Hi, Alle mercoledì 3 luglio 2013, Roger Leigh ha scritto: > On Tue, Jul 02, 2013 at 01:32:07PM +0200, Mattias Ellert wrote: > > tis 2013-07-02 klockan 11:43 +0100 skrev Roger Leigh: > > > What's different about /tmp inside the chroot? Is it bind > > > mounted (or the hurd equivalent) inside the chroot? What is the > > > configuration you are using for this chroot; did you make any > > > particular customisations? > > > > If you want to know details about how the schroot is configured on > > the porterbox, only the administrators of that machine can tell > > you. I was using the porterbox to figure out why one of my > > packages fails to build in the Debian build system for gnu/hurd. I > > did not do any tweaking of the configuration myself, but I have no > > idea if the porterbox admin has done something. > > Could the hurd buildd people possibly take a look here? The difference here (at least on exodar' setup, but I guess it is the same on the buildds) is that /tmp in the chroot is done using the firmlink translator to the system /tmp, which creates a "link" much similar to a bind mount. Indeed, setting up a firmlink translator and trying the commands earlier mentioned in this bug (#714733) gives the same failure, while they work when the root directory of both source and destination is not under a firmlink-ed node. Though, I have not investigated further whether the problem is in firmlink itself, in the libraries it uses or somewhere else. -- Pino Toscano
Description: This is a digitally signed message part.