diff mbox series

[v2] c++/modules: Ensure deduction guides are always reachable [PR115231]

Message ID 66b210fc.630a0220.268b03.48d0@mx.google.com
State New
Headers
Series [v2] c++/modules: Ensure deduction guides are always reachable [PR115231] |

Checks

Context Check Description
linaro-tcwg-bot/tcwg_gcc_build--master-arm success Build passed
linaro-tcwg-bot/tcwg_gcc_build--master-aarch64 success Build passed
linaro-tcwg-bot/tcwg_gcc_check--master-aarch64 success Test passed
linaro-tcwg-bot/tcwg_gcc_check--master-arm success Test passed

Commit Message

Nathaniel Shead Aug. 6, 2024, 12:03 p.m. UTC 
  On Fri, Jul 26, 2024 at 01:17:57PM -0400, Jason Merrill wrote:
> On 7/26/24 12:52 AM, Nathaniel Shead wrote:
> > On Tue, Jul 23, 2024 at 04:17:22PM -0400, Jason Merrill wrote:
> > > On 6/15/24 10:29 PM, Nathaniel Shead wrote:
> > > > Bootstrapped and regtested on x86_64-pc-linux-gnu, OK for trunk?
> > > > 
> > > > This probably isn't the most efficient approach, since we need to do
> > > > name lookup to find deduction guides for a type which will also
> > > > potentially do a bunch of pointless lazy loading from imported modules,
> > > > but I wasn't able to work out a better approach without completely
> > > > reworking how deduction guides are stored and represented.
> > > 
> > > Indeed.  We likely want to find them more directly from the template; it's
> > > not clear to me that DECL_INITIAL is used for TEMPLATE_DECL, or we could put
> > > them in an internal attribute or a separate hash table.
> > > 

I had a go at exploring some other representations of deduction guides
but I didn't really land anywhere; most things I tried would require
reimplementing a lot of the merging logic handled by name lookup
currently.  Potentially just having an additional hash table for
deduction guides maintained by pushdecl would work but that felt like
unnecessary duplication, I'm not sure that the benefits outweight the
cost of the name lookup.  (But I haven't measured this.)

> > > > -- >8 --
> > > > 
> > > > Deduction guides are represented as 'normal' functions currently, and
> > > > have no special handling in modules.  However, this causes some issues;
> > > > by [temp.deduct.guide] a deduction guide is not found by normal name
> > > > lookup and instead all reachable deduction guides for a class template
> > > > should be considered, but this does not happen currently.
> > > > 
> > > > To solve this, this patch ensures that all deduction guides are
> > > > considered exported to ensure that they are always visible to importers
> > > > if they are reachable.  Another alternative here would be to add a new
> > > > kind of "all reachable" flag to name lookup, but that is complicated by
> > > > some difficulties in handling GM entities; this may be a better way to
> > > > go if more kinds of entities end up needing this handling, however.
> > > > 
> > > > Another issue here is that because deduction guides are "unrelated"
> > > > functions, they will usually get discarded from the GMF, so this patch
> > > > ensures that when finding dependencies, GMF deduction guides will also
> > > > have bindings created.  We do this in find_dependencies so that we don't
> > > > unnecessarily create bindings for GMF deduction guides that are never
> > > > reached; for consistency we do this for *all* deduction guides, not just
> > > > GM ones.
> > > 
> > > If you fixed the dependency calculation, why do they also need to be
> > > exported?
> > 
> > Deduction guides aren't found using normal name lookup, but any
> > reachable deduction guide must be considered.  This means that even if
> > the module interface exports no declarations whatsoever, a deduction
> > guide declared in the module purview must still be considered by
> > importers.
> 
> Ah, I was missing the name lookup issue.
> 
> > The other option I've considered is adding a new "ANY_REACHABLE" flag to
> > name lookup which would also consider non-exported reachable decls.  On
> > further consideration I might actually go this way; I've been thinking
> > about how to resolve some issues adjacent to supporting textual
> > redefinitions that I believe this will be necessary for anyway, and we
> > can probably use this in tsubst_friend_class as well rather than the
> > current relatively ad-hoc solution.
> 
> There's also my hack in lookup_elaborated_type for ABI namespace types.
> 
> I'm not sure that it should be necessary to do this for redefinitions,
> though; what's the advantage over merging in check_module_override (apart
> from that needing to be fixed)?
> 
> > That said, I've realised that this patch isn't completely sufficient
> > anyway; consider:
> > 
> >    // m.cpp
> >    module;
> >    template <typename T> struct S;
> >    export module M;
> >    S(int) -> S<double>;
> > 
> >    // x.cpp
> >    template <typename T> struct S { S(int); };
> >    import M;
> >    int main() {
> >      S s(10);  // should be S<double> s;
> >    }
> > 
> > This patch doesn't correctly handle this case yet, we need to also
> > consider cases where only the deduction guide is in purview.
> 
> Indeed.
> 
> Jason
> 

