c++/modules: Propagate BINDING_VECTOR_*_DUPS_P on realloc [PR99242]
Checks
| Context |
Check |
Description |
| linaro-tcwg-bot/tcwg_gcc_build--master-arm |
success
|
Build passed
|
| linaro-tcwg-bot/tcwg_gcc_check--master-arm |
success
|
Test passed
|
| linaro-tcwg-bot/tcwg_gcc_build--master-aarch64 |
success
|
Build passed
|
| linaro-tcwg-bot/tcwg_gcc_check--master-aarch64 |
success
|
Test passed
|
Commit Message
Bootstrapped and regtested (so far just modules.exp) on
x86_64-pc-linux-gnu, OK for trunk/14 if full regtest succeeds?
-- >8 --
When importing modules, when a binding vector for a name runs out of
slots it gets reallocated with a larger size, and existing bindings are
copied across. However, the flags to indicate whether deduping needs to
occur did not: this causes ICEs, as it allows a duplicate binding to be
added which then violates assumptions later on.
PR c++/99242
gcc/cp/ChangeLog:
* name-lookup.cc (append_imported_binding_slot): Propagate dups
flags.
gcc/testsuite/ChangeLog:
* g++.dg/modules/pr99242_a.H: New test.
* g++.dg/modules/pr99242_b.H: New test.
* g++.dg/modules/pr99242_c.H: New test.
* g++.dg/modules/pr99242_d.C: New test.
Signed-off-by: Nathaniel Shead <nathanieloshead@gmail.com>
---
gcc/cp/name-lookup.cc | 4 ++++
gcc/testsuite/g++.dg/modules/pr99242_a.H | 3 +++
gcc/testsuite/g++.dg/modules/pr99242_b.H | 3 +++
gcc/testsuite/g++.dg/modules/pr99242_c.H | 3 +++
gcc/testsuite/g++.dg/modules/pr99242_d.C | 7 +++++++
5 files changed, 20 insertions(+)
create mode 100644 gcc/testsuite/g++.dg/modules/pr99242_a.H
create mode 100644 gcc/testsuite/g++.dg/modules/pr99242_b.H
create mode 100644 gcc/testsuite/g++.dg/modules/pr99242_c.H
create mode 100644 gcc/testsuite/g++.dg/modules/pr99242_d.C
Comments
On Mon, 8 Jul 2024, Nathaniel Shead wrote:
> Bootstrapped and regtested (so far just modules.exp) on
> x86_64-pc-linux-gnu, OK for trunk/14 if full regtest succeeds?
>
> -- >8 --
>
> When importing modules, when a binding vector for a name runs out of
> slots it gets reallocated with a larger size, and existing bindings are
> copied across. However, the flags to indicate whether deduping needs to
> occur did not: this causes ICEs, as it allows a duplicate binding to be
> added which then violates assumptions later on.
LGTM
>
> PR c++/99242
>
> gcc/cp/ChangeLog:
>
> * name-lookup.cc (append_imported_binding_slot): Propagate dups
> flags.
>
> gcc/testsuite/ChangeLog:
>
> * g++.dg/modules/pr99242_a.H: New test.
> * g++.dg/modules/pr99242_b.H: New test.
> * g++.dg/modules/pr99242_c.H: New test.
> * g++.dg/modules/pr99242_d.C: New test.
>
> Signed-off-by: Nathaniel Shead <nathanieloshead@gmail.com>
> ---
> gcc/cp/name-lookup.cc | 4 ++++
> gcc/testsuite/g++.dg/modules/pr99242_a.H | 3 +++
> gcc/testsuite/g++.dg/modules/pr99242_b.H | 3 +++
> gcc/testsuite/g++.dg/modules/pr99242_c.H | 3 +++
> gcc/testsuite/g++.dg/modules/pr99242_d.C | 7 +++++++
> 5 files changed, 20 insertions(+)
> create mode 100644 gcc/testsuite/g++.dg/modules/pr99242_a.H
> create mode 100644 gcc/testsuite/g++.dg/modules/pr99242_b.H
> create mode 100644 gcc/testsuite/g++.dg/modules/pr99242_c.H
> create mode 100644 gcc/testsuite/g++.dg/modules/pr99242_d.C
>
> diff --git a/gcc/cp/name-lookup.cc b/gcc/cp/name-lookup.cc
> index b57893116eb..3f0b8a59ff4 100644
> --- a/gcc/cp/name-lookup.cc
> +++ b/gcc/cp/name-lookup.cc
> @@ -352,6 +352,10 @@ append_imported_binding_slot (tree *slot, tree name, unsigned ix)
>
> tree new_vec = make_binding_vec (name, want);
> BINDING_VECTOR_NUM_CLUSTERS (new_vec) = have + 1;
> + BINDING_VECTOR_GLOBAL_DUPS_P (new_vec)
> + = BINDING_VECTOR_GLOBAL_DUPS_P (*slot);
> + BINDING_VECTOR_PARTITION_DUPS_P (new_vec)
> + = BINDING_VECTOR_PARTITION_DUPS_P (*slot);
> memcpy (BINDING_VECTOR_CLUSTER_BASE (new_vec),
> BINDING_VECTOR_CLUSTER_BASE (*slot),
> have * sizeof (binding_cluster));
> diff --git a/gcc/testsuite/g++.dg/modules/pr99242_a.H b/gcc/testsuite/g++.dg/modules/pr99242_a.H
> new file mode 100644
> index 00000000000..2df0b4184ab
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/modules/pr99242_a.H
> @@ -0,0 +1,3 @@
> +// { dg-additional-options "-fmodule-header" }
> +// { dg-module-cmi {} }
> +bool __is_constant_evaluated();
> diff --git a/gcc/testsuite/g++.dg/modules/pr99242_b.H b/gcc/testsuite/g++.dg/modules/pr99242_b.H
> new file mode 100644
> index 00000000000..2df0b4184ab
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/modules/pr99242_b.H
> @@ -0,0 +1,3 @@
> +// { dg-additional-options "-fmodule-header" }
> +// { dg-module-cmi {} }
> +bool __is_constant_evaluated();
> diff --git a/gcc/testsuite/g++.dg/modules/pr99242_c.H b/gcc/testsuite/g++.dg/modules/pr99242_c.H
> new file mode 100644
> index 00000000000..2df0b4184ab
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/modules/pr99242_c.H
> @@ -0,0 +1,3 @@
> +// { dg-additional-options "-fmodule-header" }
> +// { dg-module-cmi {} }
> +bool __is_constant_evaluated();
> diff --git a/gcc/testsuite/g++.dg/modules/pr99242_d.C b/gcc/testsuite/g++.dg/modules/pr99242_d.C
> new file mode 100644
> index 00000000000..1046d1af984
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/modules/pr99242_d.C
> @@ -0,0 +1,7 @@
> +// { dg-additional-options "-fmodules-ts" }
> +bool __is_constant_evaluated();
> +import "pr99242_a.H";
> +void f() { __is_constant_evaluated(); }
> +import "pr99242_b.H";
> +import "pr99242_c.H";
> +void g() { __is_constant_evaluated(); }
> --
> 2.43.2
>
>
On 7/8/24 8:30 AM, Nathaniel Shead wrote:
> Bootstrapped and regtested (so far just modules.exp) on
> x86_64-pc-linux-gnu, OK for trunk/14 if full regtest succeeds?
OK.
> -- >8 --
>
> When importing modules, when a binding vector for a name runs out of
> slots it gets reallocated with a larger size, and existing bindings are
> copied across. However, the flags to indicate whether deduping needs to
> occur did not: this causes ICEs, as it allows a duplicate binding to be
> added which then violates assumptions later on.
>
> PR c++/99242
>
> gcc/cp/ChangeLog:
>
> * name-lookup.cc (append_imported_binding_slot): Propagate dups
> flags.
>
> gcc/testsuite/ChangeLog:
>
> * g++.dg/modules/pr99242_a.H: New test.
> * g++.dg/modules/pr99242_b.H: New test.
> * g++.dg/modules/pr99242_c.H: New test.
> * g++.dg/modules/pr99242_d.C: New test.
