On Mon, Sep 26, 2011 at 10:54:16PM -0500, Dan McGee wrote: > On Mon, Sep 26, 2011 at 9:07 PM, Dave Reisner <d...@falconindy.com> wrote: > > On Mon, Sep 26, 2011 at 08:54:28PM -0500, Dan McGee wrote: > >> On Mon, Sep 26, 2011 at 8:22 PM, Dave Reisner <d...@falconindy.com> wrote: > >> > This allows a user to provide their own binaries and override anything > >> > that might be provided by busybox. > >> > > >> > Signed-off-by: Dave Reisner <dreis...@archlinux.org> > >> > --- > >> > init | 2 +- > >> > 1 files changed, 1 insertions(+), 1 deletions(-) > >> > > >> > diff --git a/init b/init > >> > index ff4917a..e8f70a1 100644 > >> > --- a/init > >> > +++ b/init > >> > @@ -1,6 +1,6 @@ > >> > #!/bin/busybox ash > >> > # Install busybox's applets as symlinks > >> > -PATH=/usr/sbin:/usr/bin:/sbin:/bin > >> > +PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin > >> This is totally nitpicking, but we have this order by default in > >> /etc/profile: > >> PATH="/usr/local/bin:/usr/bin:/bin:/usr/local/sbin:/usr/sbin:/sbin" > >> > >> Shouldn't we try to be consistent? Or at least figure out what normal is? > > > > Maybe it's another case of not being clear enough in the commit msg. > > We're sort of peeing all over /usr/bin on the busybox --install, so we > > need to be able to let the user have some control. We do that by letting > > them override binaries in /usr/local/bin and putting this bin above the > > rest. An extremely rare use case, but a use case none the less. I could, > > also, be talked out of this patch along with the busybox on build rather > > than run if it seems insane. > I think you misinterpreted me. Note the order difference in yours vs. > what I pasted from the stock Arch /etc/profile. Of course /local/ has > to come first, but I was simply pointing out the "all /bin/ first" vs > "all /local/ first" and other odd ordering choices. >
Yep, I see it now. This is all somewhat suspect now anyways, and I probably can throw this out, as it was "needed" for the patchwork that installs busybox in the build, not the boot. > > mkinitcpio is sort of longer overdue for some documentation regarding > > the rules and regulations of hook writing. Merging this would sort of > > reinforce that point. > > > >> > >> Of course, from a CentOS box... > >> $ echo $PATH > >> /usr/sbin:/sbin:/usr/local/bin:/bin:/usr/bin:/home/dan/bin > >> > >> -Dan > > > > Yeah, CentOS and RHEL both leave the /sbin's out of non root profiles. > > Not a concern here, for obvious reasons. > Also not what I meant to point out- here, /sbin/ takes preference over > /bin/, always, including local. So another totally different take. Also, I apparently can't read properly at 11pm. At least what I said about RHEL is true, but it clearly has no bearing here... > And for some more orderings... > $ pacman -Qlq initscripts | xargs grep PATH= > /etc/rc.d/functions:export > PATH="/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin" > /sbin/rc.d: ENV=("PATH=/bin:/usr/bin:/sbin:/usr/sbin" > Yay!!! We probably had a reason for this, but I believe ordering the /usr/local bins first has caused us pain with a user at least once. d