Here's a small update to the patch which fixes the above bug by ensuring
that dependencies are created for purview deduction guides rather than
just being skipped entirely, and includes that testcase with dguide-3.

This patch is still using force-exported bindings rather than attempting
to do anything more clever with name lookup just yet.

Bootstrapped and regtested (so far just modules.exp) on
x86_64-pc-linux-gnu, OK for trunk if full regtest succeeds?

-- >8 --

Deduction guides are represented as 'normal' functions currently, and
have no special handling in modules.  However, this causes some issues;
by [temp.deduct.guide] a deduction guide is not found by normal name
lookup and instead all reachable deduction guides for a class template
should be considered, but this does not happen currently.

To solve this, this patch ensures that all deduction guides are
considered exported to ensure that they are always visible to importers
if they are reachable.  Another alternative here would be to add a new
kind of "all reachable" flag to name lookup, but that is complicated by
some difficulties in handling GM entities; this may be a better way to
go if more kinds of entities end up needing this handling, however.

Another issue here is that because deduction guides are "unrelated"
functions, they will usually get discarded from the GMF, so this patch
ensures that when finding dependencies, GMF deduction guides will also
have bindings created.  We do this in find_dependencies so that we don't
unnecessarily create bindings for GMF deduction guides that are never
reached; for consistency we do this for *all* deduction guides, not just
GM ones.  We also make sure that the opposite (a deduction guide being
the only purview reference to a GMF template) correctly marks it as
reachable.

Finally, when merging deduction guides from multiple modules, the name
lookup code may now return two-dimensional overload sets, so update
callers to match.

As a small drive-by improvement this patch also updates the error pretty
printing code to add a space before the '->' when printing a deduction
guide, so we get 'S(int) -> S<int>' instead of 'S(int)-> S<int>'.

	PR c++/115231

gcc/cp/ChangeLog:

	* error.cc (dump_function_decl): Add a space before '->' when
	printing deduction guides.
	* module.cc (depset::hash::add_binding_entity): Don't create
	bindings for guides here, only mark dependencies.
	(depset::hash::add_deduction_guides): New.
	(depset::hash::find_dependencies): Add deduction guide
	dependencies for a class template.
	(module_state::write_cluster): Always consider deduction guides
	as exported.
	* pt.cc (deduction_guides_for): Use 'lkp_iterator' instead of
	'ovl_iterator'.

gcc/testsuite/ChangeLog:

	* g++.dg/modules/dguide-1_a.C: New test.
	* g++.dg/modules/dguide-1_b.C: New test.
	* g++.dg/modules/dguide-2_a.C: New test.
	* g++.dg/modules/dguide-2_b.C: New test.
	* g++.dg/modules/dguide-3_a.C: New test.
	* g++.dg/modules/dguide-3_b.C: New test.
	* g++.dg/modules/dguide-3_c.C: New test.
	* g++.dg/modules/dguide-3_d.C: New test.

