Track pre-existing device integration test timeouts #1

New issue

Open

opened 2026-11-06 20:06:16 +00:00 by kayg · 0 comments

kayg commented

2026-11-06 20:06:16 +00:00

Owner

Summary

The six network integration test timeouts reported by the J225 full suite are reproducible before the parallel crypto change. They occur in the existing one-second eavesdrop/data wait in the device test harness. This issue tracks them separately so the parallel crypto change does not absorb or obscure the pre-existing test instability.

Evidence

Plain fork main at 2a867b3, cargo test --locked --workspace --all-features: 92 passed; 2 failed; 7 ignored in 1.59 s. device::tests::one_handshake and device::tests::test_indices hit eavesdrop timeout.
Direct pre-J225 base 2bb5347, same command: 140 passed; 6 failed; 7 ignored in 1.71 s. Failures: ipv6_isnt_used, number_of_packets, one_handshake, one_keepalive, test_indices, and wg_data_length_is_x16.
At J225 head 13ed31c, all six pass when run alone. The parallel full suite reports the same six; a serial full-suite rerun leaves test_indices timed out.
The matched logs are on compute01 under /home/claude/j225-evidence/. The full comparison, exact commands, and log paths are recorded in infraternal #260 comment 82860: kayg/infraternal#260 (comment).

Follow-up

Investigate the harness timing sensitivity separately from the ordered crypto work. Preserve the standalone and full-suite evidence before changing the timeout or test scheduling.

## Summary The six network integration test timeouts reported by the J225 full suite are reproducible before the parallel crypto change. They occur in the existing one-second eavesdrop/data wait in the device test harness. This issue tracks them separately so the parallel crypto change does not absorb or obscure the pre-existing test instability. ## Evidence - Plain fork `main` at `2a867b3`, `cargo test --locked --workspace --all-features`: `92 passed; 2 failed; 7 ignored` in 1.59 s. `device::tests::one_handshake` and `device::tests::test_indices` hit `eavesdrop timeout`. - Direct pre-J225 base `2bb5347`, same command: `140 passed; 6 failed; 7 ignored` in 1.71 s. Failures: `ipv6_isnt_used`, `number_of_packets`, `one_handshake`, `one_keepalive`, `test_indices`, and `wg_data_length_is_x16`. - At J225 head `13ed31c`, all six pass when run alone. The parallel full suite reports the same six; a serial full-suite rerun leaves `test_indices` timed out. - The matched logs are on compute01 under `/home/claude/j225-evidence/`. The full comparison, exact commands, and log paths are recorded in infraternal #260 comment 82860: https://git.kayg.org/kayg/infraternal/issues/260#issuecomment-82860. ## Follow-up Investigate the harness timing sensitivity separately from the ordered crypto work. Preserve the standalone and full-suite evidence before changing the timeout or test scheduling.

kayg referenced this issue from kayg/infraternal

2026-11-06 20:06:52 +00:00

Head-to-head: infraternal mesh vs NetBird/Tailscale, infraternal sessions vs tmux/Zellij #260

kayg referenced this issue from kayg/infraternal

2026-11-07 13:48:43 +00:00

Head-to-head: infraternal mesh vs NetBird/Tailscale, infraternal sessions vs tmux/Zellij #260