[07:28] <dholbach> good morning
[11:08] <Mithrandir> StevenK: did you get anywhere with the EINVAL from mount when trying to mount squashfs?
[11:10] <StevenK> Mithrandir: Yup
[11:10] <StevenK> Mithrandir: New mksquashfs using LZMA, and a kernel that didn't support it
[11:11] <StevenK> Mithrandir: See my mail to mobile
[11:13] <Mithrandir> StevenK: d'oh, ok thanks.  That's ever so slightly unfortunate.
[11:14] <StevenK> Mithrandir: Indeed.
[11:14] <StevenK> Mithrandir: I wish the news were different
[11:14] <Mithrandir> you mailed the list and said you stumbled on an unionfs bug?
[11:15] <StevenK> Right. I sent photos of the "kernel BUG" message to Amit and he said, "Argh! unionfs!"
[11:15] <Mithrandir> and that was with .24?
[11:15] <StevenK> That was with the kernel you built
[11:16] <Mithrandir> does the kernel even support lzma yet?
[11:16] <mjg59> No
[11:16] <mjg59> There's an external patch for it
[11:16] <StevenK> squashfs 3.3 says it does, and pkl merged it in
[11:16] <mjg59> Orly? Excellent.
[11:16] <lool> StevenK: It seems you can pass -nolzma to newer mksquashfs for now  :-/
[11:16] <mjg59> If we use that for the livecd, we potentially get win
[11:17] <StevenK> lool: Right. I hacked m-i-c to do so
[11:17] <StevenK> mjg59: LZMA looks to be the default for squashfs-tools 3.3, so let's see what Alpha2 gives us
[11:30] <StevenK> Mithrandir: I was going to keep pitti happy and look at the promotion problems next
[11:31] <Mithrandir> StevenK: makes sense.  I'd liked us to have something we could give intel to start porting their graphics driver on, but it doesn't look too good then. :-/
[11:35] <StevenK> Mithrandir: I discovered when I maintained ALSA that I'm not smart enough to work on the kernel
[12:14] <amitk_> StevenK: Did pkl confirm that he merged LZMA support into LUM?
[12:15] <StevenK> He didn't confirm, but I read the changelog of 3.3, and that's what was pulled into l-u-m
[12:15] <amitk_> StevenK: and you used the updated LUM too, right?
[12:31] <amitk_> StevenK: ^^^
[12:34] <StevenK> I used 2.2
[13:59] <StevenK> Mithrandir: What do you think about the latest e-mail on u-m?
[14:04] <Mithrandir> the one about UME X startup?  Just replied.
[14:04] <StevenK> Kay
[15:12] <smagoun> Is there a good way to add a diversion (via dpkg-divert) to a package that's already installed? The moblin guys and I have both run into problems because diversions are only supposed to happen on install, not upgrade.
[15:14] <smagoun> For example: A is installed. I want to add a file F to A'. F is already installed as package B. I want users upgrading from A to A' to get the the A' version of F.
[15:21] <smagoun> Mithrandir: lool StevenK  ^^^^^
[15:24] <lool> smagoun: You should be able to call dpkg-divert during upgrade in the very same way
[15:24] <lool> Make sure you test properly when to call it and think of all cases where you should remove it
[15:25] <lool> smagoun: How is your current problematic code looking like?
[15:25] <smagoun> lool: That's what I hoped, but this debian policy manual indicates I shouldn't check for upgrade: http://www.debian.org/doc/debian-policy/ap-pkg-diversions.html
[15:25] <lool> smagoun: Oh right, I understand the limitation now
[15:26] <smagoun> lool: What I'm interested in right now is moblin's xf86-video-psb driver, which blindly copies a library on top of another one becuase diversion doesn't work in the upgrade case: http://moblin.org/repos/projects/xf86-video-psb.git/
[15:26] <lool> smagoun: What the policy page says is that you should only call dpkg-divert once
[15:27] <lool> smagoun: If you simply call dpkg-divert without checking for $1 = install, it will run for upgrade, and other cases
[15:27] <smagoun> lool: right. So presumably I could check to see if the diversion's already been done by dpkg-diver --list?
[15:27] <lool> smagoun: The best thing to do is this indeed
[15:28] <lool> I'm checking whether I can hand a good example
[15:28] <smagoun> Ok, I think I understand. Since I'm only ever installing or upgrading that particular package version once, it's ok to divert in both the install and upgrade cases and I can skip the '$1 = install'
[15:29] <lool> smagoun: But basically your preinst should be something among the lines of: if install or upgrade and diversion_not_installed; then divert; fi
[15:29] <lool> if ( install or upgrade ) and diversion_not_installed
[15:30] <lool> Interestingly, I can only find examples contradicting policy :)
[15:30] <lool> if [ "$1" = install ] || [ "$1" = upgrade ]
[15:30] <lool> then
[15:30] <lool>     dpkg-divert --add --package perl-doc --rename \
[15:30] <lool>         --divert /usr/bin/perldoc.stub /usr/bin/perldoc
[15:32] <smagoun> lool: thanks. I'll try that approach.
[16:53] <Mithrandir> StevenK: if you're still awake, prod me, I have a kernel I'd like you to try on your CB.
[17:59] <lool> How does one rebuild the amd64 kernel image?
[17:59] <lool> (From the git tree)
[17:59] <lool> StevenK: You had the command line for lpia?  O:-)
[18:08] <Mithrandir> lool: debuild -e CONCURRENCY_LEVEL=13 -e PATH=/usr/lib/ccache:$PATH -uc -us -B is what I used
[18:09] <lool> Thanks!
[18:11] <Mithrandir> that concurrency level was fine with an 8 core machine, adjust as appropriate for your setup