diff mbox series

sra: Fix corner case of total scalarization with virtual inheritance (PR 102505)

Message ID ri635oubq7c.fsf@suse.cz
State New
Headers show
Series sra: Fix corner case of total scalarization with virtual inheritance (PR 102505) | expand

Commit Message

Martin Jambor Oct. 21, 2021, 10:57 a.m. UTC
Hi,

PR 102505 is a situation where of SRA takes its initial top-level
access size from a get_ref_base_and_extent called on a COMPONENT_REF,
and thus derived frm the FIELD_DECL, which however does not include a
virtual base.  Total scalarization then goes on traversing the type,
which however has virtual base past the non-virtual bits, tricking SRA
to create sub-accesses outside of the supposedly encompassing
accesses, which in turn triggers the verifier within the pass.

The patch below fixes that by failing total scalarization when this
situation is detected.

Bootstrapped and tested on x86_64-linux, OK for trunk and (after testing
there) on the branches?

Thanks,

Martin


gcc/ChangeLog:

2021-10-20  Martin Jambor  <mjambor@suse.cz>

	PR tree-optimization/102505
	* tree-sra.c (totally_scalarize_subtree): Check that the
	encountered field fits within the acces we would like to put it
	in.

gcc/testsuite/ChangeLog:

2021-10-20  Martin Jambor  <mjambor@suse.cz>

	PR tree-optimization/102505
	* g++.dg/torture/pr102505.C: New test.
---
 gcc/testsuite/g++.dg/torture/pr102505.C | 15 +++++++++++++++
 gcc/tree-sra.c                          |  2 ++
 2 files changed, 17 insertions(+)
 create mode 100644 gcc/testsuite/g++.dg/torture/pr102505.C

Comments

Richard Biener Oct. 21, 2021, 11:21 a.m. UTC | #1
On Thu, Oct 21, 2021 at 12:57 PM Martin Jambor <mjambor@suse.cz> wrote:
>
> Hi,
>
> PR 102505 is a situation where of SRA takes its initial top-level
> access size from a get_ref_base_and_extent called on a COMPONENT_REF,
> and thus derived frm the FIELD_DECL, which however does not include a
> virtual base.  Total scalarization then goes on traversing the type,
> which however has virtual base past the non-virtual bits, tricking SRA
> to create sub-accesses outside of the supposedly encompassing
> accesses, which in turn triggers the verifier within the pass.
>
> The patch below fixes that by failing total scalarization when this
> situation is detected.
>
> Bootstrapped and tested on x86_64-linux, OK for trunk and (after testing
> there) on the branches?

OK.

> Thanks,
>
> Martin
>
>
> gcc/ChangeLog:
>
> 2021-10-20  Martin Jambor  <mjambor@suse.cz>
>
>         PR tree-optimization/102505
>         * tree-sra.c (totally_scalarize_subtree): Check that the
>         encountered field fits within the acces we would like to put it
>         in.
>
> gcc/testsuite/ChangeLog:
>
> 2021-10-20  Martin Jambor  <mjambor@suse.cz>
>
>         PR tree-optimization/102505
>         * g++.dg/torture/pr102505.C: New test.
> ---
>  gcc/testsuite/g++.dg/torture/pr102505.C | 15 +++++++++++++++
>  gcc/tree-sra.c                          |  2 ++
>  2 files changed, 17 insertions(+)
>  create mode 100644 gcc/testsuite/g++.dg/torture/pr102505.C
>
> diff --git a/gcc/testsuite/g++.dg/torture/pr102505.C b/gcc/testsuite/g++.dg/torture/pr102505.C
> new file mode 100644
> index 00000000000..a846751a0d6
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/torture/pr102505.C
> @@ -0,0 +1,15 @@
> +struct D  { int i;  int pad alignas(16); };
> +struct B : virtual D
> +{
> +  int j =84;
> +  int k =84;
> +};
> +
> +struct C: B { };
> +
> +int main()
> +{
> +  C c;
> +  if (c.j != 84 || c.k != 84)
> +    __builtin_abort();
> +}
> diff --git a/gcc/tree-sra.c b/gcc/tree-sra.c
> index 9b786e29e4e..f561e1a2133 100644
> --- a/gcc/tree-sra.c
> +++ b/gcc/tree-sra.c
> @@ -3288,6 +3288,8 @@ totally_scalarize_subtree (struct access *root)
>               continue;
>
>             HOST_WIDE_INT pos = root->offset + int_bit_position (fld);
> +           if (pos + fsize > root->size)
> +             return false;
>             enum total_sra_field_state
>               state = total_should_skip_creating_access (root,
>                                                          &last_seen_sibling,
> --
> 2.33.0
>
diff mbox series

Patch

diff --git a/gcc/testsuite/g++.dg/torture/pr102505.C b/gcc/testsuite/g++.dg/torture/pr102505.C
new file mode 100644
index 00000000000..a846751a0d6
--- /dev/null
+++ b/gcc/testsuite/g++.dg/torture/pr102505.C
@@ -0,0 +1,15 @@ 
+struct D  { int i;  int pad alignas(16); };
+struct B : virtual D
+{
+  int j =84;
+  int k =84;
+};
+
+struct C: B { };
+
+int main()
+{
+  C c;
+  if (c.j != 84 || c.k != 84)
+    __builtin_abort();
+}
diff --git a/gcc/tree-sra.c b/gcc/tree-sra.c
index 9b786e29e4e..f561e1a2133 100644
--- a/gcc/tree-sra.c
+++ b/gcc/tree-sra.c
@@ -3288,6 +3288,8 @@  totally_scalarize_subtree (struct access *root)
 	      continue;
 
 	    HOST_WIDE_INT pos = root->offset + int_bit_position (fld);
+	    if (pos + fsize > root->size)
+	      return false;
 	    enum total_sra_field_state
 	      state = total_should_skip_creating_access (root,
 							 &last_seen_sibling,