On Thu, Aug 01, 2019 at 05:11:47PM -0400, Jason Merrill wrote:
> On 8/1/19 3:50 PM, Marek Polacek wrote:
> > We started rejecting this test with r268321, whereby we process the
> > compound_literal in finish_compound_literal normally even in a template
> > if it's not instantiation-dependent. This broke with __PRETTY_FUNCTION__
> > in a template function, because instantiation_dependent_expression_p didn't
> > consider it dependent, which resulted in a bogus error later.
> >
> > cp_make_fname_decl has TYPE_DEP indicating that the fname is dependent, so
> > I think value_dependent_expression_p needs to be fixed as below, and
> > actually
> > treat such a fname as dependent.
> >
> > Bootstrapped/regtested on x86_64-linux, ok for trunk and 9?
> >
> > 2019-08-01 Marek Polacek <pola...@redhat.com>
> >
> > PR c++/91230 - wrong error with __PRETTY_FUNCTION__ and generic lambda.
> > * pt.c (value_dependent_expression_p): Consider __PRETTY_FUNCTION__
> > inside a template function value-dependent.
> >
> > * g++.dg/cpp1y/lambda-generic-pretty1.C: New test.
> >
> > diff --git gcc/cp/pt.c gcc/cp/pt.c
> > index 91a46745447..94706bc5ad1 100644
> > --- gcc/cp/pt.c
> > +++ gcc/cp/pt.c
> > @@ -25553,7 +25553,11 @@ value_dependent_expression_p (tree expression)
> > if (DECL_HAS_VALUE_EXPR_P (expression))
> > {
> > tree value_expr = DECL_VALUE_EXPR (expression);
> > - if (value_dependent_expression_p (value_expr))
> > + if (value_dependent_expression_p (value_expr)
> > + /* __PRETTY_FUNCTION__ inside a template function is dependent
> > + on the name of the function. */
> > + || (DECL_PRETTY_FUNCTION_P (expression)
> > + && value_expr == error_mark_node))
>
> Hmm, when would value_expr not be error_mark_node for __PRETTY_FUNCTION__ in
> a template?
E.g.
template<typename T>
struct S {
T *t{__PRETTY_FUNCTION__};
};
S<const char> s;
here cp_make_fname_decl is called in a template, but in_template_function() is
false, so we create a decl whose DECL_VALUE_EXPR is "top level".
Marek
