Compat: the copy's pending tasks aren't this run's to wait for

The run against INBUXA's store hung printing "Waiting for pending task
AcmeRenewal(...)": the copy carries that server's task queue, and a renewal
due in 2026-11 will not come due while a test watches it.

Under NO_INSERT the wait now skips tasks that aren't due and ones that have
permanently failed, which leaves the tasks the test itself caused — a
restore in undelete_compat comes due at once — and gives up after a minute
with the offending task printed. A test that was really waiting on its own
work now fails on its assertion, which says more than a spinner.

Ordinary runs are untouched: system_tests, which waits on tasks throughout,
still passes in 135s.
This commit is contained in:
2026-09-19 21:17:21 -07:00
parent 7714bca8d3
commit bfd2784819
2 changed files with 28 additions and 0 deletions
+9
View File
@@ -109,6 +109,15 @@ fails on one of its own (`tests/src/utils/server.rs`). The copy is not
edited to achieve this; its listeners are simply not what a compat run edited to achieve this; its listeners are simply not what a compat run
needs. needs.
**Why the copy's task queue is ignored.** A real server's pending tasks come
with its store: ACME renewals due months from now, work that can never run
on this machine. The harness waits for tasks to drain, so the first run
against INBUXA's copy sat printing `Waiting for pending task
AcmeRenewal(...)` until it was interrupted. Under `NO_INSERT` it now skips
tasks that aren't due and ones that have failed for good, and gives up
after a minute with the task printed, so the test fails on its own
assertion instead of spinning.
**Why `compat-` listeners appear in the copy.** The harness needs listeners **Why `compat-` listeners appear in the copy.** The harness needs listeners
on its own ports, and the registry keys listeners by name. A real server on its own ports, and the registry keys listeners by name. A real server
has its own, and a production listener called `jmap` or `imap` collided has its own, and a production listener called `jmap` or `imap` collided
+19
View File
@@ -281,6 +281,19 @@ async fn build_search_store(typ: SearchStoreType, _path: &str) -> SearchStore {
} }
pub async fn wait_for_tasks(server: &Server, skip_not_due: bool, skip_permanent_failures: bool) { pub async fn wait_for_tasks(server: &Server, skip_not_due: bool, skip_permanent_failures: bool) {
// inbuxa: a compat run opens a copy of a real server's store, and its
// task queue comes too: ACME renewals due months from now, work that
// can never run on this machine. Waiting for that to drain never ends,
// so skip what isn't due and what has already failed for good, and give
// up after a minute rather than hang. Whatever the test was waiting for
// then fails on its own assertion, which says more than a spinner.
let compat = std::env::var("NO_INSERT").is_ok();
let (skip_not_due, skip_permanent_failures) = if compat {
(true, true)
} else {
(skip_not_due, skip_permanent_failures)
};
let give_up_at = std::time::Instant::now() + std::time::Duration::from_secs(60);
let mut count = 0; let mut count = 0;
loop { loop {
let mut has_index_tasks = None; let mut has_index_tasks = None;
@@ -315,6 +328,12 @@ pub async fn wait_for_tasks(server: &Server, skip_not_due: bool, skip_permanent_
if count % 10 == 0 { if count % 10 == 0 {
println!("Waiting for pending task {:?}...", task); println!("Waiting for pending task {:?}...", task);
} }
if compat && std::time::Instant::now() > give_up_at {
println!(
"Gave up waiting after 60s; this task is the copy's, not this run's: {task:?}"
);
break;
}
tokio::time::sleep(std::time::Duration::from_millis(200)).await; tokio::time::sleep(std::time::Duration::from_millis(200)).await;
} else { } else {
break; break;