Signed-off-by: Nathaniel Shead <nathanieloshead@gmail.com>
---
 gcc/cp/error.cc                           |  1 +
 gcc/cp/module.cc                          | 64 +++++++++++++++++++++++
 gcc/cp/pt.cc                              |  2 +-
 gcc/testsuite/g++.dg/modules/dguide-1_a.C | 44 ++++++++++++++++
 gcc/testsuite/g++.dg/modules/dguide-1_b.C | 20 +++++++
 gcc/testsuite/g++.dg/modules/dguide-2_a.C | 24 +++++++++
 gcc/testsuite/g++.dg/modules/dguide-2_b.C | 19 +++++++
 gcc/testsuite/g++.dg/modules/dguide-3_a.C | 10 ++++
 gcc/testsuite/g++.dg/modules/dguide-3_b.C | 10 ++++
 gcc/testsuite/g++.dg/modules/dguide-3_c.C |  6 +++
 gcc/testsuite/g++.dg/modules/dguide-3_d.C | 30 +++++++++++
 11 files changed, 229 insertions(+), 1 deletion(-)
 create mode 100644 gcc/testsuite/g++.dg/modules/dguide-1_a.C
 create mode 100644 gcc/testsuite/g++.dg/modules/dguide-1_b.C
 create mode 100644 gcc/testsuite/g++.dg/modules/dguide-2_a.C
 create mode 100644 gcc/testsuite/g++.dg/modules/dguide-2_b.C
 create mode 100644 gcc/testsuite/g++.dg/modules/dguide-3_a.C
 create mode 100644 gcc/testsuite/g++.dg/modules/dguide-3_b.C
 create mode 100644 gcc/testsuite/g++.dg/modules/dguide-3_c.C
 create mode 100644 gcc/testsuite/g++.dg/modules/dguide-3_d.C

Comments

Jason Merrill Aug. 6, 2024, 9:39 p.m. UTC | #1 
On 8/6/24 8:03 AM, Nathaniel Shead wrote:
> On Fri, Jul 26, 2024 at 01:17:57PM -0400, Jason Merrill wrote:
>> On 7/26/24 12:52 AM, Nathaniel Shead wrote:
>>> On Tue, Jul 23, 2024 at 04:17:22PM -0400, Jason Merrill wrote:
>>>> On 6/15/24 10:29 PM, Nathaniel Shead wrote:
>>>>> Bootstrapped and regtested on x86_64-pc-linux-gnu, OK for trunk?
>>>>>
>>>>> This probably isn't the most efficient approach, since we need to do
>>>>> name lookup to find deduction guides for a type which will also
>>>>> potentially do a bunch of pointless lazy loading from imported modules,
>>>>> but I wasn't able to work out a better approach without completely
>>>>> reworking how deduction guides are stored and represented.
>>>>
>>>> Indeed.  We likely want to find them more directly from the template; it's
>>>> not clear to me that DECL_INITIAL is used for TEMPLATE_DECL, or we could put
>>>> them in an internal attribute or a separate hash table.
>>>>
> 
> I had a go at exploring some other representations of deduction guides
> but I didn't really land anywhere; most things I tried would require
> reimplementing a lot of the merging logic handled by name lookup
> currently.  Potentially just having an additional hash table for
> deduction guides maintained by pushdecl would work but that felt like
> unnecessary duplication, I'm not sure that the benefits outweight the
> cost of the name lookup.  (But I haven't measured this.)
> 
>>>>> -- >8 --
>>>>>
>>>>> Deduction guides are represented as 'normal' functions currently, and
>>>>> have no special handling in modules.  However, this causes some issues;
>>>>> by [temp.deduct.guide] a deduction guide is not found by normal name
>>>>> lookup and instead all reachable deduction guides for a class template
>>>>> should be considered, but this does not happen currently.
>>>>>
>>>>> To solve this, this patch ensures that all deduction guides are
>>>>> considered exported to ensure that they are always visible to importers
>>>>> if they are reachable.  Another alternative here would be to add a new
>>>>> kind of "all reachable" flag to name lookup, but that is complicated by
>>>>> some difficulties in handling GM entities; this may be a better way to
>>>>> go if more kinds of entities end up needing this handling, however.
>>>>>
>>>>> Another issue here is that because deduction guides are "unrelated"
>>>>> functions, they will usually get discarded from the GMF, so this patch
>>>>> ensures that when finding dependencies, GMF deduction guides will also
>>>>> have bindings created.  We do this in find_dependencies so that we don't
>>>>> unnecessarily create bindings for GMF deduction guides that are never
>>>>> reached; for consistency we do this for *all* deduction guides, not just
>>>>> GM ones.
>>>>
>>>> If you fixed the dependency calculation, why do they also need to be
>>>> exported?
>>>
>>> Deduction guides aren't found using normal name lookup, but any
>>> reachable deduction guide must be considered.  This means that even if
>>> the module interface exports no declarations whatsoever, a deduction
>>> guide declared in the module purview must still be considered by
>>> importers.
>>
>> Ah, I was missing the name lookup issue.
>>
>>> The other option I've considered is adding a new "ANY_REACHABLE" flag to
>>> name lookup which would also consider non-exported reachable decls.  On
>>> further consideration I might actually go this way; I've been thinking
>>> about how to resolve some issues adjacent to supporting textual
>>> redefinitions that I believe this will be necessary for anyway, and we
>>> can probably use this in tsubst_friend_class as well rather than the
>>> current relatively ad-hoc solution.
>>
>> There's also my hack in lookup_elaborated_type for ABI namespace types.
>>
>> I'm not sure that it should be necessary to do this for redefinitions,
>> though; what's the advantage over merging in check_module_override (apart
>> from that needing to be fixed)?
>>
>>> That said, I've realised that this patch isn't completely sufficient
>>> anyway; consider:
>>>
>>>     // m.cpp
>>>     module;
>>>     template <typename T> struct S;
>>>     export module M;
>>>     S(int) -> S<double>;
>>>
>>>     // x.cpp
>>>     template <typename T> struct S { S(int); };
>>>     import M;
>>>     int main() {
>>>       S s(10);  // should be S<double> s;
>>>     }
>>>
>>> This patch doesn't correctly handle this case yet, we need to also
>>> consider cases where only the deduction guide is in purview.
>>
>> Indeed.
>>
>> Jason
>>
> 
> Here's a small update to the patch which fixes the above bug by ensuring
> that dependencies are created for purview deduction guides rather than
> just being skipped entirely, and includes that testcase with dguide-3.
> 
> This patch is still using force-exported bindings rather than attempting
> to do anything more clever with name lookup just yet.
> 
> Bootstrapped and regtested (so far just modules.exp) on
> x86_64-pc-linux-gnu, OK for trunk if full regtest succeeds?
> 
> -- >8 --
> 
> Deduction guides are represented as 'normal' functions currently, and
> have no special handling in modules.  However, this causes some issues;
> by [temp.deduct.guide] a deduction guide is not found by normal name
> lookup and instead all reachable deduction guides for a class template
> should be considered, but this does not happen currently.
> 
> To solve this, this patch ensures that all deduction guides are
> considered exported to ensure that they are always visible to importers
> if they are reachable.  Another alternative here would be to add a new
> kind of "all reachable" flag to name lookup, but that is complicated by
> some difficulties in handling GM entities; this may be a better way to
> go if more kinds of entities end up needing this handling, however.
> 
> Another issue here is that because deduction guides are "unrelated"
> functions, they will usually get discarded from the GMF, so this patch
> ensures that when finding dependencies, GMF deduction guides will also
> have bindings created.  We do this in find_dependencies so that we don't
> unnecessarily create bindings for GMF deduction guides that are never
> reached; for consistency we do this for *all* deduction guides, not just
> GM ones.  We also make sure that the opposite (a deduction guide being
> the only purview reference to a GMF template) correctly marks it as
> reachable.
> 
> Finally, when merging deduction guides from multiple modules, the name
> lookup code may now return two-dimensional overload sets, so update
> callers to match.
> 
> As a small drive-by improvement this patch also updates the error pretty
> printing code to add a space before the '->' when printing a deduction
> guide, so we get 'S(int) -> S<int>' instead of 'S(int)-> S<int>'.
> 
> @@ -15158,6 +15218,10 @@ module_state::write_cluster (elf_out *to, depset *scc[], unsigned size,
>   		      flags |= cbf_hidden;
>   		    else if (DECL_MODULE_EXPORT_P (STRIP_TEMPLATE (bound)))
>   		      flags |= cbf_export;
> +		    else if (deduction_guide_p (bound))
> +		      /* Deduction guides are always exported so that they are
> +			 reachable whenever their class template is.  */

I'd adjust "reachable" to "visible to name lookup"; they don't need to 
be exported to be reachable.  OK with that change.

