On 11/20/2010 12:47 AM, Sim IJskes - QCG wrote:
On 11/19/2010 10:51 PM, Patricia Shanahan wrote:
I'm having a problem with "ant run-tests" doing too much, and taking too
long to get to the test in question.
Even when run immediately after an all.build, so that all jars are up to
date, it goes through a lot of deleting and build jars.
Can anything be done about this? Is there some alternative target I can
use that does not force rebuilding? Quick running of a single test is
very important for debug.
I'm assuming you work with the latest version. Do you want to keep
running a single test in a loop? Thats the only reason i can think of
where you wouldn't like to build the harness-runtime and underlying
river-runtime.
I usually update to the latest version between debug efforts. I tend not
to update in the middle of a debug effort so that I don't confuse
matters by getting changes due to check-ins.
See http://www.patriciashanahan.com/debug/index.html for how I approach
debug. In the debug loop, I think of a theory about what is going wrong,
design an experiment to test it, and run the experiment.
Running the experiment may require source code changes, but often only
needs changes in break point placement, or the logging configuration
file, or even just in the questions I intend to ask when the program
reaches an existing break point.
And, yes, there are times that I run the same tests repeatedly in a
"while true" loop. For example, I ran the tests that fail on Peter's
system but not mine a couple of hundred times, to see if it were a
timing problem with different failure frequency.
The alternative is to make the rebuild really, really fast if there are
no source code changes. I know you have been making valuable progress
in that direction.
Patricia
