13:13 < ths> fakeroot is a wrapper lib which catches privileged accesses.
13:13 < `Kumba|work> sounds like the sandbox we use in portage
13:14 < ths> That way you can build a tarball with proper root permissions and device files as non-root.
13:14 < `Kumba|work> sandbox has never worked for us on mips, though
13:14 < Skylark> Ah. A nice thing! I think I like the idea.
13:14 < ths> It's the standard way to build Debian packages.
13:14 < ths> `Kumba|work: What failed?
13:16 < `Kumba|work> ths: don't recall exactly.  The one report we had was in building ncurses, tic would run
                     out of memory
13:16 < ths> Ah.
13:17 < ths> The ldso bug triggered by fakeroot causes the linker to recursively resolve the same symbol
             until it ate all the stack space. Then segfault.
13:18  * Skylark just got 2kg of sodium thiosulphate :)
13:19 < ths> Skylark: Your daily terrorist training?
13:19 < `Kumba|work> ths: does kswapd handle stack space by chance?
13:19 < ths> `Kumba|work: It should. It's just paged memory.
13:20 < Skylark> ths: It is interesting. Extremely easy to melt (48 degC) and just as easy to overcool (even
                 to 20 degC) without solidifying. The resulting solidification after you break the symmetry
                 (jostle it, add something porous, whatever) is fast and heats the thing up nicely.
13:21 < `Kumba|work> hmm, it's possible the fakeroot problem might be related.  From what I've understood
                     about libsandbox, is it intercepts all file access calls so any package trying to touch
                     a file not explicitly allowed in the ebuild raises a sandbox violation.
13:22 < Skylark> ths: So, with an out-of-order version, it has the same benefits a normal out-of-order
                 superscalar has, except it is simpler to design and should be clockable higher. I think it
                 might offset the penalty of doing rl/rs.
13:22 < ths> `Kumba|work: Fakeroot just caches the permissions.
13:24 < ths> Skylark: I still don't see how you avoid hazards.
13:24 < ths> `Kumba|work: But the bug is generic for all wrapper libraries, so libsandbox will fail the same
             way as fakeroot, I guess.
13:25 < Skylark> ths: Well, each of the cores has normal interlocks.
13:26 < `Kumba|work> ths: interesting.  I'll have to make note of that in our bug database for when one of
                     our portage devs gets around to taking a look at it again