* [PATCH] Compute proper length for dynamic types of TYPE_CODE_TYPEDEF
@ 2020-06-10 11:35 Tom de Vries
2020-06-11 10:46 ` Pedro Alves
0 siblings, 1 reply; 5+ messages in thread
From: Tom de Vries @ 2020-06-10 11:35 UTC (permalink / raw)
To: gdb-patches; +Cc: Keith Seitz, Pedro Alves
Hi,
[ This patch was submitted earlier here (
https://sourceware.org/legacy-ml/gdb-patches/2017-06/msg00679.html ).
I've:
- made the patch applicable to current master
- made the log message narrower
- added a missing initialization in the test-case:
vla_struct_typedef_struct_member_object.something = n * 2;
- renamed things in the test-case as suggested by Pedro's review comments
( https://sourceware.org/legacy-ml/gdb-patches/2017-06/msg00696.html ).
- re-tested on x86_64-linux.
]
This patch fixes gdb/21356 in which we hit an assertion in
value_contents_bits_eq:
(gdb) p container_object2
(gdb) p container_object2
$1 = {_container_member2 = 15, _vla_struct_object2 = {_some_member = 0,
_vla_field = {
../../src/gdb/value.c:829: internal-error: \
int value_contents_bits_eq(const value*, int, const value*, int, int): \
Assertion `offset1 + length \
<= TYPE_LENGTH (val1->enclosing_type) * TARGET_CHAR_BIT' failed.
This is happening because TYPE_LENGTH (val1->enclosing_type) is erroneously
based on enclosing_type, which is a typedef, instead of the actual underlying
type.
This can be traced back to resolve_dynamic_struct, where the size of the
type is computed:
...
TYPE_FIELD_TYPE (resolved_type, i)
= resolve_dynamic_type_internal (TYPE_FIELD_TYPE (resolved_type, i),
&pinfo, 0);
gdb_assert (TYPE_FIELD_LOC_KIND (resolved_type, i)
== FIELD_LOC_KIND_BITPOS);
new_bit_length = TYPE_FIELD_BITPOS (resolved_type, i);
if (TYPE_FIELD_BITSIZE (resolved_type, i) != 0)
new_bit_length += TYPE_FIELD_BITSIZE (resolved_type, i);
else
new_bit_length += (TYPE_LENGTH (TYPE_FIELD_TYPE (resolved_type, i))
* TARGET_CHAR_BIT);
...
In this function, resolved_type is TYPE_CODE_TYPEDEF which is not what we
want to use to calculate the size of the actual field.
This patch fixes this and the similar problem in resolve_dynamic_union.
Any comments?
Thanks,
- Tom
Compute proper length for dynamic types of TYPE_CODE_TYPEDEF
gdb/ChangeLog:
yyyy-mm-dd Keith Seitz <keiths@redhat.com>
PR gdb/21356
* gdbtypes.c (resolve_dynamic_union, resolve_dynamic_struct):
Resolve typedefs for type length calculations.
gdb/testsuite/ChangeLog:
yyyy-mm-dd Keith Seitz <keiths@redhat.com>
PR gdb/21356
* gdb.base/vla-datatypes.c (vla_factory): Add typedef for struct
vla_struct.
Add new struct vla_typedef and union vla_typedef_union and
corresponding instantiation objects.
Initialize new objects.
* gdb.base/vla-datatypes.exp: Add tests for vla_typedef_struct_object
and vla_typedef_union_object.
Fixup type for vla_struct_object.
