[00:55] <psusi> so... there seem to be two different versions of libboost... 1.46 and 1.48... and 1.46 is the default... this is because the libboost devs don't understand that they aren't supposed to break ABI on minor revisions?
[00:55] <RAOF> Yes. And probably partially because it's moderately difficult to not break ABI in C++.
[00:56] <broder> boost uses minor revision numbers like everyone else uses major revision numbers
[00:56] <psusi> great
[00:57] <ajmitch> boost1.48 became default in debian a little late for it to happen in precise as well
[00:57] <psusi> I'm trying to build and package this nifty game I found called freeorion but it seems that it needs 1.48 of boost, but it also needs libogre, and libogre depends on libboost, which points to 1.46... and the -dev packages of libboost 1.46 and 1.48 are not coinstallable
[00:58] <psusi> so if I install libboost1.48-dev to make the game happy, it removes libboost1.46 and thus, libogre, which the game also needs...
[00:58] <psusi> so I guess I'm stuck between a rock and a hard place?
[01:02] <ajmitch> it does make it just a bit awkward
[01:04] <psusi> why does C++ make it harder to not break ABI?
[01:04] <ajmitch> are you trying to build this for precise or quantal?
[01:04] <psusi> precise
[01:06] <ajmitch> ScottK is the boost expert, I'm sure he'd be happy to help :)
[01:08]  * micahg thinks ajmitch found a way to make ScottK hide
[01:09] <ajmitch> micahg: he's fobbed boost off onto me before, I have to return the favour
[01:10] <psusi> so... in precise, libboost 1.48 is available, but 1.46 is still the default... so to fix this... a special -xxx build of libogre would need made to build against libboost1.48 right?  but quantal has moved to 1.48 as the default, so there it shouldn't be a problem?
[01:10] <ajmitch> quantal has 1.49 as default
[01:10] <ajmitch> just to be different
[01:11] <psusi> that should work... presuming that libogre has been rebuilt against it
[01:11] <ajmitch> http://people.canonical.com/~ubuntu-archive/transitions/boost1.49.html shows ogre as 'good'
[01:11] <ajmitch> so it should be rebuilt against 1.49
[01:11] <psusi> maybe I just need to upgrade to quantal
[01:12] <psusi> so why are the 1.46 and 1.48 not coinstallable?
[01:12] <psusi> I thought that was the point of creating packages with the version number in the name?
[01:13] <ajmitch> right, but they ship files in unversioned paths
[01:13] <ajmitch> & the same binaries, etc
[01:14] <psusi> when you add the version number to the name, aren't you supposed to add it to the paths of the binaries too?
[01:14] <psusi> I mean, what's the point of having two source packages if their binaries can't both be installed?
[01:14] <psusi> or was it that some of the binary packages are coinstallable, but not all of them?
[01:15] <psusi> hrm...
[01:15] <ajmitch> the library packages should be coinstallable, just not the -dev packages, which leads to this sort of mess
[01:15] <psusi> ok, that's what I thought... so why are the libs coinstallable, but not the -dev?
[01:16] <ajmitch> because libraries have the version in the SONAME
[01:16] <psusi> but the headers don't?
[01:16] <ajmitch> but not in the header paths for the -dev packages (and lots of build tools)
[01:17] <psusi> ahh, and libogre-dev depends on libboost-dev... argh
[01:17] <ajmitch> right, which means ogre shouldn't need any particular boost version
[01:18] <ajmitch> I think boost is set up this way because it would be a lot of work to make it all parallal installable
[01:18] <psusi> hrm... if libboost-dev were a provides: of both 1.46 and 1.48 then if I installed 1.48, I could install libogre, right?  but it looks like it isn't... libboost-dev is a binary package provided by the 1.46 sources, not 1.48
[01:19] <psusi> so when libogre specifies depends: liboost, it requires 1.46, not 1.48
[01:21]  * ajmitch should upgrade to quantal this weekend & see what breaks
[01:23]  * psusi needs to start using btrfs so that such upgrades can easily be reverted when the shit hits the fan
[01:37] <psusi> how the hell does g++ use that much ram?  I just saw a cc1plus process using 1.2g rss
[01:40]  * ajmitch blames templates
[01:40] <psusi> that's like blaming header files or macros
[01:42] <psusi> seems like you'd need like 10kloc of templates instantiated 10,000 different ways to do that
[01:42] <psusi> and even then...
[03:27] <ScottK> The fact that boost -dev packages aren't co-installable means related sets of packages need to be on the same version of boost.  It's not usually an issue.
[07:00] <dholbach> good morning
[08:08] <geser> good morning
[14:48] <\sh> moins