From 759d4e644d340b464bad49f6b1c6ffc9b1855dfe Mon Sep 17 00:00:00 2001 From: serrrfirat Date: Sun, 21 Jun 2026 23:54:23 +0300 Subject: [PATCH] ci(reborn): retry crates.io network failures in the closure (CARGO_NET_RETRY) The 64-crate closure runs ~64 jobs that each resolve+download the dependency tree from crates.io in parallel. A transient registry hiccup (seen twice: 'download of once_cell_polyfill failed' HTTP2 framing, and 'agent-client-protocol ... curl [55] SSL_ERROR_SYSCALL') hits many jobs at once and reddens the whole run even though nothing is wrong with the code. Bump Cargo's network retry from the default 3 to 10 and fetch git deps via the git CLI, so transient download failures self-heal instead of failing the job. The heavier fix (build-once `nextest archive` + shard the run, eliminating 64x dependency resolution) remains a separate follow-up. Co-Authored-By: Claude Opus 4.8 --- .github/workflows/reborn-tests.yml | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/.github/workflows/reborn-tests.yml b/.github/workflows/reborn-tests.yml index 99db88014ca..f5ed48f9367 100644 --- a/.github/workflows/reborn-tests.yml +++ b/.github/workflows/reborn-tests.yml @@ -44,6 +44,13 @@ env: RUSTFLAGS: "-C linker=clang -C link-arg=--ld-path=/usr/bin/mold" CARGO_PROFILE_DEV_DEBUG: 0 CARGO_PROFILE_TEST_DEBUG: 0 + # The 64-crate closure runs ~64 jobs that each resolve+download the dependency + # tree from crates.io in parallel, so a transient registry SSL/HTTP2 hiccup + # (e.g. "download of failed: curl [55] SSL_ERROR_SYSCALL") tends to hit + # many jobs at once and redden the whole run. Make Cargo retry network failures + # (default 3) far more aggressively, and use the git CLI for git deps. + CARGO_NET_RETRY: "10" + CARGO_NET_GIT_FETCH_WITH_CLI: "true" jobs: changes: