Perform destructive actions only when ACTUALLY_BAN_USERS=true

README.md

@@ -13,10 +13,22 @@
   locally in `db.json`, no concrete action is taken. (Ultimately we will want to
   lock/delete accounts, etc.)
 
+## Configuration
+
+Forgery reads the following environment variables:
+- `ACTUALLY_BAN_USERS`: define it (e.g. to `true`) to actually lock user
+  accounts, send notification emails and eventually delete user accounts. If not
+  defined (the default), no actual action is taken, spammers are only listed in
+  the database. The variable should be set in production, but probably not for
+  testing.
+
+Environment variables that are relevant when `ACTUALLY_BAN_USERS=true`:
+- `SMTP_ADDRESS`: address of the SMTP relay used to send email notifications
+- `SMTP_USERNAME`: SMTP username
+- `SMTP_PASSWORD`: SMTP password
+
 ## Todos
 
-- gate the actual account lock/email/deletion behind an environment variable
-  for easy testing and to limit bad surprises
 - discuss the current design choices for when locking the account/sending a
   notification email fails.
   (Current behavior is to periodically retry, avoid deleting if the account

src/main.rs

@@ -42,6 +42,13 @@ const GRACE_PERIOD: Duration = Duration::from_secs(30 * 24 * 3600); // 30 days
 const GUESS_SPAM_THRESHOLD: f32 = 0.8;
 const GUESS_LEGIT_THRESHOLD: f32 = 0.3;
 
+// Whether we are actually banning users or are instead in "testing" mode where
+// we don't do anything. (Defaults to "No".)
+pub enum ActuallyBan {
+    Yes { smtp: SmtpConfig },
+    No,
+}
+
 fn forge() -> anyhow::Result<Forgejo> {
     let api_token = std::fs::read_to_string(Path::new("api_token"))?
         .trim()
@@ -164,9 +171,9 @@ fn set_spam(
 
 async fn apply_classification(
     forge: &Forgejo,
-    mailer: &SmtpConfig,
     db: Arc<Mutex<Db>>,
     classifier: &mut Classifier,
+    actually_ban: Arc<ActuallyBan>,
     ids: &[(UserId, bool)],
     overwrite: bool,
 ) {
@@ -176,7 +183,7 @@ async fn apply_classification(
         let login = db.lock().unwrap().users.get(&user).unwrap().login.clone();
         // It is ok for any of these calls to fail now: a worker will periodically retry
         // TODO: signal the worker to wake up instead of performing a manual call here
-        workers::try_lock_and_notify_user(forge, mailer, db.clone(), user)
+        workers::try_lock_and_notify_user(forge, db.clone(), actually_ban.clone(), user)
             .await
             .unwrap_or_else(|err| eprintln!("Failed to lock or notify user {login}: {err}"));
     }
@@ -198,7 +205,7 @@ struct AppState {
     db: Arc<Mutex<Db>>,
     classifier: Arc<Mutex<Classifier>>,
     forge: Arc<Forgejo>,
-    mailer: Arc<SmtpConfig>,
+    actually_ban: Arc<ActuallyBan>,
 }
 
 #[derive(Debug, Deserialize)]
@@ -306,9 +313,9 @@ async fn post_classified(
 
     apply_classification(
         &data.forge,
-        &data.mailer,
         data.db.clone(),
         classifier,
+        data.actually_ban.clone(),
         &updates,
         overwrite,
     )
@@ -382,8 +389,14 @@ async fn main() -> anyhow::Result<()> {
     eprintln!("Eval templates");
     let _ = *TEMPLATES;
 
+    let actually_ban = Arc::new(match std::env::var("ACTUALLY_BAN_USERS") {
+        Ok(_) => ActuallyBan::Yes {
+            smtp: SmtpConfig::from_env().await?,
+        },
+        Err(_) => ActuallyBan::No,
+    });
+
     let forge = Arc::new(forge()?);
-    let mailer = Arc::new(SmtpConfig::from_env().await?);
     eprintln!("Load users and repos");
     let (db, classifier) = load_db(&forge).await?;
     let db = Arc::new(Mutex::new(db));
@@ -393,7 +406,7 @@ async fn main() -> anyhow::Result<()> {
         db: db.clone(),
         classifier: classifier.clone(),
         forge: forge.clone(),
-        mailer: mailer.clone(),
+        actually_ban: actually_ban.clone(),
     });
 
     let _ = {
@@ -405,13 +418,14 @@ async fn main() -> anyhow::Result<()> {
     let _ = {
         let forge = forge.clone();
         let db = db.clone();
-        tokio::spawn(async move { workers::purge_spammer_accounts(forge, db) })
+        let actually_ban = actually_ban.clone();
+        tokio::spawn(async move { workers::purge_spammer_accounts(forge, db, actually_ban) })
     };
     let _ = {
         let forge = forge.clone();
-        let mailer = mailer.clone();
         let db = db.clone();
-        tokio::spawn(async move { workers::lock_and_notify_users(forge, mailer, db) })
+        let actually_ban = actually_ban.clone();
+        tokio::spawn(async move { workers::lock_and_notify_users(forge, db, actually_ban) })
     };
 
     println!("Listening on http://127.0.0.1:8080");

src/workers.rs

@@ -1,7 +1,7 @@
 use crate::classifier::Classifier;
 use crate::data::UserId;
 use crate::db::{Db, IsSpam};
-use crate::email::SmtpConfig;
+use crate::email;
 use crate::scrape;
 use anyhow::anyhow;
 use forgejo_api::Forgejo;
@@ -9,6 +9,7 @@ use std::collections::HashMap;
 use std::path::Path;
 use std::sync::{Arc, Mutex};
 