---
gdb/gdbtypes.c | 14 ++++++++++----
gdb/testsuite/gdb.base/vla-datatypes.c | 23 ++++++++++++++++++++++-
gdb/testsuite/gdb.base/vla-datatypes.exp | 6 +++++-
3 files changed, 37 insertions(+), 6 deletions(-)
diff --git a/gdb/gdbtypes.c b/gdb/gdbtypes.c
index b5a13107ed..cdf88a4a7d 100644
--- a/gdb/gdbtypes.c
+++ b/gdb/gdbtypes.c
@@ -2264,8 +2264,10 @@ resolve_dynamic_union (struct type *type,
t = resolve_dynamic_type_internal (resolved_type->field (i).type (),
addr_stack, 0);
resolved_type->field (i).set_type (t);
- if (TYPE_LENGTH (t) > max_len)
- max_len = TYPE_LENGTH (t);
+
+ struct type *real_type = check_typedef (t);
+ if (TYPE_LENGTH (real_type) > max_len)
+ max_len = TYPE_LENGTH (real_type);
}
TYPE_LENGTH (resolved_type) = max_len;
@@ -2521,8 +2523,12 @@ resolve_dynamic_struct (struct type *type,
if (TYPE_FIELD_BITSIZE (resolved_type, i) != 0)
new_bit_length += TYPE_FIELD_BITSIZE (resolved_type, i);
else
- new_bit_length += (TYPE_LENGTH (resolved_type->field (i).type ())
- * TARGET_CHAR_BIT);
+ {
+ struct type *real_type
+ = check_typedef (resolved_type->field (i).type ());
+
+ new_bit_length += (TYPE_LENGTH (real_type) * TARGET_CHAR_BIT);
+ }
/* Normally, we would use the position and size of the last field
to determine the size of the enclosing structure. But GCC seems
diff --git a/gdb/testsuite/gdb.base/vla-datatypes.c b/gdb/testsuite/gdb.base/vla-datatypes.c
index ecdcf4313c..6f95ea0649 100644
--- a/gdb/testsuite/gdb.base/vla-datatypes.c
+++ b/gdb/testsuite/gdb.base/vla-datatypes.c
@@ -50,7 +50,9 @@ vla_factory (int n)
{
int something;
int vla_field[n];
- } vla_struct_object;
+ };
+ typedef struct vla_struct vla_struct_typedef;
+ vla_struct_typedef vla_struct_object;
struct inner_vla_struct
{
@@ -59,14 +61,29 @@ vla_factory (int n)
int after;
} inner_vla_struct_object;
+ struct vla_struct_typedef_struct_member
+ {
+ int something;
+ vla_struct_typedef vla_object;
+ } vla_struct_typedef_struct_member_object;
+
union vla_union
{
int vla_field[n];
} vla_union_object;
+ union vla_struct_typedef_union_member
+ {
+ int something;
+ vla_struct_typedef vla_object;
+ } vla_struct_typedef_union_member_object;
+
vla_struct_object.something = n;
inner_vla_struct_object.something = n;
inner_vla_struct_object.after = n;
+ vla_struct_typedef_struct_member_object.something = n * 2;
+ vla_struct_typedef_struct_member_object.vla_object.something = n * 3;
+ vla_struct_typedef_union_member_object.vla_object.something = n + 1;
for (i = 0; i < n; i++)
{
int_vla[i] = i*2;
@@ -85,6 +102,10 @@ vla_factory (int n)
vla_struct_object.vla_field[i] = i*2;
vla_union_object.vla_field[i] = i*2;
inner_vla_struct_object.vla_field[i] = i*2;
+ vla_struct_typedef_struct_member_object.vla_object.vla_field[i]
+ = i * 3;
+ vla_struct_typedef_union_member_object.vla_object.vla_field[i]
+ = i * 3 - 1;
}
size_t int_size = sizeof(int_vla); /* vlas_filled */
diff --git a/gdb/testsuite/gdb.base/vla-datatypes.exp b/gdb/testsuite/gdb.base/vla-datatypes.exp
index fbc2c97936..e8d84579fe 100644
--- a/gdb/testsuite/gdb.base/vla-datatypes.exp
+++ b/gdb/testsuite/gdb.base/vla-datatypes.exp
@@ -45,6 +45,10 @@ gdb_test "print vla_struct_object" \
"\\\{something = 5, vla_field = \\\{0, 2, 4, 6, 8\\\}\\\}"
gdb_test "print vla_union_object" \
"\\\{vla_field = \\\{0, 2, 4, 6, 8\\\}\\\}"
+gdb_test "print vla_struct_typedef_struct_member_object" \
+ "\\\{something = 10, vla_object = \\\{something = 15, vla_field = \\\{0, 3, 6, 9, 12\\\}\\\}\\\}"
+gdb_test "print vla_struct_typedef_union_member_object" \
+ "\\\{something = 6, vla_object = \\\{something = 6, vla_field = \\\{-1, 2, 5, 8, 11\\\}\\\}\\\}"
# Check whatis of VLA's.
gdb_test "whatis int_vla" "type = int \\\[5\\\]"
@@ -61,7 +65,7 @@ gdb_test "whatis unsigned_short_vla" \
gdb_test "whatis unsigned_char_vla" "type = unsigned char \\\[5\\\]"
gdb_test "whatis foo_vla" "type = struct foo \\\[5\\\]"
gdb_test "whatis bar_vla" "type = BAR \\\[5\\\]"
-gdb_test "whatis vla_struct_object" "type = struct vla_struct"
+gdb_test "whatis vla_struct_object" "type = vla_struct_typedef"
gdb_test "whatis vla_union_object" "type = union vla_union"
