#!/usr/bin/env bash
#
# Copyright (c) 2024 YunoHost Contributors
#
# This file is part of YunoHost (see https://yunohost.org)
#
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU Affero General Public License as
# published by the Free Software Foundation, either version 3 of the
# License, or (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU Affero General Public License for more details.
#
# You should have received a copy of the GNU Affero General Public License
# along with this program. If not, see <http://www.gnu.org/licenses/>.
#

# Create a dedicated systemd config
#
# usage: ynh_config_add_systemd [--mount=mount] [--service=service] [--template=templatename]
# | arg: --mount=       - Mount name (optional)
# | arg: --service=     - Service name (optional, `$app` by default)
# | arg: --template=    - Name of template file (optional, 'systemd.service' by default, meaning `../conf/systemd.service` will be used as template)
#
# This will use the template `../conf/<templatename>`.
#
# See the documentation of `ynh_config_add` for a description of the template
# format and how placeholders are replaced with actual variables.
ynh_config_add_systemd() {
    # ============ Argument parsing =============
    local -A args_array=([m]=mount= [s]=service= [t]=template=)
    local mount
    local service
    local template
    ynh_handle_getopts_args "$@"
    mount="${mount:-}"
    service="${service:-$app}"
    template="${template:-systemd.service}"
    # ===========================================

    if [[ -n "$mount" ]]; then
        ynh_config_add --template="$template" --destination="/etc/systemd/system/$mount.mount"
        systemctl enable "$mount.mount" --quiet
    else
        ynh_config_add --template="$template" --destination="/etc/systemd/system/$service.service"
        systemctl enable "$service.service" --quiet
    fi
    systemctl daemon-reload
}

# Remove the dedicated systemd config
#
# usage: ynh_config_remove_systemd service
# | arg: service   - Service name (optional, $app by default)
ynh_config_remove_systemd() {
    local service="${1:-$app}"
    if [ -e "/etc/systemd/system/$service.service" ]; then
        ynh_systemctl --service="$service" --action=stop
        systemctl disable "$service.service" --quiet
        ynh_safe_rm "/etc/systemd/system/$service.service"
        systemctl daemon-reload
    fi
    if [ -e "/etc/systemd/system/$service.mount" ]; then
        ynh_systemctl --service="$service" --action=stop
        systemctl disable "$service.mount" --quiet
        ynh_safe_rm "/etc/systemd/system/$service.mount"
        systemctl daemon-reload
    fi
}

# Start (or other actions) a service,  print a log in case of failure and optionaly wait until the service is completely started
#
# usage: ynh_systemctl [--service=service] [--action=action] [ [--wait_until="line to match"] [--log_path=log_path] [--timeout=300] [--length=20] ]
# | arg: --service=      - Name of the service to start. Default : `$app`
# | arg: --mount=        - Name of the mount. If set, service="$mount.mount"
# | arg: --action=       - Action to perform with systemctl. Default: start
# | arg: --wait_until=   - The pattern to find in the log to attest the service is effectively fully started.
# | arg: --log_path=     - Log file - Path to the log file. Default : `/var/log/$app/$app.log`; `systemd` to listen on `journalctl --unit=$service`
# | arg: --timeout=      - Timeout - The maximum time to wait before ending the watching. Default : 60 seconds.
# | arg: --length=       - Length of the error log displayed for debugging : Default : 20
ynh_systemctl() {
    # ============ Argument parsing =============
    local -A args_array=([n]=service= [m]=mount= [a]=action= [w]=wait_until= [p]=log_path= [t]=timeout= [e]=length=)
    local service
    local mount
    local action
    local wait_until
    local length
    local log_path
    local timeout
    ynh_handle_getopts_args "$@"
    service="${service:-$app}"
    mount="${mount:-}"
    action=${action:-start}
    wait_until=${wait_until:-}
    length=${length:-20}
    log_path="${log_path:-/var/log/$service/$service.log}"
    # ===========================================

    if [ -n "$mount" ]; then
        service="$mount.mount"
    fi

    if ynh_in_ci_tests && [ "$length" -le 20 ]; then
        # Use a shorter timeout than in production to save CI time (services should rarely take up more than 2 minutes to start with the CI hardware)
        timeout=${timeout:-120}
        # On CI, use length=100 because it's sometime hell to debug otherwise for super-long output
        length=100
    else
        # 300 is a conservative value, especially because it may take quite
        # long for service to start on low-end hardware, and CI should already
        # have validate that the service is expected to be working
        timeout=${timeout:-300}
    fi

