marek-miklewicz/alt-mysql
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

docs: fix subshell-wrapping bug in Task 2's OS-guard test design

Browse Source 
Sourcing native_mariadb_port_takeover.sh for unit tests means a failing
check's die() -> exit 1 would terminate the sourcing test script itself
rather than just failing an `if` condition. Wrap the three
expected-to-fail calls in subshells so exit is confined to the subshell.
This commit is contained in:
Marek Miklewicz
2026-07-05 21:26:55 +02:00
parent f15da3b7aa
commit 15044f0dea
1 changed files with 14 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/superpowers/plans/2026-07-05-native-port-takeover.md
+14 -3
View File
@@ -244,7 +244,7 @@ ID="cloudlinux"
VERSION_ID="8.9"
ID_LIKE="rhel fedora centos"
EOF
if OS_RELEASE_FILE="$TMP_DIR/os-release-cloudlinux" check_os_guard 2>/dev/null; then
if ( OS_RELEASE_FILE="$TMP_DIR/os-release-cloudlinux" check_os_guard 2>/dev/null ); then
fail "CloudLinux must be refused by the OS guard"
fi
@@ -253,7 +253,7 @@ cat > "$TMP_DIR/os-release-alma7" <<'EOF'
ID="almalinux"
VERSION_ID="7.9"
EOF
if OS_RELEASE_FILE="$TMP_DIR/os-release-alma7" check_os_guard 2>/dev/null; then
if ( OS_RELEASE_FILE="$TMP_DIR/os-release-alma7" check_os_guard 2>/dev/null ); then
fail "AlmaLinux 7 must be refused by the OS guard (only 8/9 supported)"
fi
@@ -313,13 +313,24 @@ chmod 755 "$TMP_DIR/bin/ss"
PATH="$TMP_DIR/bin:$PATH" NEW_NATIVE_PORT="3307" preflight_new_port_free \
|| fail "preflight_new_port_free should pass when the target port is free"
if PATH="$TMP_DIR/bin:$PATH" NEW_NATIVE_PORT="9999" preflight_new_port_free 2>/dev/null; then
if ( PATH="$TMP_DIR/bin:$PATH" NEW_NATIVE_PORT="9999" preflight_new_port_free 2>/dev/null ); then
fail "preflight_new_port_free should refuse when the target port is already occupied"
fi
echo "native_mariadb_port_takeover_detect_test: OK"
```
Note: the three "expected to fail" calls above (`check_os_guard` for CloudLinux and
AlmaLinux 7, and `preflight_new_port_free` for an occupied port) are each wrapped in a
subshell `( ... )` rather than called bare. This test sources the script directly rather
than executing it as a subprocess, so a bare failing call's `die()` → `exit 1` would
terminate the whole sourcing shell (this test script itself), never letting the `if`
statement observe the failure. Wrapping the call in `( ... )` confines that `exit` to the
subshell, so `if ( failing_call ); then` correctly observes a non-zero status without
killing the test. Do not "fix" this by changing `die()` to use `return` instead of
`exit` — `die()` is also used when this script is executed directly (not sourced), where
`exit` is the correct, intended behavior.
- [ ] **Step 2: Run test to verify it fails**
Run: `bash tests/native_mariadb_port_takeover_detect_test.sh`