aglinxinyuan opened a new pull request, #6951:
URL: https://github.com/apache/texera/pull/6951

   ### What changes were proposed in this PR?
   
   The `backport (release/*)` CI legs rebuild the workspace as the release 
target tree (`prepare-backport-checkout.sh`) while executing **main's** 
`build.yml`. Steps that invoke `.github/scripts/smoke-boot.sh` therefore fail 
with exit 127 when the target branch predates the harness (added in #6319): the 
script is simply not in the checked-out tree. `release/v1.2` is in that state 
today, and the backport leg of #6910 — the first `release/*`-labeled PR to fire 
`amber-integration` since the harness landed — is blocked by it:
   
   ```
   .github/scripts/smoke-boot.sh: No such file or directory
   ##[error]Process completed with exit code 127.
   ```
   (https://github.com/apache/texera/actions/runs/30297805997/job/90102731299)
   
   This PR keys every boot-smoke step on the script's presence in the 
checked-out tree, following the workflow's existing tree-adaptive pattern (`if 
[ -f amber/requirements.txt ]`):
   
   - `amber-integration`: the dist/unzip/smoke-test unit (4 steps) gains 
`hashFiles('.github/scripts/smoke-boot.sh') != ''` alongside the existing 
ubuntu-only condition.
   - `platform-integration`: the whole job is the boot smoke test, so its 
dist/unzip/MinIO/LakeFS/smoke steps (5 steps) gain the same guard.
   
   `hashFiles` is evaluated on the runner against the workspace *after* the 
backport tree swap, so main/PR runs are unaffected (script present → everything 
runs exactly as before), while backport legs against pre-harness targets skip 
the steps visibly instead of failing. The guard self-retires the moment 
`smoke-boot.sh` is backported to the target branch — doing that for 
`release/v1.2` to restore boot-smoke coverage on its backport legs is a 
sensible follow-up, but shouldn't hold the currently blocked security backport 
hostage.
   
   ### Any related issues, documentation, discussions?
   
   - Unblocks the `release/v1.2` backport leg of #6910 (torch CVE bump).
   - Harness introduced in #6319 / #6274; never backported to `release/v1.2`.
   
   ### How was this PR tested?
   
   - `build.yml` parses (`yaml.safe_load`); expression semantics reviewed: 
`matrix.object_store && hashFiles(...) != ''` preserves the skip for 
non-object-store services (`null && …` stays falsy).
   - The script-present path is exercised by this PR's own CI: touching 
`.github/**` applies the `ci` label, so all stacks including both smoke jobs 
run here with the guard evaluating true.
   - The skip path can't run pre-merge (it needs a backport leg against 
`release/v1.2`); after merge it can be observed by re-triggering #6910's 
Required Checks (e.g. remove and re-add its `release/v1.2` label).
   
   ### Was this PR authored or co-authored using generative AI tooling?
   
   Co-authored with Claude Fable 5 in compliance with ASF.


-- 
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