On Tue, 2022-07-19 at 14:18 -0400, David Malcolm wrote:
> On Tue, 2022-07-19 at 21:36 +0530, Immad Mir wrote:
>
[...snip...]
>
>
> > +void
> > +fd_state_machine::check_for_fd_attrs (
> > + sm_context *sm_ctxt, const supernode *node, const gimple
> > *stmt,
> > + const tree callee_fndecl, const char *attr_name,
> > + access_directions fd_attr_access_dir) const
> > +{
I had another idea about the patch: we have attr_name here, and thus I
think it's available when creating the various fd_param_diagnostic
instances, so why not convert the:
bool m_attr;
in fd_param_diagnostic to:
const char *m_attr_name; // can be NULL if no attribute involved
and then, when it's non-NULL we can use it directly in the inform
message, so that the attribute name we report is the one that the user
used. I think that's clearer, both for us and for the end-user.
Dave
