Question

I have a Mercurial repository set up on a Linux server, and some (but not all) users have permission to push to it. They connect to the repository over ssh.

These users are members of a unix group together. Below is the script I’m using to alter a repository to allow it to receive pushes from them.

Can this be improved? Are there unnecessary operations in here? Is anything bad style for a bash script?

#!/bin/bash                                                                     

if [[ $# -lt 2 ]]; then
    echo Usage: $0 directory groupname
    exit 1
fi

if ! chown -R :$2 $1; then
    echo chown failure
    exit 2
fi

if ! find $1/.hg -type d -exec chmod g+s {} \;; then
    echo chmod failure
    exit 3
fi

if ! find $1 -perm -u+r -exec chmod g+r {} \;; then
    echo chmod failure 2
    exit 4
fi

if ! find $1 -perm -u+w -exec chmod g+w {} \;; then
    echo chmod failure 3
    exit 5
fi

if ! find $1 -perm -u+x -exec chmod g+x {} \;; then
    echo chmod failure 4
    exit 6
fi

Answer 1

Running find with -exec launches a separate chown process for each file. You’ll get a lot less process thrash (and more speed) if you do:

find "$1"/.hg -type d -print0 | xargs chmod g+s
find "$1" -perm -u+r -print0 | xargs chmod g+r
find "$1" -perm -u+w -print0 | xargs chmod g+w
find "$1" -perm -u+x -print0 | xargs chmod g+x

As an aside, have you looked at Mercurial’s ACL Extension in combination with hg-ssh? So long as the only access is ssh, it does the same sort of thing.