Files
PaperClipAI/scripts
Priya RamanandPaperclip 6d643cc068 fix(cargo): raise network retry and timeout to survive a transient stall
Cargo's default network settings abort a crate download after 30
seconds below 10 bytes per second, then retry 3 times. On a shared CI
runner, a brief network stall can cross that 30-second window on every
attempt, so a build with hundreds of uncached crate downloads can fail
even though the network recovers soon after.

Add a repository-root .cargo/config.toml that raises net.retry to 5
and http.timeout to 60 seconds. Cargo finds this file by an upward
directory search from the current working directory, independent of
any --manifest-path argument, so every cargo invocation in the
repository picks it up.

Add a deterministic local harness that serves a synthetic crate over
a stalling HTTP server, then proves the old defaults fail with the
exact CI error text and the new values survive the same stall. Add a
regression test that asserts the config values stay at or above the
chosen minimums, wired into the existing release-registry test list
that CI already runs.

Co-authored-by: Paperclip <noreply@paperclip.ing>
2026-09-04 00:07:04 +00:00
..
2026-03-09 10:03:45 -05:00
2026-03-03 14:00:10 -06:00
2026-04-09 06:14:12 -05:00