Jason
diff mbox series

Patch

diff --git a/gcc/cp/error.cc b/gcc/cp/error.cc
index d80bac822ba..601e5672174 100644
--- a/gcc/cp/error.cc
+++ b/gcc/cp/error.cc
@@ -1871,6 +1871,7 @@  dump_function_decl (cxx_pretty_printer *pp, tree t, int flags)
 	dump_type_suffix (pp, ret, flags);
       else if (deduction_guide_p (t))
 	{
+	  pp->set_padding (pp_before);
 	  pp_cxx_ws_string (pp, "->");
 	  dump_type (pp, TREE_TYPE (TREE_TYPE (t)), flags);
 	}
diff --git a/gcc/cp/module.cc b/gcc/cp/module.cc
index d1607a06757..aef010e91d1 100644
--- a/gcc/cp/module.cc
+++ b/gcc/cp/module.cc
@@ -2589,6 +2589,9 @@  public:
     void add_partial_entities (vec<tree, va_gc> *);
     void add_class_entities (vec<tree, va_gc> *);
 
+  private:
+    void add_deduction_guides (tree decl);
+
   public:    
     void find_dependencies (module_state *);
     bool finalize_dependencies ();
@@ -13172,6 +13175,15 @@  depset::hash::add_binding_entity (tree decl, WMB_Flags flags, void *data_)
 	/* Ignore NTTP objects.  */
 	return false;
 
+      if (deduction_guide_p (decl))
+	{
+	  /* Ignore deduction guides, bindings for them will be created within
+	     find_dependencies for their class template.  But still build a dep
+	     for them so that we don't discard them.  */
+	  data->hash->make_dependency (decl, EK_FOR_BINDING);
+	  return false;
+	}
+
       if (!(flags & WMB_Using) && CP_DECL_CONTEXT (decl) != data->ns)
 	{
 	  /* An unscoped enum constant implicitly brought into the containing
@@ -13600,6 +13612,50 @@  find_pending_key (tree decl, tree *decl_p = nullptr)
   return ns;
 }
 
+/* Creates bindings and dependencies for all deduction guides of
+   the given class template DECL as needed.  */
+
+void
+depset::hash::add_deduction_guides (tree decl)
+{
+  /* Alias templates never have deduction guides.  */
+  if (DECL_ALIAS_TEMPLATE_P (decl))
+    return;
+
+  /* We don't need to do anything for class-scope deduction guides,
+     as they will be added as members anyway.  */
+  if (!DECL_NAMESPACE_SCOPE_P (decl))
+    return;
+
+  tree ns = CP_DECL_CONTEXT (decl);
+  tree name = dguide_name (decl);
+
+  /* We always add all deduction guides with a given name at once,
+     so if there's already a binding there's nothing to do.  */
+  if (find_binding (ns, name))
+    return;
+
+  tree guides = lookup_qualified_name (ns, name, LOOK_want::NORMAL,
+				       /*complain=*/false);
+  if (guides == error_mark_node)
+    return;
+
+  /* We have bindings to add.  */
+  depset *binding = make_binding (ns, name);
+  add_namespace_context (binding, ns);
+
+  depset **slot = binding_slot (ns, name, /*insert=*/true);
+  *slot = binding;
+
+  for (lkp_iterator it (guides); it; ++it)
+    {
+      gcc_checking_assert (!TREE_VISITED (*it));
+      depset *dep = make_dependency (*it, EK_FOR_BINDING);
+      binding->deps.safe_push (dep);
+      dep->deps.safe_push (binding);
+    }
+}
+
 /* Iteratively find dependencies.  During the walk we may find more
    entries on the same binding that need walking.  */
 
@@ -13659,6 +13715,10 @@  depset::hash::find_dependencies (module_state *module)
 		}
 	      walker.end ();
 
