On Mar 20, 2012, at 22:37, Ted Kremenek wrote: > On Mar 20, 2012, at 5:28 PM, Jordan Rose <[email protected]> wrote: > >>>> Oh, it's not a visitor callback -- the interesting symbols and regions are >>>> stored in the BugReport. So it's just communication between the BugReport >>>> and the BugReporter. >>> >>> Sorry, You are right. I am a bit concerned with this method. It assumes >>> that the symbols are never deleted from the set. Is this currently true? >> >> Yes, it is. If that changes, the implementation of this method will of >> course have to change. Is it worth proactively adding a change counter to >> BugReport? > > I'm missing something. Why do we need a change counter, or anything, in > BugReport? What we're suggesting should only impact some logic in > BugReporter, and maybe something that allows us to roll back the > BugReporterVisitor state, but both of these can be done without changing > BugReport.
Interesting symbols/regions are per-BugReport, and as such only the BugReport knows when they change. BugReporterVisitor needs to be able to see that to decide whether to start over. _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
