…r ones (pingdotgg#378/pingdotgg#315) `cargo test` is fail-fast ACROSS TARGETS and backend has seven of them. When one fails, cargo stops and the remaining targets are never BUILT -- their tests are not reported as failed or skipped, they are absent from the output entirely. A reader, or a test-count ratchet, sees a smaller suite with no indication anything is missing. Measured here: a --lib failure hid restart_reconnect (15 tests), shell_session (4), log_backpressure (1) and agent_default_model (1). Twenty-one tests that looked like they did not exist. This is not hypothetical and it is not only this repo. I published "do-rs 469 / 0" for several hours today off a fail-fast run whose later targets were never built; the real number was 472 / 2. clau-be20 hit the same shape from the other side -- a killed `cargo test` reads as "fewer tests, all passing", which is indistinguishable from this. There is no cargo key that makes plain `cargo test` non-fail-fast, so the fix is the tool's own mechanism: an alias, so the project's test command carries the flag by construction rather than by everyone remembering it. [alias] test-all = "test --no-fail-fast" Deliberately NOT solved by #[ignore]-ing the flaky target -- that removes the flake and the guarantee together. Both hunks are pure additions on top of current main: the config.toml block appends after the existing pingdotgg#408 [net] section (which is untouched), and the AGENTS.md section appends at the end. `diff` against main shows insertions only, no context changed. Parented on shared t3code-new main f5507d9 -- 1 ahead, 0 behind, a clean fast-forward.

Please integrate a Notion view #408

Description

Activity

shivamhwp commented on Mar 24, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions