Skip to content

Release 0.5.0: @abap2ui5/node-runtime 1.147.0, the express adapter found in either output layout - #107

Merged
oblomov-dev merged 2 commits into
mainfrom
claude/inspiring-volta-dh7djv
Oct 14, 2026
Merged

oblomov-dev merged 2 commits into
mainfrom
claude/inspiring-volta-dh7djv

Conversation

@oblomov-dev

Copy link
Copy Markdown
Member

The release PR for 0.5.0, per AGENTS.md "Releasing": plugin/package.json moves to 0.5.0 and pins the published @abap2ui5/node-runtime at 1.147.0 (abap2UI5 1.147.0, tagged 2026-10-14); the Unreleased lines of plugin/CHANGELOG.md move under [0.5.0], with a line for the runtime.

Two things the nightly run against upstream main found (38419982897), both from that runtime:

  • The output layout. 1.147.0's output/ is a folder per origin (@abaplint/transpiler 2.14), and locate( ) named the express adapter by the flat path - the conformance host died on the import with ERR_MODULE_NOT_FOUND. shimPath( ) takes output/express-icf-shim/ first and the flat output/ of 1.146.0 and older second, so an overrides to an older runtime still boots, and a package with neither says so instead of failing on the import. The accelerate test's stand-ins hard-coded the flat path too; they take it from what locate( ) found.
  • "the framework's 500 does not reflect the client's url" was the same mismatch from the other side: runtime/ held 1.147.0 while the plugin's nested node_modules kept the pinned 1.146.0, which predates the core fix. With 1.147.0 resolved everywhere the test passes - the URL comes back stripped.

Checked here on the published 1.147.0 (scripts/assemble-runtime.sh --package 1.147.0): npm run lint, npm test (222 tests, 221 pass, 1 skipped: the MCP_SERVER_DIR one), npm run consumer-test (both tarballs packed and installed into a throwaway CAP project, the runtime that loads is 1.147.0, the roundtrips, abap2js and the agent endpoint answer).

After the merge: git tag v0.5.0 && git push origin v0.5.0 - release.yml fills runtime/ from the published 1.147.0, runs the suite and consumer-test on the tag and publishes by trusted publishing. Then cap2UI5/samples#11 (peer range ^0.5.0) can go out as 0.3.0.

🤖 Generated with Claude Code

https://claude.ai/code/session_01TrfF51pia7m6tLZQF2G2ua


Generated by Claude Code

…und in either output layout

The plugin pins the published runtime at 1.147.0 (abap2UI5 1.147.0,
tagged 2026-10-14). That release's output/ is a folder per origin
(@abaplint/transpiler 2.14), and locate( ) named the adapter by the flat
path - the conformance host died on the import with ERR_MODULE_NOT_FOUND
in the nightly run against upstream main (38419982897). shimPath( ) takes
output/express-icf-shim/ first and the flat output/ of 1.146.0 and older
second, so an `overrides` to an older runtime still boots, and a package
with neither says so instead of failing on the import.

The nightly's other failure, "the framework's 500 does not reflect the
client's url", was the same mismatch from the other side: runtime/ held
1.147.0 while the plugin's nested node_modules kept the pinned 1.146.0,
which predates the core fix. With 1.147.0 resolved everywhere the test
passes; lint and the 222 bookshop tests are green here.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TrfF51pia7m6tLZQF2G2ua
…ne has it

The stand-ins symlink the real runtime's output/ and hard-coded the flat
shim path, which the 1.147.0 layout (output/express-icf-shim/) does not
have; the seven tests died on the import. The path is taken from what
locate( ) found, so the fixture follows either layout.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01TrfF51pia7m6tLZQF2G2ua
@oblomov-dev
oblomov-dev merged commit 61c7604 into main Oct 14, 2026
5 checks passed
@oblomov-dev
oblomov-dev deleted the claude/inspiring-volta-dh7djv branch October 14, 2026 07:22
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants