W dniu 26 października 2010 00:39 użytkownik Ted Kremenek <
[email protected]> napisał:

>
> On Oct 25, 2010, at 3:25 PM, Marcin Świderski wrote:
>
> > Because:
> > 1. removing the check after swaping the order did not crash tests,
>
> The existing test probably only checked the inverse case.  When in doubt,
> it is worth verifying the original test that failed (before your changes, by
> removing the sanity checking), and seeing if it needs to modified/copied to
> the new case.  This kind of sanity checking in the CFGBuilder was not added
> arbitrarily; it has been built up from esoteric cases people have provided
> over the last couple years.
>
> Crash (not fail) is in misc-ps-region-store.m somewhere in VisitForStmt. I
could not locate it, would need to isolate every instance of for loop in
file and test it independently, but it's not for today (for ma at least).


> > 2. VisitChildren() is implemented this way,
> > I assumed that this is the right way to go. If I'm mistaken then
> shouldn't we add such check in other places with similar code e.g.
> VisitChildren()?
>
> Good point.  It probably needs to be modified as well.


If we want to be on the safe side. But I'm really intrigued by this.
_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits

Reply via email to