(1) Build analysis culprit identification:
There is some problem with the build sensor.
In version 6 build script, javac is executed in local.build.xml, all
junit and checkstyle are executed in build.util.xml file (in
hackyBuild). Therefore, only for compilation task the workingDirectory
is correct. For junit and checkstyle task, the working directory is
always hackyBuild. The problem can be overcome by using a new listener
interface in ant 1.6. Another minor problem is that the sensor always
sends the absolute path of the working directory, but the analysis
expects only module. The result is the culprit is alaways NULL. Since
we will be using version 7 in a week or two, I am not going to do
anything with the version 6 build. I'll just edit the sensor data on the
server each day for the next week.
I have modified the sensor to work with version 7 build script. The
change has been committed. It should work, but I have not tested it yet.
(2) JBlanket
The code is really complicated. I know it rewrites byte code after javac
and before junit, but I still have not figured out how everything's put
together exactly. It'll take sometime.
Cheers,
Cedric
Philip Johnson wrote:
A couple of things:
(1) The failure is the infamous JBlanket problem:
csdl.jblanket.JBlanketException: Unable to read file
c:\temp\jb\COVER-org.hackystat.kernel.installer.view.swing.main.TestHostKey
ButtonPanel.xml
Tim--do you have any advice for Cedric as he delves into the code?
Are we even using the most recent version of JBlanket on hackydev?
(2) The BuildAnalysis report shows only one committer (me) but no
culprit identified:
Integration build for hacky2004-all project on 16-Oct-2005 FAILED.
[C:\CruiseControls\instance-ALL\hacky\hackyBuild] [BuildScript, ]
[Plausible Culprit: N/A]
Individual Developer Commit and Test Behavior: TotalCommitBatch
[FullyTested / PartiallyTested / UnTested / FailedTest]
[EMAIL PROTECTED]: 1 [0 / 0 / 1 / 0]
That seems weird in a way--couldn't one infer that if there's only one
committer, that person might be the culprit? (Of course, it would have
been wrong.) On the other hand, the failure was in JBlanket. Couldn't
the build analysis report a bit more specifically about the error
(i.e. JBlanket target failure)? Could it figure out it was an internal
build problem?
Cheers,
Philip
--On Sunday, October 16, 2005 3:22 AM -1000 Qin Zhang
<[EMAIL PROTECTED]> wrote:
Hackystat build (configuration Hackystat-ALL) failed.
Build report is available at
http://xenia.ics.hawaii.edu/hackyDevSite/configurationBuildReport.do?year
=2005&month=10&day=16&configuration=Hackystat-ALL Build Time Stamp: Sun
Oct 16 03:22:49 HST 2005
