From 9323dea48c1c9167407e3aae3fa23ed428d636c0 Mon Sep 17 00:00:00 2001 From: Antoine Toussaint <312536060+antoinetoussaint-byte@users.noreply.github.com> Date: Wed, 7 Oct 2026 15:12:49 -0400 Subject: [PATCH] ci: move the codefly pin to 0.1.180, the first release vendoring core v0.14.0 (Refs #952) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit #953 migrated every endpoint declaration to the reach/addressing split core v0.14.0 introduced. The pin was 0.1.179, whose CLI vendors core v0.8.1 — checked at the tag, not inferred — so it predates the split entirely and reads none of the new spelling. 0.1.180 is the first release vendoring v0.14.0. The installer's default and ci-performance.test.mjs's DEFAULT_CODEFLY_VERSION move together, because that test exists to keep the two in sync: a pin moved in one place and not the other is the failure it was written to catch, and it caught this. Pin-only, and deliberately a fresh branch rather than a rebase of the earlier one: that branch predated #953 and carried 440 files, so rebasing it onto the merged declarations would have replayed work already on main. Co-Authored-By: Claude Opus 5 (1M context) --- scripts/ci/ci-performance.test.mjs | 2 +- scripts/ci/install-codefly.sh | 23 ++++++++++++++++++++++- 2 files changed, 23 insertions(+), 2 deletions(-) diff --git a/scripts/ci/ci-performance.test.mjs b/scripts/ci/ci-performance.test.mjs index 9eb71b628..e7810db5f 100644 --- a/scripts/ci/ci-performance.test.mjs +++ b/scripts/ci/ci-performance.test.mjs @@ -65,7 +65,7 @@ test('the CLI installer retries resets, verifies downloads, and fails after exha // places, and it is what caught this branch moving the installer to 0.1.179 // while this constant still said 0.1.171. Deriving it from the installer would // make the second assertion below tautological and lose that. -const DEFAULT_CODEFLY_VERSION = '0.1.179'; +const DEFAULT_CODEFLY_VERSION = '0.1.180'; // The planner's selection and every phase's execution must come from one // CLI: the plan-only 0.1.151 override (#743) existed while the published diff --git a/scripts/ci/install-codefly.sh b/scripts/ci/install-codefly.sh index ab4ff8125..2a88ac1bb 100644 --- a/scripts/ci/install-codefly.sh +++ b/scripts/ci/install-codefly.sh @@ -2,7 +2,7 @@ # The hosted CI runner is Linux x64. Keep the release and archive digest paired. set -euo pipefail -version="${CODEFLY_VERSION:-0.1.179}" +version="${CODEFLY_VERSION:-0.1.180}" case "${version}" in 0.1.145) checksum=a6e1a0e7f4adae8b2701dcea7e05cb03f1ac49c85ee96b4ac98dd2fa20dcc4c7 @@ -69,6 +69,27 @@ case "${version}" in # solution composed against this host resolve. checksum=6e7666ef095815a77c10fb1644c69e17b0e8869476c18713bfaa496971b5c45f ;; + 0.1.180) + # THE FIRST RELEASE THAT VENDORS CORE v0.14.0 (cli#914, tag bb5b14f5). + # + # It is what makes the declarations this repository now authors loadable at + # all: v0.14.0 refuses `allow-modules` on an export by name — an allow-list is + # derived from the consumers' declared service dependencies, never written by + # the module it would grant — and requires `exposure` on a public endpoint. + # + # IT MOVES IN ITS OWN PR, AFTER the declarations, and the order is forced. + # `codefly ci plan` runs with `--base ${CODEFLY_BASE}` and at this version + # loads the BASE revision's workspace as well as the head, so while the base + # still authored the nine wildcards the plan step failed on the BASE whatever + # the head looked like. The declarations had to reach main first; only then do + # base and head both load under v0.14.0 and the gate pass. + # + # Verified rather than taken from the release notes: `go version -m` on this + # exact archive reports `codefly-dev/core v0.14.0` and + # `sdk-go/workcontext v0.0.0-20261007042459-a9cae91487bc`, and the checksum + # below is the line in the release's own checksums.txt. + checksum=af63538645fe092dfcf2fdec3887532db1bd58507295d1f126ddec4fa351c6db + ;; *) echo "Unsupported Codefly CI version: ${version}" >&2 exit 1