* [PATCH] c++ modules: ICE with dynamic_cast [PR106304]
@ 2022-10-13 19:04 Patrick Palka
2022-10-14 11:20 ` Nathan Sidwell
0 siblings, 1 reply; 2+ messages in thread
From: Patrick Palka @ 2022-10-13 19:04 UTC (permalink / raw)
To: gcc-patches; +Cc: jason, nathan, Patrick Palka
The FUNCTION_DECL we build for __dynamic_cast has an empty DECL_CONTEXT,
but trees_out::tree_node expects all FUNCTION_DECLs to have non-empty
DECL_CONTEXT thus we crash when streaming out the dynamic_cast in the
below testcase.
This patch naively fixes this by setting DECL_CONTEXT for __dynamic_cast
appropriately. Like for __cxa_atexit which is similarly lazily declared,
I suppose we should push it into the namespace too.
Bootstrapped and regtested on x86_64-pc-linux-gnu, does this look OK for
trunk?
PR c++/106304
gcc/cp/ChangeLog:
* constexpr.cc (cxx_dynamic_cast_fn_p): Check for abi_node
instead of global_namespace.
* rtti.cc (build_dynamic_cast_1): Set DECL_CONTEXT and
DECL_SOURCE_LOCATION on dynamic_cast_node, and push it
into the namespace.
gcc/testsuite/ChangeLog:
* g++.dg/modules/pr106304_a.C: New test.
* g++.dg/modules/pr106304_b.C: New test.
---
gcc/cp/constexpr.cc | 2 +-
gcc/cp/rtti.cc | 4 ++++
gcc/testsuite/g++.dg/modules/pr106304_a.C | 12 ++++++++++++
gcc/testsuite/g++.dg/modules/pr106304_b.C | 8 ++++++++
4 files changed, 25 insertions(+), 1 deletion(-)
create mode 100644 gcc/testsuite/g++.dg/modules/pr106304_a.C
create mode 100644 gcc/testsuite/g++.dg/modules/pr106304_b.C
diff --git a/gcc/cp/constexpr.cc b/gcc/cp/constexpr.cc
index 06dcd71c926..5939d2882f8 100644
--- a/gcc/cp/constexpr.cc
+++ b/gcc/cp/constexpr.cc
@@ -2108,7 +2108,7 @@ cxx_dynamic_cast_fn_p (tree fndecl)
{
return (cxx_dialect >= cxx20
&& id_equal (DECL_NAME (fndecl), "__dynamic_cast")
- && CP_DECL_CONTEXT (fndecl) == global_namespace);
+ && CP_DECL_CONTEXT (fndecl) == abi_node);
}
/* Often, we have an expression in the form of address + offset, e.g.
diff --git a/gcc/cp/rtti.cc b/gcc/cp/rtti.cc
index f5b43ec0fb2..a85c7b56409 100644
--- a/gcc/cp/rtti.cc
+++ b/gcc/cp/rtti.cc
@@ -787,6 +787,10 @@ build_dynamic_cast_1 (location_t loc, tree type, tree expr,
NULL_TREE));
dcast_fn = (build_library_fn_ptr
(fn_name, fn_type, ECF_LEAF | ECF_PURE | ECF_NOTHROW));
+ /* As with __cxa_atexit in get_atexit_node. */
+ DECL_CONTEXT (dcast_fn) = FROB_CONTEXT (current_namespace);
+ DECL_SOURCE_LOCATION (dcast_fn) = BUILTINS_LOCATION;
+ dcast_fn = pushdecl (dcast_fn, /*hiding=*/true);
pop_abi_namespace (flags);
dynamic_cast_node = dcast_fn;
}
diff --git a/gcc/testsuite/g++.dg/modules/pr106304_a.C b/gcc/testsuite/g++.dg/modules/pr106304_a.C
new file mode 100644
index 00000000000..b999eeccf4a
--- /dev/null
+++ b/gcc/testsuite/g++.dg/modules/pr106304_a.C
@@ -0,0 +1,12 @@
+// PR c++/106304
+// { dg-additional-options -fmodules-ts }
+// { dg-module-cmi pr106304 }
+
+export module pr106304;
+
+struct A { virtual ~A() = default; };
+struct B : A { };
+
+inline const B* as_b(const A& a) {
+ return dynamic_cast<const B*>(&a);
+}
diff --git a/gcc/testsuite/g++.dg/modules/pr106304_b.C b/gcc/testsuite/g++.dg/modules/pr106304_b.C
new file mode 100644
index 00000000000..e8333909c8d
--- /dev/null
+++ b/gcc/testsuite/g++.dg/modules/pr106304_b.C
@@ -0,0 +1,8 @@
+// PR c++/106304
+// { dg-additional-options -fmodules-ts }
+
+module pr106304;
+
+void f(A& a) {
+ as_b(a);
+}
--
2.38.0.68.ge85701b4af
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: [PATCH] c++ modules: ICE with dynamic_cast [PR106304]
2022-10-13 19:04 [PATCH] c++ modules: ICE with dynamic_cast [PR106304] Patrick Palka
@ 2022-10-14 11:20 ` Nathan Sidwell
0 siblings, 0 replies; 2+ messages in thread
From: Nathan Sidwell @ 2022-10-14 11:20 UTC (permalink / raw)
To: Patrick Palka, gcc-patches; +Cc: jason
On 10/13/22 15:04, Patrick Palka wrote:
> The FUNCTION_DECL we build for __dynamic_cast has an empty DECL_CONTEXT,
> but trees_out::tree_node expects all FUNCTION_DECLs to have non-empty
> DECL_CONTEXT thus we crash when streaming out the dynamic_cast in the
> below testcase.