# Check ptype of VLA's.
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] Compute proper length for dynamic types of TYPE_CODE_TYPEDEF
2020-06-10 11:35 [PATCH] Compute proper length for dynamic types of TYPE_CODE_TYPEDEF Tom de Vries
@ 2020-06-11 10:46 ` Pedro Alves
2020-06-11 12:38 ` Tom de Vries
0 siblings, 1 reply; 5+ messages in thread
From: Pedro Alves @ 2020-06-11 10:46 UTC (permalink / raw)
To: Tom de Vries, gdb-patches
On 6/10/20 12:35 PM, Tom de Vries wrote:
> Hi,
>
> [ This patch was submitted earlier here (
> https://sourceware.org/legacy-ml/gdb-patches/2017-06/msg00679.html ).
>
> I've:
> - made the patch applicable to current master
> - made the log message narrower
> - added a missing initialization in the test-case:
> vla_struct_typedef_struct_member_object.something = n * 2;
> - renamed things in the test-case as suggested by Pedro's review comments
> ( https://sourceware.org/legacy-ml/gdb-patches/2017-06/msg00696.html ).
> - re-tested on x86_64-linux.
> ]
>
>
> Any comments?
LGTM.
Thanks,
Pedro Alves
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] Compute proper length for dynamic types of TYPE_CODE_TYPEDEF
2020-06-11 10:46 ` Pedro Alves
@ 2020-06-11 12:38 ` Tom de Vries
0 siblings, 0 replies; 5+ messages in thread
From: Tom de Vries @ 2020-06-11 12:38 UTC (permalink / raw)
To: Pedro Alves, gdb-patches
On 11-06-2020 12:46, Pedro Alves wrote:
> On 6/10/20 12:35 PM, Tom de Vries wrote:
>> Hi,
>>
>> [ This patch was submitted earlier here (
>> https://sourceware.org/legacy-ml/gdb-patches/2017-06/msg00679.html ).
>>
>> I've:
>> - made the patch applicable to current master
>> - made the log message narrower
>> - added a missing initialization in the test-case:
>> vla_struct_typedef_struct_member_object.something = n * 2;
>> - renamed things in the test-case as suggested by Pedro's review comments
>> ( https://sourceware.org/legacy-ml/gdb-patches/2017-06/msg00696.html ).
>> - re-tested on x86_64-linux.
>> ]
>>
>>
>> Any comments?
>
> LGTM.
>
Thanks for the review.
I've just realized that Keith submitted a v2 here (
https://sourceware.org/pipermail/gdb-patches/2017-December/145898.html
), so I seem to have duplicated some work.
Anyway, I've copied the comments in the test-case from that version, and
committed.
Thanks,
- Tom
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] Compute proper length for dynamic types of TYPE_CODE_TYPEDEF
2017-06-24 0:59 Keith Seitz
@ 2017-06-26 11:01 ` Pedro Alves
0 siblings, 0 replies; 5+ messages in thread
From: Pedro Alves @ 2017-06-26 11:01 UTC (permalink / raw)
To: Keith Seitz, gdb-patches
Hi Keith,
On 06/24/2017 01:59 AM, Keith Seitz wrote:
> --- a/gdb/testsuite/gdb.base/vla-datatypes.c
> +++ b/gdb/testsuite/gdb.base/vla-datatypes.c
> @@ -46,11 +46,12 @@ vla_factory (int n)
> BAR bar_vla[n];
> int i;
>
> - struct vla_struct
> + typedef struct vla_struct
> {
> int something;
> int vla_field[n];
> - } vla_struct_object;
> + } vla_s;
> + vla_s vla_struct_object;
This object is now defined with a typedef type, and that confused me a lot.
After staring at this for a while, I think you did intend that. I just
find the naming very confusing. Can we improve on that? Maybe add
some comments too.
>
> struct inner_vla_struct
> {
> @@ -59,14 +60,28 @@ vla_factory (int n)
> int after;
> } inner_vla_struct_object;
>
> + struct vla_typedef
... because this type is called "typedef", while
the above isn't.
> + {
> + int something;
> + vla_s vla_object;
> + } vla_typedef_struct_object;
... and this object here is not a a type that is
defined as a typedef, while it's named "..._typedef_...".
So at first is looked like you have the "typedef vs non-typedef"
cases backwards. I understand now that you're referring to
the "vla_object" member, but it was totally non-obvious to me.
Can we rename things a bit to avoid this confusion?
Maybe
struct vla_typedef -> struct vla_s_struct_member
union vla_typedef_union -> union vla_s_union_member
?
and add some comment about using "vla_s" throughout
because it's a typedef. And/or rename it
to "vla_struct_typedef".
