Pádraig Brady wrote: > Jim Meyering wrote: >>>From ce9ff50623ad010fb52c3a4bf18a232eb875137c Mon Sep 17 00:00:00 2001 >> From: Jim Meyering <meyer...@redhat.com> >> Date: Thu, 19 Nov 2009 10:13:22 +0100 >> Subject: [PATCH] tests: avoid spurious failures due to insecure directory in >> PATH >> >> These tests perform no PATH search, and used to simply delete PATH from >> the environment. However, that is not portable, as seen on Cygwin, >> where cygwin.dll must be resolvable via PATH when starting a sub-shell. >> With commit 0cc04241, we took the alternate approach of untaining the > > s/untaining/untainting/
Thanks! Fixed. > Off the top of my head the following might point > out dodgy $PATH entries to the reporters: > > echo $PATH | tr : '\n' | sed -n '/^\//!p' > find -H $( > ( > echo $PATH | tr : '\n' | xargs -n1 readlink -f > echo $PATH | tr : '\n' > ) | > sed -n ':s; p; s#/\{0,\}[^/]\{1,\}$##; /./b s' | > sort -u > ) -maxdepth 0 -perm -o=w > > Hmm, as we're adding /abs/path/to/coreutils/src to the start > of the $PATH, then I'm guessing that this issue is caused > by running from /tmp which is often o+w > Maybe we need to just remove the coreutils/src from > the $PATH for these tests? I can attest that building from a subdirectory of /tmp (which has o+t set) causes no problem. I do that all the time and it never fails. It's only if your path contains a directory that is o+w, or if any path-dir *ancestor* is o+w but without o+t.