This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH v6 2/2] Implement pahole-like 'ptype /o' option
- From: Pedro Alves <palves at redhat dot com>
- To: Sergio Durigan Junior <sergiodj at redhat dot com>, GDB Patches <gdb-patches at sourceware dot org>
- Cc: Tom Tromey <tom at tromey dot com>, Eli Zaretskii <eliz at gnu dot org>, Simon Marchi <simon dot marchi at ericsson dot com>, Keith Seitz <keiths at redhat dot com>
- Date: Thu, 14 Dec 2017 14:19:26 +0000
- Subject: Re: [PATCH v6 2/2] Implement pahole-like 'ptype /o' option
- Authentication-results: sourceware.org; auth=none
- References: <20171121160709.23248-1-sergiodj@redhat.com> <20171214024816.29629-1-sergiodj@redhat.com> <20171214024816.29629-3-sergiodj@redhat.com>
Something broke in v5 -> v6 in the offset and hole computation.
I'm seeing this:
v6:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
(top-gdb) ptype/o minimal_symbol
/* offset | size */ type = struct minimal_symbol {
/* 0 | 32 */ struct general_symbol_info {
/* 0 | 8 */ const char *name;
/* 8 | 8 */ union {
/* 8 */ LONGEST ivalue;
/* 8 */ const block *block;
/* 8 */ const gdb_byte *bytes;
/* 8 */ CORE_ADDR address;
/* 8 */ const common_block *common_block;
/* 8 */ symbol *chain;
/* total size (bytes): 8 */
} value;
/* 24 | 8 */ union {
/* 8 */ obstack *obstack;
/* 8 */ const char *demangled_name;
/* total size (bytes): 8 */
} language_specific;
/* 48:27 | 4 */ language language : 5;
/* 48:26 | 4 */ unsigned int ada_mangled : 1;
/* XXX 2-bit hole */
/* XXX 1-byte hole */
/* 50 | 2 */ short section;
/* total size (bytes): 32 */
} mginfo;
/* 56 | 8 */ unsigned long size;
/* 64 | 8 */ const char *filename;
/* 72:28 | 4 */ minimal_symbol_type type : 4;
/* 72:27 | 4 */ unsigned int created_by_gdb : 1;
/* 72:26 | 4 */ unsigned int target_flag_1 : 1;
/* 72:25 | 4 */ unsigned int target_flag_2 : 1;
/* 72:24 | 4 */ unsigned int has_size : 1;
/* XXX 7-byte hole */
/* 80 | 8 */ minimal_symbol *hash_next;
/* 88 | 8 */ minimal_symbol *demangled_hash_next;
/* total size (bytes): 72 */
}
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
This says that "language_specific" is at 24, with size 8 => next
byte is at 32. "language" is at 48. How come no "XXX hole"
is printed? Similarly in the "8 + 8 -> 24" jump.
It turns out the byte offsets are wrong. Here's the actual
byte offset for language_specific is:
(top-gdb) p /d &((minimal_symbol *) 0) ->mginfo.language_specific
$1 = 16
while v6 says 24.
v5 seemed to get it right. Here's what it would show:
(top-gdb) ptype /o minimal_symbol
/* offset | size */
type = struct minimal_symbol {
/* 0 | 32 */ struct general_symbol_info {
/* 0 | 8 */ const char *name;
/* 8 | 8 */ union {
/* 8 */ LONGEST ivalue;
/* 8 */ const block *block;
/* 8 */ const gdb_byte *bytes;
/* 8 */ CORE_ADDR address;
/* 8 */ const common_block *common_block;
/* 8 */ symbol *chain;
} /* total size: 8 bytes */ value;
/* 16 | 8 */ union {
/* 8 */ obstack *obstack;
/* 8 */ const char *demangled_name;
} /* total size: 8 bytes */ language_specific;
/* 24:27 | 4 */ language language : 5;
/* 24:26 | 4 */ unsigned int ada_mangled : 1;
/* XXX 2-bit hole */
/* XXX 1-byte hole */
/* 26 | 2 */ short section;
} /* total size: 32 bytes */ mginfo;
/* 32 | 8 */ unsigned long size;
/* 40 | 8 */ const char *filename;
/* 48:28 | 4 */ minimal_symbol_type type : 4;
/* 48:27 | 4 */ unsigned int created_by_gdb : 1;
/* 48:26 | 4 */ unsigned int target_flag_1 : 1;
/* 48:25 | 4 */ unsigned int target_flag_2 : 1;
/* 48:24 | 4 */ unsigned int has_size : 1;
/* XXX 7-byte hole */
/* 56 | 8 */ minimal_symbol *hash_next;
/* 64 | 8 */ minimal_symbol *demangled_hash_next;
} /* total size: 72 bytes */
(top-gdb)
Surprised this wasn't caught by the testcases.
I do like the new output format a lot better. Thanks much for
implementing the suggestions!
Pedro Alves