Overlord/porthole
1
1
Fork 0
Code Releases 1 Activity

Replace mid-sentence dashes in comments with commas/semicolons

Browse Source 
Swapped " - " for a semicolon where it joined two independent clauses,
or a comma where the following text was an appositive/dependent phrase
with no subject of its own. Markdown-style list bullets in
tests/live_test.sh's header (leading "- Item") are unaffected - those
are structural, not sentence punctuation.

Also dropped a "Spec §5.4:" citation in edit.rs missed by the earlier
spec-citation cleanup (case-sensitive grep at the time didn't match the
capitalized "Spec").

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
This commit is contained in:
Overlord
2026-08-14 12:36:59 +02:00
parent 40437bf78c
commit 6e6b5c1393
11 changed files with 29 additions and 29 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
tests/live_test.sh
View File

@@ -5,7 +5,7 @@
# auth (success and failure), a real two-hop ProxyJump, a real SOCKS proxy
# carrying real traffic. Deliberately NOT part of `cargo test` (same reason
# as vmic's tests/live_test.sh: this needs a real remote server, not a CI
# sandbox) - live-only, opt-in, run by hand.
# sandbox), live-only, opt-in, run by hand.
#
# Usage:
# tests/live_test.sh [options]
@@ -22,21 +22,21 @@
# -h, --help
#
# Known limitation: only one real server is available, so multi-hop (`-J`)
# is tested by chaining the server through itself (--via user@host,user@host)
# - a real two-hop ProxyJump handshake, just with both hops the same box.
# is tested by chaining the server through itself (--via user@host,user@host),
# a real two-hop ProxyJump handshake, just with both hops the same box.
# There is no way to test a genuine distinct-host chain without a second
# server.
#
# Safety:
# - Profiles/instances/locks/logs are sandboxed for the whole run under
# one `PORTHOLE_STATE_DIR_OVERRIDE` temp dir (profile.rs/instance.rs
# both honor it) - this suite NEVER touches the real
# both honor it); this suite NEVER touches the real
# ~/.config/porthole or ~/.local/state/porthole.
# - Every profile created is named "$NAME_..." (default prefix
# porttestsuite); no operation targets anything outside that prefix.
# - `wipe` (src/commands/wipe.rs) kills ANY process on the
# whole system whose cmdline contains "__supervise", regardless of
# which state dir it belongs to - it is NOT scoped by the sandboxing
# which state dir it belongs to; it is NOT scoped by the sandboxing
# above. Before running it, this script scans the real process table
# and skips the wipe phase entirely (not "wipe only the safe parts")
# if it finds a live __supervise process that isn't one of this run's
@@ -48,10 +48,10 @@
# never writes to your real ~/.ssh/known_hosts (only porthole's own
# spawned ssh does, against the real file, using the accept-new policy
# already forced in src/ssh.rs).
# - An EXIT trap always attempts full cleanup - closes/removes every
# - An EXIT trap always attempts full cleanup, closes/removes every
# test-prefixed profile in every sandbox dir used, force-kills any
# stray matching __supervise process, deletes the throwaway bad-auth
# key - even on failure or Ctrl-C.
# key, even on failure or Ctrl-C.
set -uo pipefail
cd "$(dirname "${BASH_SOURCE[0]}")/.."
@@ -117,7 +117,7 @@ p_run() { LAST_OUT="$(PORTHOLE_STATE_DIR_OVERRIDE="$1" "$BIN" "${@:2}" 2>&1)"; L
# status --json's "state" field is always one of closed/up/reconnecting/error
# (print_json normalizes a dead-supervisor instance file to "error" too, see
# status.rs) - polling that key is far more robust than scraping the padded
# status.rs); polling that key is far more robust than scraping the padded
# human-readable field.
json_field() { # json_field <state-dir> <name> <field> -> value, empty if absent/no instance
p "$1" status "$2" --json 2>/dev/null | sed -n "s/.*\"$3\": \"\\{0,1\\}\\([^\",]*\\)\"\\{0,1\\},\\{0,1\\}\$/\\1/p" | head -1
@@ -140,7 +140,7 @@ wait_port_closed() { # wait_port_closed <port> [tries, x0.5s]
}
SSH_PROBE_OPTS=(-o BatchMode=yes -o ConnectTimeout=5 -o StrictHostKeyChecking=accept-new -o UserKnownHostsFile=/dev/null)
# mktemp's output is captured via $(...), which forks a subshell - any
# mktemp's output is captured via $(...), which forks a subshell; any
# array append done *inside* a function called that way would be lost when
# the subshell exits, so state dirs are appended here at the call site
# instead of through a helper function.
@@ -408,7 +408,7 @@ expect_contains "warns it's left running untracked" "left running untracked"
p_run "$STATE_DIR" status "${NAME}_keep"; assert_eq "profile is gone from tracking" "$LAST_CODE" "1"
assert_true "the untracked process is still actually alive" pgrep -f "__supervise ${NAME}_keep\$"
assert_true "port $KEEP_PORT is still live, untracked" port_open "$KEEP_PORT"
# left running on purpose - phase 15's wipe is what's being tested against it
# left running on purpose; phase 15's wipe is what's being tested against it
# ---------------------------------------------------------------------------
section "Phase 14: transfer round trip against a live profile"