Hi Hartmut, Great that you’re working on this long-standing issue, that’ll make all of us happier! :-)
Hartmut Goebel <h.goe...@crazy-compilers.com> skribis: > This also closes bug #15890. Please use the same wording as in other commits, to simplify grepping and access: Fixes <http://bugs.gnu.org/15890>. > * nix/libstore/worker-protocol.hh (PROTOCOL_VERSION): Increment it. > * guix/store.scm (protocol-version): Increment it to the same value. > (set-build-options): Send uid and gid of evoking user. > * nix/libstore/globals.hh (Settings) Add clientUid and clientGid. > * nix/nix-daemon/nix-daemon.cc (performOp)[wopSetOptions] Read clientUid > and clientGid. > * nix/libstore/build.cc (DerivationGoal::deleteTmpDir): Change > ownership of build directory if it is kept. [...] > + (when (>= (nix-server-minor-version server) 16) > + ;; Send uid and gid of evoking user. If either of it is zero > (root), > + ;; send -1 wich means: do not change. > + (send (integer (or (getuid) -1)) (integer (or (getgid) -1)))) I think this can and should be avoided (the client could pass in any UID). In nix-daemon.cc:921, the UID and GID of the caller are retrieved using SO_PEERCRED. I think we should just pass them down in ‘processConnection’ and then probably in the ‘LocalStore’ constructor. WDYT? > + if (settings.clientUid != (uid_t) -1) { > + // FIXME: Is this the correct way to to it? Or it there a > better > + // way? > + Strings args; > + char ids[32]; > + if (settings.clientGid != (gid_t) -1) > + // both uid and gid are set, change owner and group > + snprintf(ids, 31, "%lu:%lu", > + (unsigned long int) settings.clientUid, > + (unsigned long int) settings.clientGid); > + else > + // only uid is set, change only the owner > + snprintf(ids, 31, "%lu", > + (unsigned long int) settings.clientUid); > + args.push_back("-R"); > + args.push_back(ids); > + args.push_back(tmpDir.c_str()); > + string signature = runProgram("chown", true, args); This should be done in C++, with code comparable to ‘_deletePath’ in util.cc. BTW, in Emacs I use: C-c . stroustrup RET. That roughly corresponds to the existing style. So, how does that sound? :-) Thanks, Ludo’.
