hubcio opened a new pull request, #3124:
URL: https://github.com/apache/iggy/pull/3124

   Releases must not push a git tag until the target registry is
   actually serving the artifact, and reruns must converge instead
   of leaving half-published state behind. Today cargo search is
   the availability oracle (rate-limited, laggy, and --limit 1
   returns a relevance-ranked hit rather than an exact match),
   the foreign-SDK matrix pushes tags with no wait gate at all,
   and create-git-tag's early-skip silently accepts a pre-existing
   tag at the wrong commit.
   
   A new wait-for-crate composite queries the crates.io sparse
   index directly (CDN-fronted, unrate-limited, authoritative as
   of cargo publish return) and is reused at max_attempts=1 as
   the rust/post-merge idempotency pre-check so reruns skip
   already-published crates cleanly. A new wait-for-url composite
   gates every foreign-SDK tag step on HTTP 200 from its registry
   (PyPI, npm, Maven Central with ~25 min propagation budget,
   NuGet); Go stays tag-only. create-git-tag's early-skip now
   peels the remote ref and compares against the requested commit,
   so a wrong-target tag is a hard failure with recovery commands
   inline. The tag input regex now accepts semver build metadata.
   
   Plus smaller tightenings: env: indirection on every run block
   in rust/post-merge; dry_run kept there as a deprecated short-
   circuiting input so forks pinning by SHA do not silently start
   real-publishing; check-tags queries the remote; and
   _publish_rust_crates.yml requires non-empty inputs.commit
   instead of bypassing master-ancestry via a github.sha fallback.
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to