On Sun, 2011-02-20 at 08:43 +0100, ext Julia Lawall wrote:
> On Sun, 20 Feb 2011, Oliver McFadden wrote:
> 
> > Hi,
> > 
> > I have just installed the latest spatch release hoping this might be a
> > bug in my older version, 0.2.3, but I'm now using 0.2.5-rc6, and I have
> > the same problem.
> > 
> > The idea was to expand on the "drop unnecessary continue" SmPL
> > (http://coccinelle.lip6.fr/impact/continue.html)
> > 
> > This works fine on the simple test code (test.c) but when it's applied
> > onto a larger code base, it hangs. I've attached the debug output from
> > test.c (good.txt) and the debug output from a file in my code base.
> > (bad.txt) I have to killed the bad spatch after about 30 seconds.
> 
> Unfortnuately, it is the bad file I would need to see.  Perhaps you can 
> localize the problem with the argument -show_trying, and then make an 
> example that illustrates the problem?

Yes, I will try this option tonight as soon as I have time and send you
the results along with code.

> It is possible that it is just very slow.  When there are many nested 
> loops with many nested conditionals, then it is following all of the 
> control-flow paths, which can take a lot of time.

Perhaps in such cases spatch could output a "spinner" showing it's still
working and (slowly) making progress (something like the wget progress
bar?)

> You can use the options 
> -no_loops and -no_gotos which makes it not follow paths through loops or 
> gotos.  This is not safe in general, but it should be ok for what is to be 
> done here.

Right, but I wouldn't really want to be telling an analysis tool to
ignore major features of a language. Seems like a bad idea in my
opinion.

> 
> julia


_______________________________________________
Cocci mailing list
[email protected]
http://lists.diku.dk/mailman/listinfo/cocci
(Web access from inside DIKUs LAN only)

Reply via email to