Some Eclipse Foundation services are deprecated, or will be soon. Please ensure you've read this important communication.

Bug 367974

Summary:

ConnectorTimeoutTest and subclasses use fixed maxIdleTime

Product:

[RT] Jetty

Reporter:

Thomas Becker <tbecker>

Component:

server

Assignee:

Thomas Becker <tbecker>

Status:

RESOLVED FIXED

QA Contact:

Severity:

minor

Priority:

CC:

jetty-inbox

Version:

unspecified

Target Milestone:

7.5.x

Hardware:

All

OS:

All

Whiteboard:

Attachments:

Description	Flags
proposed patch	none

Description Thomas Becker

2012-01-05 14:13:07 EST

Build Identifier: 

ConnectorTimeoutTests and it's subclasses successful execution depend on some brittle timing constraints. 

They all have a fixed maxIdleTimeout of 250ms hardcoded and each individual test contains some hardcoded timing information themselves based on that value.

Best solution would be to use CountDownlatches, Listeners, etc. to avoid timing issues. If that's not possible or too complicated at least stuff like Thread.sleep(someValue) should have someValue being calculated based on maxIdleTimeout instead of having them hardcoded. 

Reproducible: Always

Comment 1 Thomas Becker

2012-01-06 08:10:22 EST

Created attachment 209127 [details]
proposed patch

Comment 2 Greg Wilkins

2012-01-08 19:47:47 EST

applied.  but is there more work to do here?

Comment 3 Thomas Becker

2012-01-09 04:19:12 EST

No, can be closed.

Refactoring it to make the test execute faster is not worth the work and code clutter it'll require IMHO.

Resolving the issue.

Comment 4 Thomas Becker

2012-01-09 04:20:57 EST

Btw. I discussed with Simone that the test might be fragile as it's timing is pretty precise. A GC at the wrong time for example might make the test fail as the timeouts,etc. are pretty small.
However if we make them bigger the test will last longer. I suggest to leave it as is and if it fails on hudson, we should review. Do you agree?