[PATCH] Fix crash with C++ qualified names

[PATCH] Fix crash with C++ qualified names

[Tom Tromey]

PR c++/29503 points out that something like "b->Base::member" will
crash when 'b' does not have pointer type.  This seems to be a simple
oversight in eval_op_member.

Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29503
---
 gdb/eval.c                         | 2 ++
 gdb/testsuite/gdb.cp/ambiguous.exp | 1 +
 2 files changed, 3 insertions(+)

diff --git a/gdb/eval.c b/gdb/eval.c
index d0a4a16ceb5..3757028b4a3 100644
--- a/gdb/eval.c
+++ b/gdb/eval.c
@@ -1300,6 +1300,8 @@ eval_op_member (struct type *expect_type, struct expression *exp,
 
     case TYPE_CODE_MEMBERPTR:
       /* Now, convert these values to an address.  */
+      if (check_typedef (value_type (arg1))->code () != TYPE_CODE_PTR)
+	arg1 = value_addr (arg1);
       arg1 = value_cast_pointers (lookup_pointer_type (TYPE_SELF_TYPE (type)),
 				  arg1, 1);
 
diff --git a/gdb/testsuite/gdb.cp/ambiguous.exp b/gdb/testsuite/gdb.cp/ambiguous.exp
index 5d0d2f88b6e..65b33e89a79 100644
--- a/gdb/testsuite/gdb.cp/ambiguous.exp
+++ b/gdb/testsuite/gdb.cp/ambiguous.exp
@@ -123,6 +123,7 @@ with_test_prefix "all fields" {
     gdb_test "print n.A1::y" " = 2"
     gdb_test "print n.A2::x" " = 3"
     gdb_test "print n.A2::y" " = 4"
+    gdb_test "print n->A2::y" " = 4"
     gdb_test "print n.w" " = 5"
     gdb_test "print n.r" " = 6"
     gdb_test "print n.z" " = 7"
-- 
2.38.1

Re: [PATCH] Fix crash with C++ qualified names

[Bruno Larsen]

On 23/12/2022 20:58, Tom Tromey wrote:
> PR c++/29503 points out that something like "b->Base::member" will
> crash when 'b' does not have pointer type.  This seems to be a simple
> oversight in eval_op_member.
>
> Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29503

I can confirm that this fixes the issue, and seems like a simple enough 
change.

Reviewed-By: Bruno Larsen <blarsen@redhat.com>

-- 
Cheers,
Bruno

Re: [PATCH] Fix crash with C++ qualified names

[Andrew Burgess]

Tom Tromey <tom@tromey.com> writes:

> PR c++/29503 points out that something like "b->Base::member" will
> crash when 'b' does not have pointer type.  This seems to be a simple
> oversight in eval_op_member.
>
> Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29503
> ---
>  gdb/eval.c                         | 2 ++
>  gdb/testsuite/gdb.cp/ambiguous.exp | 1 +
>  2 files changed, 3 insertions(+)
>
> diff --git a/gdb/eval.c b/gdb/eval.c
> index d0a4a16ceb5..3757028b4a3 100644
> --- a/gdb/eval.c
> +++ b/gdb/eval.c
> @@ -1300,6 +1300,8 @@ eval_op_member (struct type *expect_type, struct expression *exp,
>  
>      case TYPE_CODE_MEMBERPTR:
>        /* Now, convert these values to an address.  */
> +      if (check_typedef (value_type (arg1))->code () != TYPE_CODE_PTR)
> +	arg1 = value_addr (arg1);
>        arg1 = value_cast_pointers (lookup_pointer_type (TYPE_SELF_TYPE (type)),
>  				  arg1, 1);
>  
> diff --git a/gdb/testsuite/gdb.cp/ambiguous.exp b/gdb/testsuite/gdb.cp/ambiguous.exp
> index 5d0d2f88b6e..65b33e89a79 100644
> --- a/gdb/testsuite/gdb.cp/ambiguous.exp
> +++ b/gdb/testsuite/gdb.cp/ambiguous.exp
> @@ -123,6 +123,7 @@ with_test_prefix "all fields" {
>      gdb_test "print n.A1::y" " = 2"
>      gdb_test "print n.A2::x" " = 3"
>      gdb_test "print n.A2::y" " = 4"
> +    gdb_test "print n->A2::y" " = 4"
>      gdb_test "print n.w" " = 5"
>      gdb_test "print n.r" " = 6"
>      gdb_test "print n.z" " = 7"

LGTM.

Thanks,
Andrew

> -- 
> 2.38.1