Problem 1.
[jblanketreport] [Fatal Error] COVER-MethodSets.xml:86:88: The value of
attribute "method" associated with an element type "Method" must not contain the
'<' character.
[jblanketreport] csdl.jblanket.JBlanketException: Unable to transform XML
to HTML

Hmm. This one is definitely fixed. That's what broke the Hackystat build over the weekend. Are you sure you've got the latest jblanket.jar? Are you sure it's the only version of the jar available in the classpath?

I agree with Tim---this is almost certainly due to the fact that you've got an old jblanket.jar.

Get the latest one from:

<http://csdl.ics.hawaii.edu/Tools/JBlanket/dist/lib/ant/jblanket.jar>

I just downloaded the one from the CSDL website, and this is what I get for
diagnostics.

C:\>  java -jar jblanket.jar SysInfo
JBlanket Release: @release@

That's irritating. There was a bug in jblanket's build.xml. I fixed it, committed the change, and updated the CSDL jblanket website, and now you should see:

C:\cvs\JBlanket\lib\ant>java -jar jblanket.jar
JBlanket Release: 4.4.0309

Problem 2.
When I sent a previous email, stating that "I concurred that the problem
was fixed" I didn't do a junitAll. Philip did you do one?

Yes. As confirmation, I just checked last night's hackystat-ALL:

<http://hackydev.ics.hawaii.edu/hackyDevSite/build_log/20050309/Hackystat-ALL/ant/antlog/index.html>

You can see that JBlanket worked successfully, and that the deploySoap target 
did not
produce errors. Yeah!

Here is another problem that I've found. this is major a one. I've noticed
that when I enable Jblanket the building and testing (freshStart junitAll)
there is an significant increase (more than 80 minutes on my brand new 3.4
Ghz computer) in the testing time. Here is my supporting data.

I haven't done any testing of my own with JBlanket's overhead, but this sounds pretty serious. Our project currently can't use JBlanket since we've switched over to using only JDK 1.5 (we use Eclipse 3.1m5a), so I can't check it on any large project.

The more I've worked with the JBlanket code, the more I think it's
nearing time to try a new architecture.  Given the pain that's
happened here, I certainly won't just spring it on anyone.

Could be. We did try a few different approaches and found that byte code modification produced the fastest runs (way better than JDI or JVMPI). What kind of 'new architecture' are you thinking of?

If it were me, the first thing I would do is profile JBlanket on a minimal 
hackystat
configuration (say, just hackyKernel). I wouldn't be surprised if some fairly 
small
changes produced some fairly significant speedups, given that we haven't looked 
at
performance for quite a while.

Just out of curiosity (at this point), how come JBlanket won't compile/run/work 
under JDK
1.5?

Cheers,
Philip

Reply via email to