On Nov 12, 2013, at 22:46 , Richard Smith <[email protected]> wrote:
> The important question is what's the timing difference between:
> -Wuninitialized -Wyour-new-warning and just -Wuninitialized
>
> That way you won't be comparing the timing of CFG building and no CFG
> building, but CFG building with your warning and CFG building without your
> warning.
>
> Right. If this warning doesn't add measurable compile time in a build that's
> already constructing CFGs, then that's OK, but it needs to be off by default,
> like the other CFG-based warnings.
>
…which raises the point: should we have a -Wflow-sensitive-analysis for “all
current and future CFG-based checks”? I have the usual discoverability
concerns, but having a warning flag that means “warnings that would be on by
default if we could get a CFG for free” seems like a way to ameliorate that
issue.
Jordan
_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