+	      if (!walker.is_key_order ()
+		  && DECL_CLASS_TEMPLATE_P (decl))
+		add_deduction_guides (decl);
+
 	      if (!walker.is_key_order ()
 		  && TREE_CODE (decl) == TEMPLATE_DECL
 		  && !DECL_UNINSTANTIATED_TEMPLATE_FRIEND_P (decl))
@@ -15158,6 +15218,10 @@  module_state::write_cluster (elf_out *to, depset *scc[], unsigned size,
 		      flags |= cbf_hidden;
 		    else if (DECL_MODULE_EXPORT_P (STRIP_TEMPLATE (bound)))
 		      flags |= cbf_export;
+		    else if (deduction_guide_p (bound))
+		      /* Deduction guides are always exported so that they are
+			 reachable whenever their class template is.  */
+		      flags |= cbf_export;
 		  }
 
 		gcc_checking_assert (DECL_P (bound));
diff --git a/gcc/cp/pt.cc b/gcc/cp/pt.cc
index 77fa5907c3d..53917cb6dc6 100644
--- a/gcc/cp/pt.cc
+++ b/gcc/cp/pt.cc
@@ -30739,7 +30739,7 @@  deduction_guides_for (tree tmpl, bool &any_dguides_p, tsubst_flags_t complain)
   else
     {
       cands = ctor_deduction_guides_for (tmpl, complain);
-      for (ovl_iterator it (guides); it; ++it)
+      for (lkp_iterator it (guides); it; ++it)
 	cands = lookup_add (*it, cands);
     }
 
