* [PATCH] Fix internal errors due to missing check_typedef calls
@ 2015-02-24 22:21 Doug Evans
2015-02-26 11:59 ` Pedro Alves
0 siblings, 1 reply; 3+ messages in thread
From: Doug Evans @ 2015-02-24 22:21 UTC (permalink / raw)
To: gdb-patches
Hi.
Silly me forgot to add in the requisite check_typedef calls.
I tripped over this bug with a program that had a
typedef'd unnamed baseclass with virtual methods.
The patch includes testcases to exercise this case now.
I added tests in a couple of files because I wanted to exercise
both ptype and print, and these files looked like the best
place to put them.
Regression tested on amd64-linux.
2015-02-24 Doug Evans <dje@google.com>
* gdbtypes.c (internal_type_vptr_fieldno): Add missing call to
CHECK_TYPEDEF.
(set_type_vptr_fieldno): Ditto.
(internal_type_vptr_basetype, set_type_vptr_basetype): Ditto.
* gnu-v3-abi.c (gnuv3_dynamic_class): Ditto.
testsuite/
* class2.cc (Dbase, D): New classes.
(main): New local delta.
* class2.exp: Test printing delta.
* classes.cc (DynamicBase2, DynamicBar): New classes.
(dynbar): New global.
* classes.exp (test_ptype_class_objects): Test ptype DynamicBar.
diff --git a/gdb/gdbtypes.c b/gdb/gdbtypes.c
index a80151c..ebb047f 100644
--- a/gdb/gdbtypes.c
+++ b/gdb/gdbtypes.c
@@ -1636,6 +1636,7 @@ get_signed_type_minmax (struct type *type, LONGEST *min, LONGEST *max)
int
internal_type_vptr_fieldno (struct type *type)
{
+ CHECK_TYPEDEF (type);
gdb_assert (TYPE_CODE (type) == TYPE_CODE_STRUCT
|| TYPE_CODE (type) == TYPE_CODE_UNION);
if (!HAVE_CPLUS_STRUCT (type))
@@ -1648,6 +1649,7 @@ internal_type_vptr_fieldno (struct type *type)
void
set_type_vptr_fieldno (struct type *type, int fieldno)
{
+ CHECK_TYPEDEF (type);
gdb_assert (TYPE_CODE (type) == TYPE_CODE_STRUCT
|| TYPE_CODE (type) == TYPE_CODE_UNION);
if (!HAVE_CPLUS_STRUCT (type))
@@ -1661,6 +1663,7 @@ set_type_vptr_fieldno (struct type *type, int fieldno)
struct type *
internal_type_vptr_basetype (struct type *type)
{
+ CHECK_TYPEDEF (type);
gdb_assert (TYPE_CODE (type) == TYPE_CODE_STRUCT
|| TYPE_CODE (type) == TYPE_CODE_UNION);
gdb_assert (TYPE_SPECIFIC_FIELD (type) == TYPE_SPECIFIC_CPLUS_STUFF);
@@ -1672,6 +1675,7 @@ internal_type_vptr_basetype (struct type *type)
void
set_type_vptr_basetype (struct type *type, struct type *basetype)
{
+ CHECK_TYPEDEF (type);
gdb_assert (TYPE_CODE (type) == TYPE_CODE_STRUCT
|| TYPE_CODE (type) == TYPE_CODE_UNION);
if (!HAVE_CPLUS_STRUCT (type))
diff --git a/gdb/gnu-v3-abi.c b/gdb/gnu-v3-abi.c
index cdcb354..1dfb37b 100644
--- a/gdb/gnu-v3-abi.c
+++ b/gdb/gnu-v3-abi.c
@@ -202,6 +202,7 @@ gnuv3_dynamic_class (struct type *type)
{
int fieldnum, fieldelem;
+ CHECK_TYPEDEF (type);
gdb_assert (TYPE_CODE (type) == TYPE_CODE_STRUCT
|| TYPE_CODE (type) == TYPE_CODE_UNION);
diff --git a/gdb/testsuite/gdb.cp/class2.cc b/gdb/testsuite/gdb.cp/class2.cc
index 617a8fa..409fb29 100644
--- a/gdb/testsuite/gdb.cp/class2.cc
+++ b/gdb/testsuite/gdb.cp/class2.cc
@@ -46,6 +46,23 @@ struct C : public B
A *c2;
};
+/* Use a typedef for the baseclass, with a virtual method, to exercise
+ gnu-v3-abi.c:gnuv3_dynamic_class recursion. It's important that the
+ class itself have no name to make sure the typedef makes it through
+ to the recursive call. */
+typedef class {
+ public:
+ /* This class has no members as gcc 4.9.x doesn't emit the debug info
+ for them. */
+ virtual int get () { return 42; }
+} Dbase;
+
+class D : public Dbase
+{
+ public:
+ int d1;
+};
+
// Stop the compiler from optimizing away data.
void refer (A *)
{
@@ -65,12 +82,14 @@ int main (void)
A alpha, *aap, *abp, *acp;
B beta, *bbp;
C gamma;
+ D delta;
empty e;
A &aref (alpha);
alpha.a1 = 100;
beta.a1 = 200; beta.b1 = 201; beta.b2 = 202;
gamma.c1 = 0; gamma.c2 = (A *) ~0UL;
+ delta.d1 = 400;
aap = α refer (aap);
abp = β refer (abp);
diff --git a/gdb/testsuite/gdb.cp/class2.exp b/gdb/testsuite/gdb.cp/class2.exp
index 7fedc58..81b3561 100644
--- a/gdb/testsuite/gdb.cp/class2.exp
+++ b/gdb/testsuite/gdb.cp/class2.exp
@@ -101,17 +101,31 @@ gdb_test "print * (B *) abp" \
"= {.*a1 = 200.*b1 = 201.*b2 = 202}" \
"print * (B *) abp at marker return 0"
+# Print the "D" object.
+
+gdb_test "print delta" \
+ "= {.*d1 = 400}" \
+ "print delta with \"print object\" off"
+
# Printing the value of an object containing no data fields:
gdb_test "p e" "= \{<No data fields>\}" "print object with no data fields"
-# Printing NULL pointers with "set print object on"
+# Printing NULL pointers with "set print object on".
gdb_test_no_output "set print object on"
gdb_test "p acp" "= \\(C \\*\\) ${hex}"
gdb_test "p acp->c1" "\\(A \\*\\) 0x0"
gdb_test "p acp->c2" "\\(A \\*\\) ${hex}f"
+# Print the "D" object with "set print object on".
+# There's no difference in output, but this exercises vtable lookup
+# with a typedef'd baseclass.
+
+gdb_test "print delta" \
+ "= {.*d1 = 400}" \
+ "print delta with \"print object\" on"
+
# Regression test for PR c++/15401.
# Check that the type printed is a reference.
gdb_test "p aref" " = \\(A \\&\\) .*"
diff --git a/gdb/testsuite/gdb.cp/classes.cc b/gdb/testsuite/gdb.cp/classes.cc
index e7b7c14..648452c 100644
--- a/gdb/testsuite/gdb.cp/classes.cc
+++ b/gdb/testsuite/gdb.cp/classes.cc
@@ -434,6 +434,24 @@ Foo::operator int() { return x; }
ByAnyOtherName foo(10, 11);
Bar bar(20, 21, 22);
+/* Use a typedef for the baseclass to exercise gnu-v3-abi.c:gnuv3_dynamic_class
+ recursion. It's important that the class itself have no name to make sure
+ the typedef makes it through to the recursive call. */
+typedef class {
+ public:
+ int x;
+ virtual int get_x () { return x; }
+} DynamicBase2;
+
+class DynamicBar : public DynamicBase2
+{
+ public:
+ DynamicBar (int i, int j) { x = i; y = j; }
+ int y;
+};
+
+DynamicBar dynbar (23, 24);
+
class ClassWithEnum {
public:
enum PrivEnum { red, green, blue, yellow = 42 };
diff --git a/gdb/testsuite/gdb.cp/classes.exp b/gdb/testsuite/gdb.cp/classes.exp
index 7b658d5..1d714d3 100644
--- a/gdb/testsuite/gdb.cp/classes.exp
+++ b/gdb/testsuite/gdb.cp/classes.exp
@@ -307,6 +307,15 @@ proc test_ptype_class_objects {} {
{ method public "Bar(int, int, int);" }
}
+ # Derived class with typedef'd baseclass with virtual methods.
+
+ cp_test_ptype_class \
+ "class DynamicBar" "" "class" "DynamicBar" \
+ {
+ { base "public DynamicBase2" }
+ { field public "int y;" }
+ { method public "DynamicBar(int, int);" }
+ }
}
# Test simple access to class members.
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] Fix internal errors due to missing check_typedef calls
2015-02-24 22:21 [PATCH] Fix internal errors due to missing check_typedef calls Doug Evans
@ 2015-02-26 11:59 ` Pedro Alves
2015-02-27 1:33 ` Doug Evans
0 siblings, 1 reply; 3+ messages in thread
From: Pedro Alves @ 2015-02-26 11:59 UTC (permalink / raw)
To: Doug Evans, gdb-patches
On 02/24/2015 10:21 PM, Doug Evans wrote:
> Hi.
>
> Silly me forgot to add in the requisite check_typedef calls.
> I tripped over this bug with a program that had a
> typedef'd unnamed baseclass with virtual methods.
>
> The patch includes testcases to exercise this case now.
> I added tests in a couple of files because I wanted to exercise
> both ptype and print, and these files looked like the best
> place to put them.
>
> Regression tested on amd64-linux.
LGTM.
Thanks,
Pedro Alves
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] Fix internal errors due to missing check_typedef calls
2015-02-26 11:59 ` Pedro Alves
@ 2015-02-27 1:33 ` Doug Evans
0 siblings, 0 replies; 3+ messages in thread
From: Doug Evans @ 2015-02-27 1:33 UTC (permalink / raw)
To: Pedro Alves; +Cc: gdb-patches
On Thu, Feb 26, 2015 at 3:59 AM, Pedro Alves <palves@redhat.com> wrote:
> On 02/24/2015 10:21 PM, Doug Evans wrote:
>> Hi.
>>
>> Silly me forgot to add in the requisite check_typedef calls.
>> I tripped over this bug with a program that had a
>> typedef'd unnamed baseclass with virtual methods.
>>
>> The patch includes testcases to exercise this case now.
>> I added tests in a couple of files because I wanted to exercise
>> both ptype and print, and these files looked like the best
>> place to put them.
>>
>> Regression tested on amd64-linux.
>
> LGTM.
Committed.
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2015-02-27 1:33 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2015-02-24 22:21 [PATCH] Fix internal errors due to missing check_typedef calls Doug Evans
2015-02-26 11:59 ` Pedro Alves
2015-02-27 1:33 ` Doug Evans
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).