>
> This patch naively fixes this by setting DECL_CONTEXT for __dynamic_cast
> appropriately. Like for __cxa_atexit which is similarly lazily declared,
> I suppose we should push it into the namespace too.
>
> Bootstrapped and regtested on x86_64-pc-linux-gnu, does this look OK for
> trunk?
This is correct, there were a bunch of similar issues, surprising these
never triggered.
>
> PR c++/106304
>
> gcc/cp/ChangeLog:
>
> * constexpr.cc (cxx_dynamic_cast_fn_p): Check for abi_node
> instead of global_namespace.
> * rtti.cc (build_dynamic_cast_1): Set DECL_CONTEXT and
> DECL_SOURCE_LOCATION on dynamic_cast_node, and push it
> into the namespace.
>
> gcc/testsuite/ChangeLog:
>
> * g++.dg/modules/pr106304_a.C: New test.
> * g++.dg/modules/pr106304_b.C: New test.
> ---
> gcc/cp/constexpr.cc | 2 +-
> gcc/cp/rtti.cc | 4 ++++
> gcc/testsuite/g++.dg/modules/pr106304_a.C | 12 ++++++++++++
> gcc/testsuite/g++.dg/modules/pr106304_b.C | 8 ++++++++
> 4 files changed, 25 insertions(+), 1 deletion(-)
> create mode 100644 gcc/testsuite/g++.dg/modules/pr106304_a.C
> create mode 100644 gcc/testsuite/g++.dg/modules/pr106304_b.C
>
> diff --git a/gcc/cp/constexpr.cc b/gcc/cp/constexpr.cc
> index 06dcd71c926..5939d2882f8 100644
> --- a/gcc/cp/constexpr.cc
> +++ b/gcc/cp/constexpr.cc
> @@ -2108,7 +2108,7 @@ cxx_dynamic_cast_fn_p (tree fndecl)
> {
> return (cxx_dialect >= cxx20
> && id_equal (DECL_NAME (fndecl), "__dynamic_cast")
> - && CP_DECL_CONTEXT (fndecl) == global_namespace);
> + && CP_DECL_CONTEXT (fndecl) == abi_node);
> }
>
> /* Often, we have an expression in the form of address + offset, e.g.
> diff --git a/gcc/cp/rtti.cc b/gcc/cp/rtti.cc
> index f5b43ec0fb2..a85c7b56409 100644
> --- a/gcc/cp/rtti.cc
> +++ b/gcc/cp/rtti.cc
> @@ -787,6 +787,10 @@ build_dynamic_cast_1 (location_t loc, tree type, tree expr,
> NULL_TREE));
> dcast_fn = (build_library_fn_ptr
> (fn_name, fn_type, ECF_LEAF | ECF_PURE | ECF_NOTHROW));
> + /* As with __cxa_atexit in get_atexit_node. */
> + DECL_CONTEXT (dcast_fn) = FROB_CONTEXT (current_namespace);
> + DECL_SOURCE_LOCATION (dcast_fn) = BUILTINS_LOCATION;
> + dcast_fn = pushdecl (dcast_fn, /*hiding=*/true);
> pop_abi_namespace (flags);
> dynamic_cast_node = dcast_fn;
> }
> diff --git a/gcc/testsuite/g++.dg/modules/pr106304_a.C b/gcc/testsuite/g++.dg/modules/pr106304_a.C
> new file mode 100644
> index 00000000000..b999eeccf4a
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/modules/pr106304_a.C
> @@ -0,0 +1,12 @@
> +// PR c++/106304
> +// { dg-additional-options -fmodules-ts }
> +// { dg-module-cmi pr106304 }
> +
> +export module pr106304;
> +
> +struct A { virtual ~A() = default; };
> +struct B : A { };
> +
> +inline const B* as_b(const A& a) {
> + return dynamic_cast<const B*>(&a);
> +}
> diff --git a/gcc/testsuite/g++.dg/modules/pr106304_b.C b/gcc/testsuite/g++.dg/modules/pr106304_b.C
> new file mode 100644
> index 00000000000..e8333909c8d
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/modules/pr106304_b.C
> @@ -0,0 +1,8 @@
> +// PR c++/106304
> +// { dg-additional-options -fmodules-ts }
> +
> +module pr106304;
> +
> +void f(A& a) {
> + as_b(a);
> +}
--
Nathan Sidwell
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2022-10-14 11:20 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-10-13 19:04 [PATCH] c++ modules: ICE with dynamic_cast [PR106304] Patrick Palka
2022-10-14 11:20 ` Nathan Sidwell
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).