15:00:55 <kergoth> as i said before, the space separator between the two subshells is not added by pkg-config. compare the actual output of the pkg-config —cflags command in devshell between the two versions
May 08 15:01:05 <kergoth> afaik, anyway
May 08 15:01:12 <kergoth> i also haven't had much caffeine yet
May 08 15:01:14 <Noor> -I/home/noor/SB/release_2015.12/build_mx6q/tmp/sysroots/mx6q/usr/include/imx-mm/audio-codec -I/home/noor/SB/release_2015.12/build_mx6q/tmp/sysroots/mx6q/usr/include/imx-mm/video-codec
May 08 15:01:28 <Noor> see there is space in second out between two paths
May 08 15:01:54 <Noor> this is done in devshell
May 08 15:02:00 <Noor> kergoth: ^
May 08 15:02:06 <kergoth> ?
May 08 15:02:12 <kergoth> you just said you're echoing a variable, not running pkg-config
May 08 15:02:51 * kergoth checks something
May 08 15:03:44 <Noor> kergoth: i first assigned it to varaible then echo the value of variable
May 08 15:03:47 <Noor> in devshell
May 08 15:04:25 <kergoth> that's not useful
May 08 15:04:33 <kergoth> again, compare the exact output of the indiviudal pkg-config commands as run at a shell
May 08 15:04:39 <kergoth> but regardless, i think i found the issue
May 08 15:04:45 <Noor> but just changing the pkg-config in native sysroot adds the space
May 08 15:05:03 <kergoth> pkg-config used to leave a trailing space in —cflags output, but it was not intentional, it was due to how it did its concatenation
May 08 15:05:06 <Noor> :D
May 08 15:05:07 <kergoth> that was fixed in f26a505826acec6b1e1af8d1efdf959913421c3a upstream
May 08 15:05:15 <kergoth> relying on that behavior was not good
May 08 15:06:11 * Noor wishes that he becomes quick like kergoth :)
May 08 15:06:14 <kergoth> what recipe is this?
May 08 15:06:32 <kergoth> heh, it wasn't too tough, i just cloned the pkg-config git repo and looked through the git history for anything obvious
May 08 15:06:39 <kergoth> it was 10 commits ago or something :)
May 08 15:07:07 <Noor> gst-fsl-plugin
May 08 15:08:02 * Noor dont want to explain how much time he put to identify where the problem is :P
May 08 15:08:39 <Noor> I was wondering why this only showed up in only this recipe
May 08 15:08:58 <kergoth> its possible it could affect others, depends on who all was relying on that behavior
May 08 15:09:08 <kergoth> generally i think most include a space when referencing cflags components
May 08 15:09:15 <kergoth> e.g. CFLAGS = $(FOO_CFLAGS) $(BAR_CFLAGS)
May 08 15:09:43 <kergoth> hmm
May 08 15:09:58 <Noor> we have this
May 08 15:10:00 <Noor> FSL_MM_CORE_CFLAGS=`$PKG_CONFIG --cflags libfslaudiocodec 2>/dev/null`
May 08 15:10:00 <Noor> FSL_MM_CORE_CFLAGS+=`$PKG_CONFIG --cflags libfslvideocodec 2>/dev/null`
May 08 15:10:00 <Noor> FSL_MM_CORE_CFLAGS+=`$PKG_CONFIG --cflags libfslparser 2>/dev/null`
May 08 15:10:00 <Noor> FSL_MM_CORE_CFLAGS+=`$PKG_CONFIG --cflags libfslvpuwrap 2>/dev/null`
May 08 15:10:00 <Noor> AC_SUBST(FSL_MM_CORE_CFLAGS)
May 08 15:10:13 <Noor> in configure.ac
May 08 15:10:53 <kergoth> ah. shell +=, which doesn't add a separator. also that'd break if the —cflags output contains spaces
May 08 15:11:10 <kergoth> should 1) quote those values, 2) add the separators in the quoted value
May 08 15:11:57 <kergoth> i was thinking those lines were from a makefile :) my mistake
May 08 15:12:19 * kergoth would have taken less time to help if he was more awake, most likely
May 08 15:12:44 <Noor> :)
May 08 15:13:11 * kergoth wonders if anyone's done a world build since the pkg-config upgrade
May 08 15:49:51 <realBigFoot> khem`, thanks ! I got it fixed 
May 08 15:50:11 <khem`> np
May 08 16:11:39 <fray> I can't remember the magic, but someone asked me how to set an environment variable outside of bitbake, and then make sure it filters into the shell calls that are made with tasks..
May 08 16:11:45 <fray> can anyone remind me of the BB variable?
May 08 16:14:12 <fray> BB_PRESERVE_ENV ?
May 08 16:14:17 <kergoth> BB_ENV_EXTRAWHITE="$BB_ENV_EXTRAWHITE FOO" bitbake bar
May 08 16:14:38 <kergoth> hmm, i think you'd have to re-export those using extrawhite though
May 08 16:15:05 <fray> We've already tried setting it in extrawhite.. it's there for things like do_fetch, but for other tasks it's not there
May 08 16:15:12 <fray> maybe it's the way the other tasks are being invoked?
May 08 16:15:48 <kergoth> i think vars in extrawhite get into the metadata, but not exported into tasks unless you add 'export FOO' to the metadata
May 08 16:15:52 <kergoth> hmm
May 08 16:16:01 <fray> thats what I'm seeing..
May 08 16:17:25 <kergoth> BB_PRESERVE_ENV lets the entire env get into the metadata, which isn't ideal, and sitll might not re-export vars :)
May 08 16:17:35 <fray> ahh so thats just an on/off..
May 08 16:17:44 <fray> ya, this is one specific variable pointing to the curl CA dir
May 08 16:18:09 <kergoth> also remember that exporting a variable will affect all task checksums
May 08 16:18:14 <kergoth> all shell tasks, at least
May 08 16:18:38 <fray> thats why you have to add it to the hash exclude one
May 08 16:21:44 <fray> think I found it..
May 08 16:22:06 <fray> the runfetchcmd has a limited set of things you can pass in..  and the curl CA path isn't one of them
May 08 16:22:49 <fray> Hmm.. or maybe the symptoms I see are unrelated.. argh.. anyway.. thanks
May 08 16:28:45 <kergoth> ugh, forgot about fetch's filtering
**** ENDING LOGGING AT Sat May 09 02:59:58 2015