On Mar 20, 2012, at 5:28 PM, Jordan Rose <[email protected]> wrote:

> Fair enough. I'll go rework this. Advance question: is it better to wait 
> until the whole path is generated before restarting, or to restart as soon as 
> we see the interesting symbols have changed? Both ways could result in a 
> second round of new interesting symbols.

I would just do the simplest thing possible.  Generate the entire path by 
wrapping the calls to 
GenerateExtensivePathDiagnostic/GenerateMinimalPathDiagnostic in a while loop, 
check if the set of symbols have changed, and if so, drop the old path and 
repeat.

The performance issue here is likely non-existent.  It takes a couple orders of 
magnitude more work to find the bug in the first place.
_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits

Reply via email to