    # Manage case of service already stopped
    if [ "$action" == "stop" ] && ! systemctl is-active --quiet "$service"; then
        return 0
    fi

    # Start to read the log
    if [[ -n "$wait_until" ]]; then
        local templog="$(mktemp)"
        # Following the starting of the app in its log
        if [ "$log_path" == "systemd" ]; then
            # Read the systemd journal
            journalctl --unit="$service" --follow --since=-0 --quiet > "$templog" &
            # Get the PID of the journalctl command
            local pid_tail=$!
        else
            # Read the specified log file
            tail --follow=name --retry --lines=0 "$log_path" > "$templog" 2>&1 &
            # Get the PID of the tail command
            local pid_tail=$!
        fi
    fi

    # Use reload-or-restart instead of reload. So it wouldn't fail if the service isn't running.
    if [ "$action" == "reload" ]; then
        action="reload-or-restart"
    fi

    local time_start="$(date --utc --rfc-3339=seconds | cut -d+ -f1) UTC"

    # If the service fails to perform the action
    if ! systemctl "$action" "$service"; then
        # Show syslog for this service
        journalctl --quiet --no-hostname --no-pager --lines="$length" --unit="$service" >&2
        # If a log is specified for this service, show also the content of this log
        if [ -e "$log_path" ]; then
            tail --lines="$length" "$log_path" >&2
        fi
        _ynh_clean_check_starting
        return 1
    fi

    # Start the timeout and try to find wait_until
    if [[ -n "${wait_until:-}" ]]; then
        set +o xtrace # set +x
        local i=0
        local starttime=$(date +%s)
        for i in $(seq 1 "$timeout"); do
            # Read the log until the sentence is found, that means the app finished to start. Or run until the timeout
            if [ "$log_path" == "systemd" ]; then
                # For systemd services, we in fact dont rely on the templog, which for some reason is not reliable, but instead re-read journalctl every iteration, starting at the timestamp where we triggered the action
                if journalctl --unit="$service" --since="$time_start" --quiet --no-pager --no-hostname | grep --extended-regexp --quiet "$wait_until"; then
                    ynh_print_info "The service $service has correctly executed the action ${action}."
                    break
                fi
            else
                if grep --extended-regexp --quiet "$wait_until" "$templog"; then
                    ynh_print_info "The service $service has correctly executed the action ${action}."
                    break
                fi
            fi
            if [ "$i" -eq 30 ]; then
                echo "(this may take some time)" >&2
            fi
            # Also check the timeout using actual timestamp, because sometimes for some reason,
            # journalctl may take a huge time to run, and we end up waiting literally an entire hour
            # instead of 5 min ...
            if [[ "$(($(date +%s) - starttime))" -gt "$timeout" ]]; then
                i=$timeout
                break
            fi
            sleep 1
        done
        set -o xtrace # set -x
        if [ "$i" -ge 3 ]; then
            echo "" >&2
        fi
        if [ "$i" -eq "$timeout" ]; then
            ynh_print_warn "The service $service didn't fully execute the action ${action} before the timeout."
            ynh_print_warn "Please find here an extract of the end of the log of the service $service:"
            journalctl --quiet --no-hostname --no-pager --lines="$length" --unit="$service" >&2
            if [ -e "$log_path" ]; then
                ynh_print_warn "==="
                tail --lines="$length" "$log_path" >&2
            fi

            # If we tried to reload/start/restart but reached timeout, then handle it as a failure
            if [ "$action" == "reload" ] || [ "$action" == "start" ] || [ "$action" == "restart" ]; then
                # NB: we don't check the actual 'active' status of the service,
                # because if the service is on Restart=always, it may be in a starting loop where it's temporarily active...
                _ynh_clean_check_starting
                return 1
            fi
        fi
        _ynh_clean_check_starting
    fi
}

_ynh_clean_check_starting() {
    if [ -n "${pid_tail:-}" ]; then
        # Stop the execution of tail.
        kill -SIGTERM "$pid_tail" 2>&1
    fi
    if [ -n "${templog:-}" ]; then
        ynh_safe_rm "$templog" 2>&1
    fi
}