Soak test: a session kept alive for hours, with drops #66

New issue

Closed

opened 2026-10-17 17:54:18 +00:00 by littlefox · 1 comment

littlefox commented

2026-10-17 17:54:18 +00:00

Owner

Reconnecting (#52) is only trustworthy if it's tested over time, not just once.

What:

A test client logged in for a long time while another client posts, sends messages and logs in and out.
The connection is cut at random times (kill the relay connection or the TCP socket), as a phone or laptop would.
Afterwards, checks that unread counts, the reading position and the event log are right, and that memory use stayed flat.

Done when:

A scheduled workflow runs it nightly for an hour (like the fuzzing), and a short version runs on PRs that touch the engine.

Helps: #52.

From docs/ROADMAP.md, "Under the hood"; filed for 0.2.

Reconnecting (#52) is only trustworthy if it's tested over time, not just once. What: - A test client logged in for a long time while another client posts, sends messages and logs in and out. - The connection is cut at random times (kill the relay connection or the TCP socket), as a phone or laptop would. - Afterwards, checks that unread counts, the reading position and the event log are right, and that memory use stayed flat. Done when: - A scheduled workflow runs it nightly for an hour (like the fuzzing), and a short version runs on PRs that touch the engine. Helps: #52. From docs/ROADMAP.md, "Under the hood"; filed for 0.2.

littlefox added this to the 0.2: Feels like an app milestone

2026-10-17 17:54:18 +00:00

littlefox added the

kind/infra

area/ci

area/protocol

labels

2026-10-17 17:54:18 +00:00

littlefox referenced this issue from a pull request that will close it,

2026-10-17 23:01:07 +00:00

0.2: stay logged in, soak test, installable web app, keepalive #88

littlefox referenced this issue from a commit

2026-10-18 00:13:26 +00:00

soak: on a count mismatch, report a fresh login's count too

littlefox closed this issue

2026-10-18 00:35:15 +00:00

littlefox referenced this issue from a commit

2026-10-18 00:35:16 +00:00

Soak test: a long session with drops, nightly for an hour

littlefox referenced this issue

2026-10-18 00:44:05 +00:00

Screenshot comparison in CI; soak test fix #89

littlefox referenced this issue from a commit

2026-10-18 00:44:07 +00:00

soak: wait for a new post number, not just any status line

littlefox commented

2026-10-18 00:55:58 +00:00

Author

Owner

Merged in #88; follow-up fix in #89. soak.spec.ts: one minute in every e2e run (desktop), an hour nightly (soak.yml). The first CI failure turned out to be the test counting a post that wasn't made (app 52 = fresh login 52); it now waits for a new post number.