about the hudson server

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

about the hudson server

stephane ducasse
Hi yanni

I wanted to know if you can tell us what means (unstable) in the report summary

        http://hudson.jooshr.org/job/Pharo1.1/81/
is it because the tests cannot run?

Stef
_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Reply | Threaded
Open this post in threaded view
|

Re: about the hudson server

Yanni Chiu
stephane ducasse wrote:
> Hi yanni
>
> I wanted to know if you can tell us what means (unstable) in the report summary
>
> http://hudson.jooshr.org/job/Pharo1.1/81/
> is it because the tests cannot run?

Here's my interpretation.

The status of the last build (the big ball in the first column) is
important:

blue   - the code built okay, and all tests passed
yellow - the code built okay, but some tests failed
grey   - the build/test was aborted (either manually,
          or by the Hudson build timeout plugin I installed)

The sun/cloud icon in the second column is some fuzzy indicator that
uses the last few builds in its calculation.

In the case of Pharo1.1 build 81 (your link above), I don't see an
"unstable" report. I see the "Aborted" when I hover over the grey ball.
What's confusing is that if you hover over the shiny sun icon, you get a
"Build stability: no recent builds failed". That's misleading, because
the most recent build was aborted - but that does not count as a
failure. Now that I think about it, there might be an option on the
timeout plugin, to cause a build failure, when it's been aborted.

The timeout happens (and therefore an aborted build), when:
- PharoDebug.log is generated and debugger is opened
- during the build (i.e. loading packages or running scripts) a popup
appears that requires a user response
- when running tests, user input is required
- long running test exceeds the timeout that was set

I usually look at the workspace to see what's happening, or has
happened. I watch the package-cache or the size of .changes file to see
if code is still loading. I watch the size of .xml test result files, to
see if tests are still running. "watch" means click the refresh every
few seconds.

Back to build 81, the console output has:

   Build timed out. Aborting
   Build was aborted
   Archiving artifacts
   Recording test results
   Test reports were found but none of them are new. Did tests run?
   For example,
/home/yanni/.hudson/jobs/Pharo1.1/workspace/core11-tests/AllTestCases.xml
is 23 hr old

   Finished: ABORTED

Which means that it's stuck during the package load, since the test .xml
was never overwritten.

At this point, I would download core11 from the workspace, using the
"all files in zip" link. Open the image locally, and filein the
core11.st script, to see what happened.


_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
Reply | Threaded
Open this post in threaded view
|

Re: about the hudson server

Yanni Chiu
In reply to this post by stephane ducasse
stephane ducasse wrote:
> http://hudson.jooshr.org/job/Pharo1.1/81/
> is it because the tests cannot run?

It builds and runs the tests now (with some failure reports).

In build#82, I saw in the stack trace some EOCD message, which usually
means some kind of package missing, corruption, or download problem. I
cleared the package-cache and got the same problem.

It was updating from PharoCore-1.1-11239-UNSTABLE, so I changed to the
latest: PharoCore-1.1-11254-UNSTABLE.1, and now the build/test works.
There probably were no updates to be applied to 11254, so of course the
build worked.


_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project