On Wed, Jul 17, 2013 at 5:41 PM, Adrian Prantl <[email protected]> wrote: > > On Jul 17, 2013, at 5:35 PM, Eric Christopher <[email protected]> wrote: > >> Without looking too hard on my end I think this is something we might >> want to use in EmitFunctionStart if we don't have a valid loc for the >> function? > > The whole topic is really finicky. Currently CGDebugInfo actually _ignores_ > unknown locations and reuses the last one. I would really like to reconsider > this behavior, because then we wouldn’t need hacks like this. But I’m also > aware that changing this default will break the line table in lots of subtle > ways and we don’t necessarily have the tests that would catch this. >
I'm definitely up with revisiting the decision there. I don't know how much work it is, but the gdb testsuite is pretty good at figuring out finicky line problems. And hey, an excuse for more tests. ;) -eric _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