>
> Signed-off-by: Nathaniel Shead <nathanieloshead@gmail.com>
> ---
> gcc/cp/name-lookup.cc | 4 ++++
> gcc/testsuite/g++.dg/modules/pr99242_a.H | 3 +++
> gcc/testsuite/g++.dg/modules/pr99242_b.H | 3 +++
> gcc/testsuite/g++.dg/modules/pr99242_c.H | 3 +++
> gcc/testsuite/g++.dg/modules/pr99242_d.C | 7 +++++++
> 5 files changed, 20 insertions(+)
> create mode 100644 gcc/testsuite/g++.dg/modules/pr99242_a.H
> create mode 100644 gcc/testsuite/g++.dg/modules/pr99242_b.H
> create mode 100644 gcc/testsuite/g++.dg/modules/pr99242_c.H
> create mode 100644 gcc/testsuite/g++.dg/modules/pr99242_d.C
>
> diff --git a/gcc/cp/name-lookup.cc b/gcc/cp/name-lookup.cc
> index b57893116eb..3f0b8a59ff4 100644
> --- a/gcc/cp/name-lookup.cc
> +++ b/gcc/cp/name-lookup.cc
> @@ -352,6 +352,10 @@ append_imported_binding_slot (tree *slot, tree name, unsigned ix)
>
> tree new_vec = make_binding_vec (name, want);
> BINDING_VECTOR_NUM_CLUSTERS (new_vec) = have + 1;
> + BINDING_VECTOR_GLOBAL_DUPS_P (new_vec)
> + = BINDING_VECTOR_GLOBAL_DUPS_P (*slot);
> + BINDING_VECTOR_PARTITION_DUPS_P (new_vec)
> + = BINDING_VECTOR_PARTITION_DUPS_P (*slot);
> memcpy (BINDING_VECTOR_CLUSTER_BASE (new_vec),
> BINDING_VECTOR_CLUSTER_BASE (*slot),
> have * sizeof (binding_cluster));
> diff --git a/gcc/testsuite/g++.dg/modules/pr99242_a.H b/gcc/testsuite/g++.dg/modules/pr99242_a.H
> new file mode 100644
> index 00000000000..2df0b4184ab
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/modules/pr99242_a.H
> @@ -0,0 +1,3 @@
> +// { dg-additional-options "-fmodule-header" }
> +// { dg-module-cmi {} }
> +bool __is_constant_evaluated();
> diff --git a/gcc/testsuite/g++.dg/modules/pr99242_b.H b/gcc/testsuite/g++.dg/modules/pr99242_b.H
> new file mode 100644
> index 00000000000..2df0b4184ab
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/modules/pr99242_b.H
> @@ -0,0 +1,3 @@
> +// { dg-additional-options "-fmodule-header" }
> +// { dg-module-cmi {} }
> +bool __is_constant_evaluated();
> diff --git a/gcc/testsuite/g++.dg/modules/pr99242_c.H b/gcc/testsuite/g++.dg/modules/pr99242_c.H
> new file mode 100644
> index 00000000000..2df0b4184ab
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/modules/pr99242_c.H
> @@ -0,0 +1,3 @@
> +// { dg-additional-options "-fmodule-header" }
> +// { dg-module-cmi {} }
> +bool __is_constant_evaluated();
> diff --git a/gcc/testsuite/g++.dg/modules/pr99242_d.C b/gcc/testsuite/g++.dg/modules/pr99242_d.C
> new file mode 100644
> index 00000000000..1046d1af984
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/modules/pr99242_d.C
> @@ -0,0 +1,7 @@
> +// { dg-additional-options "-fmodules-ts" }
> +bool __is_constant_evaluated();
> +import "pr99242_a.H";
> +void f() { __is_constant_evaluated(); }
> +import "pr99242_b.H";
> +import "pr99242_c.H";
> +void g() { __is_constant_evaluated(); }
@@ -352,6 +352,10 @@ append_imported_binding_slot (tree *slot, tree name, unsigned ix)
tree new_vec = make_binding_vec (name, want);
BINDING_VECTOR_NUM_CLUSTERS (new_vec) = have + 1;
+ BINDING_VECTOR_GLOBAL_DUPS_P (new_vec)
+ = BINDING_VECTOR_GLOBAL_DUPS_P (*slot);
+ BINDING_VECTOR_PARTITION_DUPS_P (new_vec)
+ = BINDING_VECTOR_PARTITION_DUPS_P (*slot);
memcpy (BINDING_VECTOR_CLUSTER_BASE (new_vec),
BINDING_VECTOR_CLUSTER_BASE (*slot),
have * sizeof (binding_cluster));
new file mode 100644
@@ -0,0 +1,3 @@
+// { dg-additional-options "-fmodule-header" }
+// { dg-module-cmi {} }
+bool __is_constant_evaluated();
new file mode 100644
@@ -0,0 +1,3 @@
+// { dg-additional-options "-fmodule-header" }
+// { dg-module-cmi {} }
+bool __is_constant_evaluated();
new file mode 100644
@@ -0,0 +1,3 @@
+// { dg-additional-options "-fmodule-header" }
+// { dg-module-cmi {} }
+bool __is_constant_evaluated();
new file mode 100644
@@ -0,0 +1,7 @@
+// { dg-additional-options "-fmodules-ts" }
+bool __is_constant_evaluated();
+import "pr99242_a.H";
+void f() { __is_constant_evaluated(); }
+import "pr99242_b.H";
+import "pr99242_c.H";
+void g() { __is_constant_evaluated(); }