diff --git a/gcc/testsuite/g++.dg/modules/dguide-1_a.C b/gcc/testsuite/g++.dg/modules/dguide-1_a.C
new file mode 100644
index 00000000000..834e033eae3
--- /dev/null
+++ b/gcc/testsuite/g++.dg/modules/dguide-1_a.C
@@ -0,0 +1,44 @@ 
+// PR c++/115231
+// { dg-additional-options "-fmodules-ts -Wno-global-module" }
+// { dg-module-cmi M }
+
+module;
+
+template <typename T>
+struct A {
+  template <typename U> A(U);
+};
+
+template <typename T> A(T) -> A<T>;
+
+export module M;
+
+// Exporting a GMF entity should make the deduction guides reachable.
+export using ::A;
+
+
+export template <typename T>
+struct B {
+  template <typename U> B(U);
+};
+
+// Not exported, but should still be reachable by [temp.deduct.guide] p1.
+B(int) -> B<double>;
+
+
+// Class-scope deduction guides should be reachable as well, even if
+// the class body was not exported.
+export template <typename T> struct C;
+
+template <typename T>
+struct C {
+  template <typename U>
+  struct I {
+    template <typename V> I(V);
+  };
+
+  I(int) -> I<int>;
+
+  template <typename P>
+  I(const P*) -> I<P>;
+};
diff --git a/gcc/testsuite/g++.dg/modules/dguide-1_b.C b/gcc/testsuite/g++.dg/modules/dguide-1_b.C
new file mode 100644
index 00000000000..97266986d8f
--- /dev/null
+++ b/gcc/testsuite/g++.dg/modules/dguide-1_b.C
@@ -0,0 +1,20 @@ 
+// PR c++/115231
+// { dg-additional-options "-fmodules-ts" }
+
+import M;
+
+int main() {
+  // Check that deduction guides are reachable,
+  // and that they declared the right type.
+  A a(1);
+  A<int> a2 = a;
+
+  B b(2);
+  B<double> b2 = b;
+
+  C<int>::I x(10);
+  C<int>::I<int> x2 = x;
+
+  C<int>::I y("xyz");
+  C<int>::I<char> y2 = y;
+}
diff --git a/gcc/testsuite/g++.dg/modules/dguide-2_a.C b/gcc/testsuite/g++.dg/modules/dguide-2_a.C
new file mode 100644
index 00000000000..fcd6c579813
--- /dev/null
+++ b/gcc/testsuite/g++.dg/modules/dguide-2_a.C
@@ -0,0 +1,24 @@ 
+// PR c++/115231
+// { dg-additional-options "-fmodules-ts -Wno-global-module" }
+// { dg-module-cmi M }
+
+module;
+
+template <typename T>
+struct A {
+  template <typename U> A(U);
+};
+template <typename T> A(T) -> A<T>;
+
+export module M;
+
+template <typename T>
+struct B {
+  template <typename U> B(U);
+};
+B(int) -> B<int>;
+
+// Accessing deduction guides should be possible,
+// even if we can't name the type directly.
+export A<void> f();
+export B<void> g();
diff --git a/gcc/testsuite/g++.dg/modules/dguide-2_b.C b/gcc/testsuite/g++.dg/modules/dguide-2_b.C
new file mode 100644
index 00000000000..ca31306aea3
--- /dev/null
+++ b/gcc/testsuite/g++.dg/modules/dguide-2_b.C
@@ -0,0 +1,19 @@ 
+// PR c++/115231
+// { dg-additional-options "-fmodules-ts" }
+
+import M;
+
+template <typename>
+struct U;
+
+template <template <typename> typename TT, typename Inner>
+struct U<TT<Inner>> {
+  void go() {
+    TT t(10);
+  }
+};
+
+int main() {
+  U<decltype(f())>{}.go();
+  U<decltype(g())>{}.go();
+}
diff --git a/gcc/testsuite/g++.dg/modules/dguide-3_a.C b/gcc/testsuite/g++.dg/modules/dguide-3_a.C
new file mode 100644
index 00000000000..33350ce9804
--- /dev/null
+++ b/gcc/testsuite/g++.dg/modules/dguide-3_a.C
@@ -0,0 +1,10 @@ 
+// { dg-additional-options "-fmodules-ts" }
+// { dg-module-cmi A }
+
+export module A;
+
+extern "C++" {
+  template <typename T> struct S;
+  S(int) -> S<int>;
+  S(double) -> S<double>;
+}
diff --git a/gcc/testsuite/g++.dg/modules/dguide-3_b.C b/gcc/testsuite/g++.dg/modules/dguide-3_b.C
new file mode 100644
index 00000000000..d23696c74bd
--- /dev/null
+++ b/gcc/testsuite/g++.dg/modules/dguide-3_b.C
@@ -0,0 +1,10 @@ 
+// { dg-additional-options "-fmodules-ts" }
+// { dg-module-cmi B }
+
+export module B;
+
+extern "C++" {
+  template <typename T> struct S;
+  S(int) -> S<int>;
+  S(char) -> S<char>;
+}
diff --git a/gcc/testsuite/g++.dg/modules/dguide-3_c.C b/gcc/testsuite/g++.dg/modules/dguide-3_c.C
new file mode 100644
index 00000000000..95d29db724d
--- /dev/null
+++ b/gcc/testsuite/g++.dg/modules/dguide-3_c.C
@@ -0,0 +1,6 @@ 
+// { dg-additional-options "-fmodules-ts -Wno-global-module" }
+
+module;
+template <typename T> struct S;
+export module C;
+S(const char*) -> S<const char*>;
diff --git a/gcc/testsuite/g++.dg/modules/dguide-3_d.C b/gcc/testsuite/g++.dg/modules/dguide-3_d.C
new file mode 100644
index 00000000000..0d98517b7da
--- /dev/null
+++ b/gcc/testsuite/g++.dg/modules/dguide-3_d.C
@@ -0,0 +1,30 @@ 
+// { dg-additional-options "-fmodules-ts" }
+// Test merging deduction guides.
+
+template <typename T> struct S {
+  template <typename U> S(U);
+};
+
+import A;
+import B;
+import C;
+
+int main() {
+  // declared in A and B
+  S x(123);
+  S<int> x2 = x;
+
+  // declared only in A
+  S y(0.5);
+  S<double> y2 = y;
+
+  // declared only in B
+  S z('c');
+  S<char> z2 = z;
+
+  // declared only in C (and attached to named module)
+  S w("hello");
+  S<const char*> w2 = w;
+}
+
+S(char) -> S<double>;  // { dg-error "ambiguating" }