lh + upstart/rsyslog + karmic-updates

[Aymeric Mansoux]

Complex interaction:

 * rsyslog needs to be installed via hook because it tries to connect to upstart: OK
 * rsyslog and upstart gets updated only once in the chroot after its population because their updates are on a different mirror than the one used to populate the chroot: not a problem because there is a hook
 * the lh binary process calls lh chroot_sources again from a chroot producing the same issue, but there are not hooks possible here: build fails

Dirty workaround for now: disable the second lh chroot_sources install...
http://live.debian.net/gitweb?p=live-helper.git;a=blob;f=helpers/lh_binary;h=33ee5d6ee80d60e6c8a89d142c4db77a5a6b70bd;hb=HEAD#l43

[Aymeric Mansoux (aymeric)]

12:12 < aym3ric> ok so there is a chroot that is used to be the live system, then once it's finished, it gets copied and the image is built from this copied chroot using extra tools.
12:13 < hramrach> the copy is then nuked ynd you get to keep the chroot from which the image was built
12:13 < aym3ric> I see
12:13 < hramrach> they may be installed already but it depends on the packages you selected for teh live system
12:14 < hramrach> so to work in all cases theyu have to be (re)installed
12:14 < aym3ric> I still fail to understand why it goes through a lh apt_sources install, if it is a copy of a chroot that just went through this step :/
12:14 < hramrach> if you installed correct apt pinning your versions would likely persist
12:16 < hramrach> aym3ric: I don't know the exact details but the packages have to be installed so apt is et up for that
12:17 < aym3ric> but apt is already set up correctly no? again if the chroot is an exact copy, why redo all this?
12:18 < hramrach> it is not. There are two apt configurations (though the split is incomplete and buggy): one for building the chroot image and one for the resulting live system
12:19 < aym3ric> aaah
12:19 < aym3ric> the .chroot and .binary settings ?
12:19 < hramrach> yes
12:20 < aym3ric> ok so that start to make sense :)
12:20 < aym3ric> you can use indeed a different set of repos for the two stages
12:20 < hramrach> or perhaps the .binary are specifically for this chroot-copy step, not sure
12:20 < aym3ric> that would make sense
12:21 < aym3ric> which is why in my case it feels funny because all my chroot repos are also needed in the binary image
12:22 < aym3ric> so maybe it would also make sense to have a hook mechanism for the chroot copy?
12:22 < hramrach> I suspect there is
12:23 * aym3ric checks
12:23 < hramrach> I recall somebody talking about modifications to the .iso outside of the squashfs
12:23 < aym3ric> ah no this is a binary hook and yes it exists
12:24 < aym3ric> I am more thinking of another chroot hook that would happen specially for the chroot copy
12:24 < hramrach> there's not really that much to be gained - the chroot copy is just thrown away after build
12:27 < aym3ric> yes, but you see, in my very particular situation, I can only complete the installation of a certain package via a hook, so my initial chroot is all good, no rpoblem. But when the copy occurs and lh apt_sources install is called again, this packages is pulled again and hence needs again an extra help to install properly, hence the need for an extra hook here.