Model what each parser does with an empty and a dangling directive

Review of the previous commits turned up four places where the predicates
and their tests disagreed with the tools they are modelling, each checked
against udevadm verify, systemd-analyze verify and visudo -cf rather than
against reading of the sources.

An empty ExecStop= resets the list, so a unit an administrator neutralised
that way runs nothing at shutdown and is no longer ours to remove; the
predicate now tracks the last state instead of returning on the first home
path it sees. A file whose last line ends in a backslash still carries a
live directive for systemd, so the pending logical line is emitted at EOF;
udev ignores such a line and sudo rejects the file outright, so this costs
those two nothing. The scanner's taint pass now reads += appends, which its
own comment already promised: the value of an append is no use, but a name
that reaches a user root through one has to be judged on it.

Two regression guards passed against the implementations they were written
for. The udev continuation fixture put the whole RUN+= below the comment, so
it matched whether or not the pending half was carried across; the split now
falls inside the RUN+= value. The sudoers one kept its file on the strength
of a spec above the comment, so it could not fail either; the hand-written
spec now sits below. Both fail against a mutant that discards the pending
line. The comment above the second also claimed a continued comment stays a
comment, which visudo contradicts.
This commit is contained in:
acrogenesis
2026-08-29 19:23:02 -06:00
parent d593847728
commit 394c1371c9
7 changed files with 90 additions and 19 deletions
+11 -6
View File
@@ -176,16 +176,17 @@ declare -A VARS_TAINTED=()
# isolation would miss in whichever direction it picked.
collect_vars() {
local -n source_lines="$1"
local line name value
local line name value append
VARS=()
VARS_TAINTED=()
for line in "${source_lines[@]}"; do
[[ $line =~ ^[[:space:]]*# ]] && continue
[[ $line =~ ^[[:space:]]*(local|declare|export|readonly|typeset)?[[:space:]]*([A-Za-z_][A-Za-z0-9_]*)=(.*)$ ]] || continue
[[ $line =~ ^[[:space:]]*(local|declare|export|readonly|typeset)?[[:space:]]*([A-Za-z_][A-Za-z0-9_]*)(\+?)=(.*)$ ]] || continue
name=${BASH_REMATCH[2]}
value=${BASH_REMATCH[3]}
append=${BASH_REMATCH[3]}
value=${BASH_REMATCH[4]}
value=${value%%[[:space:]]#*}
value=${value%[[:space:]]}
if [[ $value == \"*\" || $value == \'*\' ]]; then
@@ -193,6 +194,11 @@ collect_vars() {
fi
mentions_user_writable_root "$value" && VARS_TAINTED["$name"]=1
# An append never wins the value -- an array grown across a file resolves to
# nothing useful -- but it does carry the taint, or a name could reach a user
# root through += and never be judged on it.
[[ -n $append ]] && continue
[[ -v VARS[$name] ]] || VARS["$name"]=$value
done
}
@@ -457,9 +463,8 @@ privileged_destination() {
fi
done < <(command_destinations "$line")
# An elevated write whose destination cannot be resolved counts as
# privileged: sudo tee is not aimed at a user's own dotfile, and assuming
# otherwise is how this bug class survived six reviews.
# An elevated write whose destination cannot be resolved counts as privileged:
# sudo tee is not aimed at a user's own dotfile.
if ((elevated == 0)) && ((${#unresolved[@]} > 0)); then
printf '%s' "${unresolved[0]} (unresolved destination of an elevated write)"
return 0