Ken Sedgwick writes:

> Is there a regression testing suite for Freenet? [...] I would 
> like to start building one if this effort is not already
> underway ...

I'm pretty new to the list, but I'm fairly sure they don't have 
a full-blown regression test suite yet.  The "Serapis" 
simulation stuff seems to be closely related, though.  That 
should be a good place to start looking.

Must... fight... urge... to be... verbose...

(Loses fight, gives in to the Digression Side)

One opinion: it seems that automatically generating regression 
testing scripts is a process quite similar to automatically 
generating user-activity logs for tutorial, help, and support 
purposes.  For example, setting up a "record and replay" script 
for in-house testing is similar -- if only in priciple -- to 
maintaining an ongoing activity log in the field.  Such logs 
could be used (for example) by a bug-report program to help the 
user more quickly find the relevant pre-existing support files.  
This sort of thing is especially appealing to "free" apps that 
don't have a paid support staff to direct users to the manual.  
If no appropriate existing help page is found, the data could 
then be included when generating a new bug report.

In this particular application, of course, we'll have to be 
extra careful about activity tracking.  We certainly don't want 
some silently-generated plaintext activity log lying around the 
user's system without their knowledge.  The "Start, Documents" 
feature in Windows comes to mind.  

<fantasy> I think it's possible, in at least some cases, that 
we could use things like activity logs and a carefully-crafted 
"program state vector" to automatically re-create the relevant 
aspects of a reported field bug.  We might even be able to 
narrow down the problem by using the regression-testing trick 
of automatically running several variants of the regression 
test (activity log) with some parts of the script deleted.  All 
this before the coder even looks at it!

Some coding shops already do this sort of thing, and others at 
least have it in their future plans.  Is it within reason to 
hope that we could do it here?  The more the program behaves 
like a kind of state vector, driven by simple serial commands 
that perform a transform on the vector, the more this seems 
possible.  Of course, many programs are not easily modelled 
that way -- at least not in their entirety.  Hmm...  </fantasy>

In other words, when we develop a general strategy for 
regression testing, we may want to simultaneously develop a 
strategy for things like user bug reports.  I am willing to 
help work on regression testing & related stuff.  Contact me 
privately if you wish.    

--Will
willdye at freedom.net


_______________________________________________
Freenet-dev mailing list
Freenet-dev at lists.sourceforge.net
http://lists.sourceforge.net/mailman/listinfo/freenet-dev

Reply via email to