Thanks,
Pedro Alves
^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH] Compute proper length for dynamic types of TYPE_CODE_TYPEDEF
@ 2017-06-24 0:59 Keith Seitz
2017-06-26 11:01 ` Pedro Alves
0 siblings, 1 reply; 5+ messages in thread
From: Keith Seitz @ 2017-06-24 0:59 UTC (permalink / raw)
To: gdb-patches
This patch fixes gdb/21356 in which we hit an assertion in
value_contents_bits_eq:
(gdb) p container_object2
(gdb) p container_object2
$1 = {_container_member2 = 15, _vla_struct_object2 = {_some_member = 0,
_vla_field = {
../../src/gdb/value.c:829: internal-error: int value_contents_bits_eq(const value*, int, const value*, int, int): Assertion `offset1 + length <= TYPE_LENGTH (val1->enclosing_type) * TARGET_CHAR_BIT' failed.
A problem internal to GDB has been detected,
further debugging may prove unreliable.
Quit this debugging session? (y or n)
This is happening because TYPE_LENGTH (val1->enclosing_type) is erroneously
based on enclosing_type, which is a typedef, instead of the actual underlying
type.
This can be traced back to resolve_dynamic_struct, where the size of the
type is computed:
2093 TYPE_FIELD_TYPE (resolved_type, i)
2094 = resolve_dynamic_type_internal (TYPE_FIELD_TYPE (resolved_type, i),
2095 &pinfo, 0);
2096 gdb_assert (TYPE_FIELD_LOC_KIND (resolved_type, i)
2097 == FIELD_LOC_KIND_BITPOS);
2098
2099 new_bit_length = TYPE_FIELD_BITPOS (resolved_type, i);
2100 if (TYPE_FIELD_BITSIZE (resolved_type, i) != 0)
2101 new_bit_length += TYPE_FIELD_BITSIZE (resolved_type, i);
2102 else
2103 new_bit_length += (TYPE_LENGTH (TYPE_FIELD_TYPE (resolved_type, i))
2104 * TARGET_CHAR_BIT);
In this function, resolved_type is TYPE_CODE_TYPEDEF which is not what we
want to use to calculate the size of the actual field.
This patch fixes this and the similar problem in resolve_dynamic_union.
gdb/ChangeLog:
yyyy-mm-dd Keith Seitz <keiths@redhat.com>
PR gdb/21356
* gdbtypes.c (resolve_dynamic_union, resolve_dynamic_struct):
Resolve typedefs for type length calculations.
gdb/testsuite/ChangeLog:
yyyy-mm-dd Keith Seitz <keiths@redhat.com>
PR gdb/21356
* gdb.base/vla-datatypes.c (vla_factory): Add typedef for struct
vla_struct.
Add new struct vla_typedef and unin vla_typedef_union and
corresponding instantiation objects.
Initialize new objects.
* gdb.base/vla-datatypes.exp: Add tests for vla_typedef_struct_object
and vla_typedef_union_object.
Fixup type for vla_struct_object.
---
gdb/gdbtypes.c | 14 ++++++++++----
gdb/testsuite/gdb.base/vla-datatypes.c | 21 +++++++++++++++++++--
gdb/testsuite/gdb.base/vla-datatypes.exp | 6 +++++-
3 files changed, 34 insertions(+), 7 deletions(-)
diff --git a/gdb/gdbtypes.c b/gdb/gdbtypes.c
index c9a9b3d..45dc141 100644
--- a/gdb/gdbtypes.c
+++ b/gdb/gdbtypes.c
@@ -2033,8 +2033,10 @@ resolve_dynamic_union (struct type *type,
t = resolve_dynamic_type_internal (TYPE_FIELD_TYPE (resolved_type, i),
addr_stack, 0);
TYPE_FIELD_TYPE (resolved_type, i) = t;
- if (TYPE_LENGTH (t) > max_len)
- max_len = TYPE_LENGTH (t);
+
+ struct type *real_type = check_typedef (t);
+ if (TYPE_LENGTH (real_type) > max_len)
+ max_len = TYPE_LENGTH (real_type);
}
TYPE_LENGTH (resolved_type) = max_len;
@@ -2100,8 +2102,12 @@ resolve_dynamic_struct (struct type *type,
if (TYPE_FIELD_BITSIZE (resolved_type, i) != 0)
new_bit_length += TYPE_FIELD_BITSIZE (resolved_type, i);
else
- new_bit_length += (TYPE_LENGTH (TYPE_FIELD_TYPE (resolved_type, i))
