public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
* C++ PATCH for c++/88196, ICE with class non-type template parameter
@ 2018-12-20 22:19 Marek Polacek
  2018-12-21  2:51 ` Jason Merrill
  0 siblings, 1 reply; 2+ messages in thread
From: Marek Polacek @ 2018-12-20 22:19 UTC (permalink / raw)
  To: GCC Patches, Jason Merrill

This testcase breaks in C++20 because convert_nontype_argument gets a PTRMEM_CST
as the expression, and we're trying to convert it to a class type, which
doesn't work, but because of the PTRMEM_CST check below we never diagnosed it.

The comment says a PTRMEM_CST is a valid template argument for a parameter of
pointer to member type, but it doesn't check that that's what we're converting
it to.  Fixed by adding such a check.

Bootstrapped/regtested on x86_64-linux, ok for trunk?

2018-12-20  Marek Polacek  <polacek@redhat.com>

	PR c++/88196 - ICE with class non-type template parameter.
	* pt.c (convert_nontype_argument): If the expr is a PTRMEM_CST, also
	check if the type we're converting it to is TYPE_PTRMEM_P.

	* g++.dg/cpp0x/ptrmem-cst-arg1.C: Tweak dg-error.
	* g++.dg/cpp2a/nontype-class10.C: New test.
	* g++.dg/template/pr54858.C: Tweak dg-error.

diff --git gcc/cp/pt.c gcc/cp/pt.c
index d4ac034671a..18b093e7d2d 100644
--- gcc/cp/pt.c
+++ gcc/cp/pt.c
@@ -6775,7 +6775,7 @@ convert_nontype_argument (tree type, tree expr, tsubst_flags_t complain)
      to a null value, but otherwise still need to be of a specific form.  */
   if (cxx_dialect >= cxx11)
     {
-      if (TREE_CODE (expr) == PTRMEM_CST)
+      if (TREE_CODE (expr) == PTRMEM_CST && TYPE_PTRMEM_P (type))
 	/* A PTRMEM_CST is already constant, and a valid template
 	   argument for a parameter of pointer to member type, we just want
 	   to leave it in that form rather than lower it to a
diff --git gcc/testsuite/g++.dg/cpp0x/ptrmem-cst-arg1.C gcc/testsuite/g++.dg/cpp0x/ptrmem-cst-arg1.C
index ed12655af32..38b3e24241c 100644
--- gcc/testsuite/g++.dg/cpp0x/ptrmem-cst-arg1.C
+++ gcc/testsuite/g++.dg/cpp0x/ptrmem-cst-arg1.C
@@ -5,5 +5,5 @@ template<int> struct A {};
 struct B
 {
     int i;
-    A<&B::i> a; // { dg-error "could not convert template argument" }
+    A<&B::i> a; // { dg-error "could not convert" }
 };
diff --git gcc/testsuite/g++.dg/cpp2a/nontype-class10.C gcc/testsuite/g++.dg/cpp2a/nontype-class10.C
new file mode 100644
index 00000000000..5cc179300ca
--- /dev/null
+++ gcc/testsuite/g++.dg/cpp2a/nontype-class10.C
@@ -0,0 +1,6 @@
+// PR c++/88196
+// { dg-do compile { target c++2a } }
+
+struct C { C *c; };
+template <C> struct D;
+D <&C::c> d; // { dg-error "could not convert" }
diff --git gcc/testsuite/g++.dg/template/pr54858.C gcc/testsuite/g++.dg/template/pr54858.C
index 51610ad689a..6d23decfbb9 100644
--- gcc/testsuite/g++.dg/template/pr54858.C
+++ gcc/testsuite/g++.dg/template/pr54858.C
@@ -5,17 +5,17 @@ template <int> struct A {};
 template <typename T, T *> struct B {};
 template <typename D> struct C
 {
-  A<0> c0; B<A<0>, &C::c0> d0;	// { dg-error "could not convert template argument" }
-  A<0> c1; B<A<0>, &C::c1> d1;	// { dg-error "could not convert template argument" }
-  A<0> c2; B<A<0>, &C::c2> d2;	// { dg-error "could not convert template argument" }
-  A<0> c3; B<A<0>, &C::c3> d3;	// { dg-error "could not convert template argument" }
-  A<0> c4; B<A<0>, &C::c4> d4;	// { dg-error "could not convert template argument" }
-  A<0> c5; B<A<0>, &C::c5> d5;	// { dg-error "could not convert template argument" }
-  A<0> c6; B<A<0>, &C::c6> d6;	// { dg-error "could not convert template argument" }
-  A<0> c7; B<A<0>, &C::c7> d7;	// { dg-error "could not convert template argument" }
-  A<0> c8; B<A<0>, &C::c8> d8;	// { dg-error "could not convert template argument" }
-  A<0> c9; B<A<0>, &C::c9> d9;	// { dg-error "could not convert template argument" }
-  A<0> ca; B<A<0>, &C::ca> da;	// { dg-error "could not convert template argument" }
-  A<0> cb; B<A<0>, &C::cb> db;	// { dg-error "could not convert template argument" }
+  A<0> c0; B<A<0>, &C::c0> d0;	// { dg-error "could not convert" }
+  A<0> c1; B<A<0>, &C::c1> d1;	// { dg-error "could not convert" }
+  A<0> c2; B<A<0>, &C::c2> d2;	// { dg-error "could not convert" }
+  A<0> c3; B<A<0>, &C::c3> d3;	// { dg-error "could not convert" }
+  A<0> c4; B<A<0>, &C::c4> d4;	// { dg-error "could not convert" }
+  A<0> c5; B<A<0>, &C::c5> d5;	// { dg-error "could not convert" }
+  A<0> c6; B<A<0>, &C::c6> d6;	// { dg-error "could not convert" }
+  A<0> c7; B<A<0>, &C::c7> d7;	// { dg-error "could not convert" }
+  A<0> c8; B<A<0>, &C::c8> d8;	// { dg-error "could not convert" }
+  A<0> c9; B<A<0>, &C::c9> d9;	// { dg-error "could not convert" }
+  A<0> ca; B<A<0>, &C::ca> da;	// { dg-error "could not convert" }
+  A<0> cb; B<A<0>, &C::cb> db;	// { dg-error "could not convert" }
 };
 C<int> e;

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: C++ PATCH for c++/88196, ICE with class non-type template parameter
  2018-12-20 22:19 C++ PATCH for c++/88196, ICE with class non-type template parameter Marek Polacek
@ 2018-12-21  2:51 ` Jason Merrill
  0 siblings, 0 replies; 2+ messages in thread
From: Jason Merrill @ 2018-12-21  2:51 UTC (permalink / raw)
  To: Marek Polacek, GCC Patches

On 12/20/18 4:54 PM, Marek Polacek wrote:
> This testcase breaks in C++20 because convert_nontype_argument gets a PTRMEM_CST
> as the expression, and we're trying to convert it to a class type, which
> doesn't work, but because of the PTRMEM_CST check below we never diagnosed it.
> 
> The comment says a PTRMEM_CST is a valid template argument for a parameter of
> pointer to member type, but it doesn't check that that's what we're converting
> it to.  Fixed by adding such a check.
> 
> Bootstrapped/regtested on x86_64-linux, ok for trunk?

OK.

Jason

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2018-12-21  2:51 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2018-12-20 22:19 C++ PATCH for c++/88196, ICE with class non-type template parameter Marek Polacek
2018-12-21  2:51 ` Jason Merrill

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).