slachiewicz opened a new issue, #3622:
URL: https://github.com/apache/maven-surefire/issues/3622

   Affected version: master at d7c7e359e.
   
   While reading the fork-to-master event path I noticed that the build result 
is reconstructed from frames with no application-layer producer binding. 
`EventDecoder.decode` 
(maven-surefire-common/src/main/java/org/apache/maven/surefire/stream/EventDecoder.java:124,192)
 maps magic/opcode frames to Event objects without any producer identity or run 
identifier, and `ForkClient.handleEvent` 
(maven-surefire-common/src/main/java/org/apache/maven/plugin/surefire/booterclient/output/ForkClient.java:347)
 dispatches them into reporter state and lifecycle flags. `ForkStarter` derives 
the result from that reporter state (around lines 686-735).
   
   The only gate is possession of the fork-channel session token. 
`SystemPropertiesListener.handle` (ForkClient.java:219) records fork-supplied 
properties without master-side authorization, and a synthetic success/goodbye 
sequence would be accepted as genuine.
   
   Expected behavior: bind events to the run and fork that produced them (or 
authenticate each frame), and derive pass/fail from data the master 
independently observed. A regression test could use a `ForkChannelTest`-style 
loopback client with a valid session token to inject a synthetic success 
sequence and assert that a failing test is not silently reported as passing.
   


-- 
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