On 2010-6-22 06:56 , Jordan K. Hubbard wrote: > > On Jun 21, 2010, at 1:28 PM, Joshua Root wrote: > >> On 2010-6-22 06:16 , Jordan K. Hubbard wrote: >>> Depends on how you constructed the chroot - are you sure you union mounted >>> all of the various important filesystems like devfs and fdesc? >> >> Nope! :-) It's blb's code, and I haven't even come close to verifying >> that it's all doing the right thing. All that's clear is that what used >> to work on 10.5 no longer does on 10.6. >> >> Do you have an example of constructing the chroot correctly? > > It's not all that up-to-date, but there's always the script in > base/portmgr/packaging/buildall.sh - if you look at the prepchroot() > function, you'll see the union mounts required. > > Note that this script also does everything on an HFSX (case-sensitive HFS) > disk image with -shadow enabled, which means that all writes to it are > trapped and stored in the shadow file rather than modifying the dmg to which > I've copied all of the system files. A rather convenient little feature > that, without disk images, would have to be done some other way (the > interposing already suggested, or some clever set of union mounts over the > writable bits).
Hm, well, that seems to be what is being done here: <http://trac.macports.org/browser/contrib/mpab/mpab-functions#L318> ... which isn't too surprising considering MPAB is inspired by DPAB which was inspired by buildall.sh. :-) - Josh _______________________________________________ macports-dev mailing list [email protected] http://lists.macosforge.org/mailman/listinfo.cgi/macports-dev