+use crate::ActuallyBan;
 use crate::FORGEJO_POLL_DELAY;
 use crate::GRACE_PERIOD;
 use crate::{GUESS_LEGIT_THRESHOLD, GUESS_SPAM_THRESHOLD};
@@ -82,17 +83,32 @@ pub async fn refresh_user_data(
 
 // Worker to delete spam accounts after their grace period expired
 
-async fn try_purge_account(forge: &Forgejo, login: &str) -> anyhow::Result<()> {
+async fn try_purge_account(
+    forge: &Forgejo,
+    login: &str,
+    actually_ban: &ActuallyBan,
+) -> anyhow::Result<()> {
+    if let ActuallyBan::No = actually_ban {
+        eprintln!("[Simulating: delete account of user {login}]");
+        return Ok(());
+    }
+
+    eprintln!("Deleting account of user {login}");
     forge
         .admin_delete_user(
             login,
             forgejo_api::structs::AdminDeleteUserQuery { purge: Some(true) },
         )
         .await?;
+    eprintln!("Success");
     Ok(())
 }
 
-pub async fn purge_spammer_accounts(forge: Arc<Forgejo>, db: Arc<Mutex<Db>>) {
+pub async fn purge_spammer_accounts(
+    forge: Arc<Forgejo>,
+    db: Arc<Mutex<Db>>,
+    actually_ban: Arc<ActuallyBan>,
+) {
     loop {
         let mut classified_users = Vec::new();
         {
@@ -129,7 +145,7 @@ pub async fn purge_spammer_accounts(forge: Arc<Forgejo>, db: Arc<Mutex<Db>>) {
                             );
                         }
 
-                        if let Err(e) = try_purge_account(&forge, &login).await {
+                        if let Err(e) = try_purge_account(&forge, &login, &actually_ban).await {
                             eprintln!("Error while deleting spammer account {login}: {:?}", e)
                         } else {
                             eprintln!("Deleted spammer account {login}");
@@ -185,8 +201,8 @@ pub async fn lock_user_account(forge: &Forgejo, username: &str) -> anyhow::Resul
 
 pub async fn try_lock_and_notify_user(
     forge: &Forgejo,
-    mailer: &SmtpConfig,
     db: Arc<Mutex<Db>>,
+    actually_ban: Arc<ActuallyBan>,
     user_id: UserId,
 ) -> anyhow::Result<()> {
     let (login, email, is_spam) = {
@@ -205,7 +221,15 @@ pub async fn try_lock_and_notify_user(
 
     if let Some((classified_at, locked, notified)) = is_spam {
         if !locked {
+            match actually_ban.as_ref() {
+                ActuallyBan::Yes { .. } => {
+                    eprintln!("Locking account of user {login}");
                     lock_user_account(forge, &login).await?;
+                    eprintln!("Success");
+                }
+                ActuallyBan::No => eprintln!("[Simulating: lock account of user {login}]"),
+            }
+
             let db = &mut db.lock().unwrap();
             db.is_spam.insert(
                 user_id,
@@ -219,7 +243,16 @@ pub async fn try_lock_and_notify_user(
         }
 
         if !notified {
-            crate::email::send_locked_account_notice(mailer, &login, &email).await?;
+            match actually_ban.as_ref() {
+                ActuallyBan::Yes { smtp } => {
+                    eprintln!("Sending notification email to user {login}");
+                    email::send_locked_account_notice(&smtp, &login, &email).await?;
+                    eprintln!("Success");
+                }
+                ActuallyBan::No => {
+                    eprintln!("[Simulating: send notification email to user {login}]")
+                }
+            }
             let db = &mut db.lock().unwrap();
             db.is_spam.insert(
                 user_id,
@@ -243,8 +276,8 @@ pub async fn try_lock_and_notify_user(
 
 pub async fn lock_and_notify_users(
     forge: Arc<Forgejo>,
-    mailer: Arc<SmtpConfig>,
     db: Arc<Mutex<Db>>,
+    actually_ban: Arc<ActuallyBan>,
 ) {
     let mut spammers = Vec::new();
     {
@@ -257,7 +290,7 @@ pub async fn lock_and_notify_users(
     }
 
     for (user_id, login) in spammers {
-        try_lock_and_notify_user(&forge, &mailer, db.clone(), user_id)
+        try_lock_and_notify_user(&forge, db.clone(), actually_ban.clone(), user_id)
             .await
             .unwrap_or_else(|err| eprintln!("Failed to lock or notify user {login}: {err}"));
     }