- * TARGET_CHAR_BIT);
+ {
+ struct type *real_type
+ = check_typedef (TYPE_FIELD_TYPE (resolved_type, i));
+
+ new_bit_length += (TYPE_LENGTH (real_type) * TARGET_CHAR_BIT);
+ }
/* Normally, we would use the position and size of the last field
to determine the size of the enclosing structure. But GCC seems
diff --git a/gdb/testsuite/gdb.base/vla-datatypes.c b/gdb/testsuite/gdb.base/vla-datatypes.c
index 4902282..e7aee85 100644
--- a/gdb/testsuite/gdb.base/vla-datatypes.c
+++ b/gdb/testsuite/gdb.base/vla-datatypes.c
@@ -46,11 +46,12 @@ vla_factory (int n)
BAR bar_vla[n];
int i;
- struct vla_struct
+ typedef struct vla_struct
{
int something;
int vla_field[n];
- } vla_struct_object;
+ } vla_s;
+ vla_s vla_struct_object;
struct inner_vla_struct
{
@@ -59,14 +60,28 @@ vla_factory (int n)
int after;
} inner_vla_struct_object;
+ struct vla_typedef
+ {
+ int something;
+ vla_s vla_object;
+ } vla_typedef_struct_object;
+
union vla_union
{
int vla_field[n];
} vla_union_object;
+ union vla_typedef_union
+ {
+ int something;
+ vla_s vla_object;
+ } vla_typedef_union_object;
+
vla_struct_object.something = n;
inner_vla_struct_object.something = n;
inner_vla_struct_object.after = n;
+ vla_typedef_struct_object.something = n * 2;
+ vla_typedef_struct_object.vla_object.something = n * 3;
for (i = 0; i < n; i++)
{
int_vla[i] = i*2;
@@ -85,6 +100,8 @@ vla_factory (int n)
vla_struct_object.vla_field[i] = i*2;
vla_union_object.vla_field[i] = i*2;
inner_vla_struct_object.vla_field[i] = i*2;
+ vla_typedef_struct_object.vla_object.vla_field[i] = i * 3;
+ vla_typedef_union_object.vla_object.vla_field[i] = i * 3 - 1;
}
size_t int_size = sizeof(int_vla); /* vlas_filled */
diff --git a/gdb/testsuite/gdb.base/vla-datatypes.exp b/gdb/testsuite/gdb.base/vla-datatypes.exp
index d32ed5a..93997ed 100644
--- a/gdb/testsuite/gdb.base/vla-datatypes.exp
+++ b/gdb/testsuite/gdb.base/vla-datatypes.exp
@@ -57,6 +57,10 @@ gdb_test "print vla_struct_object" \
"\\\{something = 5, vla_field = \\\{0, 2, 4, 6, 8\\\}\\\}"
gdb_test "print vla_union_object" \
"\\\{vla_field = \\\{0, 2, 4, 6, 8\\\}\\\}"
+gdb_test "print vla_typedef_struct_object" \
+ "\\\{something = 10, vla_object = \\\{something = 15, vla_field = \\\{0, 3, 6, 9, 12\\\}\\\}\\\}"
+gdb_test "print vla_typedef_union_object" \
+ "\\\{something = 6, vla_object = \\\{something = 6, vla_field = \\\{-1, 2, 5, 8, 11\\\}\\\}\\\}"
# Check whatis of VLA's.
gdb_test "whatis int_vla" "type = int \\\[5\\\]" "whatis int_vla"
@@ -78,7 +82,7 @@ gdb_test "whatis unsigned_char_vla" "type = unsigned char \\\[5\\\]" \
"whatis unsigned_char_vla"
gdb_test "whatis foo_vla" "type = struct foo \\\[5\\\]" "whatis foo_vla"
gdb_test "whatis bar_vla" "type = BAR \\\[5\\\]" "whatis bar_vla"
-gdb_test "whatis vla_struct_object" "type = struct vla_struct"
+gdb_test "whatis vla_struct_object" "type = vla_s"
gdb_test "whatis vla_union_object" "type = union vla_union"
# Check ptype of VLA's.
--
2.1.0
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2020-06-11 12:38 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-06-10 11:35 [PATCH] Compute proper length for dynamic types of TYPE_CODE_TYPEDEF Tom de Vries
2020-06-11 10:46 ` Pedro Alves
2020-06-11 12:38 ` Tom de Vries
-- strict thread matches above, loose matches on Subject: below --
2017-06-24 0:59 Keith Seitz
2017-06-26 11:01 ` Pedro Alves
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).