On Sat, Sep 3, 2011 at 5:31 PM, Michael Matz <m...@suse.de> wrote: > Hi, > > On Sat, 3 Sep 2011, Richard Guenther wrote: > >> >> As I feared the call to get_var_ann in set_is_used right now really >> >> is still needed, privatizing it hence isn't that straight forward. >> > >> > After pondering I concluded that it's not necessary to call >> > set_is_used for variables without var annotation. Those won't be in >> > referenced_vars (or local_decls) and therefore also won't be removed >> > from those lists no matter how hard we try. >> > >> > Regstrapped on x86_64-linux (without Ada). Okay for trunk? >> >> No. We call mark_all_vars_used on trees contained in GIMPLE non-debug >> statements. All vars we can reach that way _have_ to be in the list of >> referenced vars. > > Sometimes global variables (non-auto vars with context != cfun) can be > missing from referenced_vars. In the case where we hit bugs with this > it's the non-local variables generated for profile counters. I can add > the respective add_referenced_var calls for that, but I'm not sure I see > the point. That of course comes back to our old discussion for what > purposes referenced_vars actually is used. I looked at all users and I > think that the global counters missing from referenced_vars is harmless. > > OTOH it's a nice invariant that can actually be checked for (that all > reachable vars whatsoever have to be in referenced_vars), so I'm going to > do that.
Yes, until we get rid of referenced_vars (which we still should do at some point...) that's the best. IIRC we have some verification code even, and wonder why it doesn't trigger. Richard. >> That they are not is the bug. So - can you investigate >> which var doesn't have an annotation an why it isn't in referenced vars? > > > Ciao, > Michael.
