Corrections are committed. Sorry, guys. -Cedric

Philip Johnson wrote:


--On Monday, February 14, 2005 9:19 AM -1000 "(Cedric) Qin ZHANG"
<[EMAIL PROTECTED]> wrote:

I did freshStart and junitAll for all module before committing.

The problem is, when testing build sensor, I set
   checkstyle.failOnViolation=false,

So when junitAll finishes, all checkstyle errors have already scrolled
up the console
window (nowhere to find).


Hmm. So, looks like knowing what targets are invoked won't be good
enough for your build analysis if people can basically disable their
results!  There's no point in running checkstyle, for example, if the
results are ignored.

I think the larger moral of the story is that the ability to disable
Checkstyle failure was intended for use only in the daily build
mechanism (so we could still get other reports and a distribution even
if checkstyle failed).

It's not, however, appropriate for regular developers to disable
checkstyle or junit failure. As yesterday's build illustrates,
eventually this results in project-wide problems.

So, two defect prevention-related proposals:
- Cedric, this seems to indicate to me that we were wrong in our
previous approach to defaults.  The default should always be to fail.
- Everyone should simply delete (or at least comment out) the
checkstyle.failOnViolation and the junit.haltonfailure properties in
their local hackystat.properties file.  In combination with appropriate
defaults, this will prevent the cause of last night's build failure from
occuring in future.

Cheers,
Philip

Reply via email to