How to report bugs (was Re: new snapshot (may 11))

At 2006-05-15 16:46, Jan Kriesten wrote:
>Scott:
> > 1052 doesn't really give much information at all.  It just says facelets
> > doesn't work, which means that for us to track it down, we would need to
> > download facelets, create a test case, figure out how to reproduce it,
> > and then track down the actual API that's failing.
>
>hmm, i think this as a complaint about how users should report bugs.
>
>actually, as i'm a programmer, too, i have a different view on the 
>bug-report of
>  what is actually reported in 1052.
>
>it doesn't say 'facelets don't work', it says that the META-INF isn't searched
>by the classloader for requested files. this is imo as detailed a description
>can be.
>
>i'm a bit puzzled by your post. emerging technologies should be looked into by
>you from time to time, so you should encounter similar problems yourself. it's
>quite strange to me to let your customers report mis-behaviors and then just
>scratch the problem on the surface and even don't try to run 
>facelets (which is
>really a download/run thing!).
>
>... for a programmer a description as reported is - imo - reproducable.


I'd actually have to agree with Scott on this issue, at least partially.
Scott is a human like everyone else on this list, and thus has 
limited time to spend on coding Resin. If he has to choose between 
trying to reproduce a bug or fix other bugs, I'd have to vote for the 
other bugs. That is why it is to important to make the bug easily reproducible.

You say the Facelets thingy is just download + run. Firstly, I am not 
able to read this out of the bug report (honestly I cannot personally 
understand how to reproduce the error from 
http://bugs.caucho.com/view.php?id=1052 - what is the corresponding 
method call to "file-loading requests"?). So when you look at this 
report, you cannot decide whether you need to spend 5 minutes or 5 
hours trying to reproduce. Secondly, if Scott has not worked with 
Facelets before (although it is quite possible he has) he'd have to 
Google for it and download it. It only takes a couple of minutes - 
bug again, this time could be spent fixing *other* things. In most 
cases it is preferrable if the reporter would attach a minimized test 
case (such as a WAR including sources) to the bug report, that could 
be dropped in to reproduce the bug. Oftenly this only takes the 
reporter a maximum of 1-2 hours to do, but increases probability that 
the bug will be fixed (correctly) and decreases the time for it to be 
fixed. This also makes it easy for the reporter to verify that the 
bug is fixed in the next release/snapshop.

Besides, if you have spent time debugging the problem and finally 
concluding the problem lies within Resin, you are much more "there" 
mentally, and thus in a better position to provide the test case, 
than for Scott to start from the beginning, trying to interpret the 
error description.

Now, if you have support subscriptions I may be of another opinion 
regarding the above, but Scott still cannot do mind reading no matter 
how much $$$ you send to Caucho.