On Tue, May 31, 2022 at 12:23 PM HAGIO KAZUHITO(萩尾 一仁) <k-hagio-ab@nec.com> wrote:
Hi Lianbo,

On 2022/05/27 18:56, lijiang wrote:
> On Fri, May 27, 2022 at 5:14 PM HAGIO KAZUHITO(萩尾 一仁) <k-hagio-ab@nec.com <mailto:k-hagio-ab@nec.com>> wrote:
>
>     On 2022/05/27 16:03, lijiang wrote:
>      >      >> If no another gdb setting, we will need to rewrite the parser.
>      >      >> So first, I'd like to know whether there is no another setting.
>      >      >
>      >      > Otherwise, maybe we can patch the gdb code...
>      >      >
>      >      > That '{...}' is probably printed in c_type_print_base_struct_union().
>      >      > And found a comment for c_type_print_base_1() in gdb/c-typeprint.c:
>      >      >
>      >      >      SHOW negative means just print the type name or struct tag if there
>      >      >      is one.  If there is no name, print something sensible but concise
>      >      >      like "struct {...}".
>      >      >
>      >      > Just an idea and clue.
>      >
>      >
>      > I like the good idea.
>      >
>      >     Just a rough try and I've not tested enough, but this patch might be
>      >     somewhat good.
>      >
>      >     --- gdb-10.2.orig/gdb/c-typeprint.c     2022-05-27 14:49:53.079853333 +0900
>      >     +++ gdb-10.2/gdb/c-typeprint.c  2022-05-27 14:47:18.729165094 +0900
>      >     @@ -1043,6 +1043,8 @@
>      >          struct type_print_options local_flags = *flags;
>      >          local_flags.local_typedefs = NULL;
>      >
>      >     +  show = 1;
>      >     +
>      >          std::unique_ptr<typedef_hash_table> hash_holder;
>      >          if (!flags->raw)
>      >            {
>      >
>      >
>      > This looks more reasonable to me. Could you please post a patch with this fix?
>
>     Sure.  But more understanding and testing are needed, I will look into it
>     next week.

I found that gdb-7.6 also prints "{...}" for unnamed struct/union.

(gdb) ptype struct page
type = struct page {
     unsigned long flags;
     struct address_space *mapping;
     struct {
         union {...};
         union {...};
     };
...

And crash-7.x (gdb-7.6.patch) had the following patch to print the details
of unnamed struct/union.

--- gdb-7.6/gdb/c-typeprint.c.orig
+++ gdb-7.6/gdb/c-typeprint.c
@@ -1097,7 +1097,8 @@ c_type_print_base (struct type *type, st
                   fprintf_filtered (stream, "static ");
                 c_print_type (TYPE_FIELD_TYPE (type, i),
                               TYPE_FIELD_NAME (type, i),
-                             stream, show - 1, level + 4,
+                             stream, strlen(TYPE_FIELD_NAME (type, i)) ?
+                             show - 1 : show, level + 4,
                               &local_flags);
                 if (!field_is_static (&TYPE_FIELD (type, i))
                     && TYPE_FIELD_PACKED (type, i))


But gdb-10.2.patch does not have it, and prints "{...}" again.


Good findings, Kazu.
 
So I will post the following patch later to do the same thing in gdb-10.2.

--- gdb-10.2/gdb/c-typeprint.c.orig     2022-05-31 13:15:39.099537173 +0900
+++ gdb-10.2/gdb/c-typeprint.c  2022-05-31 13:17:02.076909579 +0900
@@ -1202,6 +1202,9 @@ c_type_print_base_struct_union (struct t
                 = podata->end_bitpos
                   - TYPE_LENGTH (type->field (i).type ()) * TARGET_CHAR_BIT;
             }
+         else if (strlen(TYPE_FIELD_NAME (type, i)) == 0)
+           /* crash: Print details for unnamed struct and union. */
+           newshow = show;

           c_print_type_1 (type->field (i).type (),
                           TYPE_FIELD_NAME (type, i),

This looks good, it can keep the same behavior as gdb-7.6.

Thanks.
Lianbo


Thanks,
Kazu