On 3/24/21 9:20 PM, Allan McRae wrote:

On 22/3/21 1:14 pm, Eli Schwartz wrote:

...

If specified, this will be used no matter what. If not, then we check if sudo exists and use that, or else fall back on su.

Implements FS#32621

Signed-off-by: Eli Schwartz --- doc/makepkg.conf.5.asciidoc | 8 ++++++++ etc/makepkg.conf.in | 7 +++++++ scripts/makepkg.sh.in | 13 ++++++++++--- 3 files changed, 25 insertions(+), 3 deletions(-)

diff --git a/doc/makepkg.conf.5.asciidoc b/doc/makepkg.conf.5.asciidoc index 2c7a54dbf..398529158 100644 --- a/doc/makepkg.conf.5.asciidoc +++ b/doc/makepkg.conf.5.asciidoc @@ -278,6 +278,14 @@ Options `.tar.lzo`, `.tar.lrz`, `.tar.lz4`, `.tar.lz` and `.tar.Z`, or simply `.tar` to disable compression entirely.

+**PACMAN_AUTH=()**:: + Specify a command prefix for running pacman as root. If unset, makepkg will + check for the presence of sudo(8) and su(1) in turn, and try the first one + it finds. + + + If present, `%q` will be replaced with the shell-quoted form of the command + to run. Otherwise, the command to run is appended to the auth command.

I found "%q" a weird choice for the command when reading this, then got even more confused with the "printf '%q ' " in the code, which is a different %q! Would %c be better?

I forget why I picked it (but now it seems weird to me too), and don't much care what we use. Sure.

...

See Also diff --git a/etc/makepkg.conf.in b/etc/makepkg.conf.in index 43a69df66..fff5b8eb2 100644 --- a/etc/makepkg.conf.in +++ b/etc/makepkg.conf.in @@ -147,3 +147,10 @@ COMPRESSLZ=(lzip -c -f) # PKGEXT='@PKGEXT@' SRCEXT='@SRCEXT@' + +######################################################################### +# OTHER +######################################################################### +# +#-- Command used to run pacman as root, instead of trying sudo and su +PACMAN_AUTH=() diff --git a/scripts/makepkg.sh.in b/scripts/makepkg.sh.in index f4a2de7d4..a0cd1a4fb 100644 --- a/scripts/makepkg.sh.in +++ b/scripts/makepkg.sh.in @@ -225,15 +225,22 @@ missing_source_file() { }

run_pacman() { - local cmd + local cmd cmdescape if [[ $1 = -@(T|Q)*([[:alpha:]]) ]]; then cmd=("$PACMAN_PATH" "$@") else cmd=("$PACMAN_PATH" "${PACMAN_OPTS[@]}" "$@") - if type -p sudo >/dev/null; then + cmdescape="$(printf '%q ' "${cmd[@]}")" + if (( ${#PACMAN_AUTH[@]} )); then + if in_array '%q' "${PACMAN_AUTH[@]}"; then + cmd=("${PACMAN_AUTH[@]/\%q/$cmdescape}") + else + cmd=("${PACMAN_AUTH[@]}" "${cmd[@]}") + fi + elif type -p sudo >/dev/null; then

Can we just put sudo in PACMAN_PATH in our makepkg.conf by deafult. Then we can get rid of the sudo path and just have su -c as a fallback.

We probably want a check for the binary at the start of the PACMAN_AUTH instead of sudo in scripts/libmakepkg/executable/sudo.sh.in too.

The current implementation was supposed to assume that the users know what they are doing in setting PACMAN_AUTH to non-default values, and refrain from second-guessing them by erroring rather than trying su as a fallback (they explicitly asked for it, don't try something else instead). sudo is our attempt to gracefully pick our recommended tool automatically, if needed/available. That being said, I guess at a minimum, executable_sudo() should not warn you if: - sudo is not installed - PACMAN_AUTH is set to something you installed

...

cmd=(sudo "${cmd[@]}") else - cmd=(su root -c "$(printf '%q ' "${cmd[@]}")") + cmd=(su root -c "$cmdescape") fi local lockfile="$(pacman-conf DBPath)/db.lck" while [[ -f $lockfile ]]; do

-- Eli Schwartz Bug Wrangler and Trusted User