Anton Vinogradov created IGNITE-28842:
-----------------------------------------

             Summary: Restore parallel codestyle build (-T 1C) and report 
compilation failures clearly
                 Key: IGNITE-28842
                 URL: https://issues.apache.org/jira/browse/IGNITE-28842
             Project: Ignite
          Issue Type: Task
            Reporter: Anton Vinogradov
            Assignee: Anton Vinogradov


h3. Background

The "Check java code" codestyle job intermittently looked broken with ~288 
\{{cannot find symbol}} errors for annotation-processor-generated classes 
(\{{*ViewWalker}}, \{{*MessageSerializer}}, \{{IDTOSerializerFactory}}). Root 
cause: a single *real* compile error in a stale PR branch (e.g. a removed 
method still referenced after a clean text merge). javac drops all 
annotation-processor output when a compilation round has errors, so one real 
error cascades into hundreds of generated-class "cannot find symbol" errors 
that bury it.

IGNITE-28840 and IGNITE-28841 chased this as a build-infra/parallelism problem; 
both were misdiagnoses. IGNITE-28841 in particular dropped the \{{-T 1C}} 
parallel build added by IGNITE-28823. \{{-T 1C}} was never the cause (the 
failure reproduces single-threaded).

h3. Change

1. Restore \{{-T 1C}} on the codestyle / \{{test-compile}} step (re-applies the 
IGNITE-28823 speed-up).

2. On compilation failure, make the job state it plainly and surface the real 
error: a GitHub \{{::error::}} annotation ("compilation failed, not a 
checkstyle violation"), plus a collapsible group listing likely root-cause 
errors with the generated-class cascade 
(\{{*Walker}}/\{{*Serializer}}/\{{*Factory}}, \{{internal.systemview}}, 
\{{codegen.idto}}) filtered out. Checkstyle-only failures are unaffected.

h3. Verification

Filtering validated against a real failing build log (PR #13298 merge): from 
288 errors it surfaces exactly the real one — \{{IgniteTxManager:[148,1] cannot 
find symbol: static deriveSyncMode}}.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to