Re: [LTP] Non-terminating behavior in pan

[<nat...@re...>]

On Fri, Jun 16, 2006 at 02:55:33PM -0600, Andrew Echols wrote:
> My colleagues and I have run into what appeared to be a bug when trying to
> run individual test cases in ltp-base-20060515.  If the test list (with
> runltp -f) or the test pattern (with runltp -s) refers to a test that
> doesn't exist, it would loop forever.   

Yup, looks like you ran over a real bug in pan.

> Here's how I percieve it should work:
> 
>     459             cpid = run_child(coll->ary[c], running + i,
> quiet_mode);
>     460             if (cpid != -1)
>     461                 ++num_active;
>     462             if ((cpid != -1 || sequential) && starts > 0)
>     463                 --starts;

That looks like a reasonable fix.  

> Is this a bug or is it actually the technically "correct" behavior? 

Most of the cases I see where you use sequential runs, you want to run
the whole list, nothing more, nothing less.  It's "correct" by the way
the code it written ;)  but IMHO not how to tool should be working.

> (See attached file: pan_loop_fix.diff)

Please apply to CVS.

Nate