* [PATCH] Fix incorrect .gdb_index with new DWARF scanner
@ 2022-10-17 19:41 Tom Tromey
2022-10-18 13:41 ` Tom de Vries
0 siblings, 1 reply; 3+ messages in thread
From: Tom Tromey @ 2022-10-17 19:41 UTC (permalink / raw)
To: gdb-patches; +Cc: Tom Tromey
PR symtab/29694 points out a regression caused by the new DWARF
scanner when the cc-with-gdb-index target board is used.
What happens here is that an older version of gdb will make an index
describing the "A" type as:
[737] A: 1 [global, type]
whereas the new gdb says:
[1008] A: 0 [global, type]
Here the old one is correct because the A in CU 0 is just a
declaration without a size:
<1><45>: Abbrev Number: 10 (DW_TAG_structure_type)
<46> DW_AT_name : A
<48> DW_AT_declaration : 1
<48> DW_AT_sibling : <0x6d>
This patch fixes the problem by introducing the idea of a "type
declaration". I think gdb still needs to recurse into these types,
searching for methods, but by marking the type itself as a
declaration, gdb can skip this type during lookups and when writing
the index.
Regression tested on x86-64 using the cc-with-gdb-index board.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29694
---
gdb/dwarf2/cooked-index.h | 15 +++++++++++++++
gdb/dwarf2/index-write.c | 5 +++++
gdb/dwarf2/read.c | 22 ++++++++++++++--------
3 files changed, 34 insertions(+), 8 deletions(-)
diff --git a/gdb/dwarf2/cooked-index.h b/gdb/dwarf2/cooked-index.h
index f3c26480a81..2ea32781be5 100644
--- a/gdb/dwarf2/cooked-index.h
+++ b/gdb/dwarf2/cooked-index.h
@@ -48,6 +48,9 @@ enum cooked_index_flag_enum : unsigned char
IS_ENUM_CLASS = 4,
/* True if this entry uses the linkage name. */
IS_LINKAGE = 8,
+ /* True if this entry is just for the declaration of a type, not the
+ definition. */
+ IS_TYPE_DECLARATION = 16,
};
DEF_ENUM_FLAGS_TYPE (enum cooked_index_flag_enum, cooked_index_flag);
@@ -76,6 +79,10 @@ struct cooked_index_entry : public allocate_on_obstack
/* Return true if this entry matches SEARCH_FLAGS. */
bool matches (block_search_flags search_flags) const
{
+ /* Just reject type declarations. */
+ if ((flags & IS_TYPE_DECLARATION) != 0)
+ return false;
+
if ((search_flags & SEARCH_STATIC_BLOCK) != 0
&& (flags & IS_STATIC) != 0)
return true;
@@ -88,6 +95,10 @@ struct cooked_index_entry : public allocate_on_obstack
/* Return true if this entry matches DOMAIN. */
bool matches (domain_enum domain) const
{
+ /* Just reject type declarations. */
+ if ((flags & IS_TYPE_DECLARATION) != 0)
+ return false;
+
switch (domain)
{
case LABEL_DOMAIN:
@@ -106,6 +117,10 @@ struct cooked_index_entry : public allocate_on_obstack
/* Return true if this entry matches KIND. */
bool matches (enum search_domain kind) const
{
+ /* Just reject type declarations. */
+ if ((flags & IS_TYPE_DECLARATION) != 0)
+ return false;
+
switch (kind)
{
case VARIABLES_DOMAIN:
diff --git a/gdb/dwarf2/index-write.c b/gdb/dwarf2/index-write.c
index f592734addc..3d215a6307b 100644
--- a/gdb/dwarf2/index-write.c
+++ b/gdb/dwarf2/index-write.c
@@ -1167,6 +1167,11 @@ write_cooked_index (cooked_index_vector *table,
be redundant are rare and not worth supporting. */
continue;
}
+ else if ((entry->flags & IS_TYPE_DECLARATION) != 0)
+ {
+ /* Don't add type declarations to the index. */
+ continue;
+ }
gdb_index_symbol_kind kind;
if (entry->tag == DW_TAG_subprogram)
diff --git a/gdb/dwarf2/read.c b/gdb/dwarf2/read.c
index b5efcb3cc09..e849f62576d 100644
--- a/gdb/dwarf2/read.c
+++ b/gdb/dwarf2/read.c
@@ -18192,14 +18192,20 @@ cooked_indexer::scan_attributes (dwarf2_per_cu_data *scanning_per_cu,
that is ok. Similarly, we allow an external variable without a
location; those are resolved via minimal symbols. */
if (is_declaration && !for_specification
- && !(abbrev->tag == DW_TAG_variable && (*flags & IS_STATIC) == 0)
- && !((abbrev->tag == DW_TAG_class_type
- || abbrev->tag == DW_TAG_structure_type
- || abbrev->tag == DW_TAG_union_type)
- && abbrev->has_children))
- {
- *linkage_name = nullptr;
- *name = nullptr;
+ && !(abbrev->tag == DW_TAG_variable && (*flags & IS_STATIC) == 0))
+ {
+ /* We always want to recurse into some types, but we may not
+ want to treat them as definitions. */
+ if ((abbrev->tag == DW_TAG_class_type
+ || abbrev->tag == DW_TAG_structure_type
+ || abbrev->tag == DW_TAG_union_type)
+ && abbrev->has_children)
+ *flags |= IS_TYPE_DECLARATION;
+ else
+ {
+ *linkage_name = nullptr;
+ *name = nullptr;
+ }
}
else if ((*name == nullptr
|| (*linkage_name == nullptr
--
2.34.3
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] Fix incorrect .gdb_index with new DWARF scanner
2022-10-17 19:41 [PATCH] Fix incorrect .gdb_index with new DWARF scanner Tom Tromey
@ 2022-10-18 13:41 ` Tom de Vries
2022-10-21 16:13 ` Tom Tromey
0 siblings, 1 reply; 3+ messages in thread
From: Tom de Vries @ 2022-10-18 13:41 UTC (permalink / raw)
To: Tom Tromey, gdb-patches
On 10/17/22 21:41, Tom Tromey via Gdb-patches wrote:
> PR symtab/29694 points out a regression caused by the new DWARF
> scanner when the cc-with-gdb-index target board is used.
>
> What happens here is that an older version of gdb will make an index
> describing the "A" type as:
>
> [737] A: 1 [global, type]
>
> whereas the new gdb says:
>
> [1008] A: 0 [global, type]
>
> Here the old one is correct because the A in CU 0 is just a
> declaration without a size:
>
> <1><45>: Abbrev Number: 10 (DW_TAG_structure_type)
> <46> DW_AT_name : A
> <48> DW_AT_declaration : 1
> <48> DW_AT_sibling : <0x6d>
>
> This patch fixes the problem by introducing the idea of a "type
> declaration". I think gdb still needs to recurse into these types,
> searching for methods, but by marking the type itself as a
> declaration, gdb can skip this type during lookups and when writing
> the index.
>
> Regression tested on x86-64 using the cc-with-gdb-index board.
>
Regression tested on openSUSE Tumbleweed x86_64 with native and
cc-with-gdb-index board, results looks good.
Patch LGTM.
Thanks,
- Tom
> Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29694
> ---
> gdb/dwarf2/cooked-index.h | 15 +++++++++++++++
> gdb/dwarf2/index-write.c | 5 +++++
> gdb/dwarf2/read.c | 22 ++++++++++++++--------
> 3 files changed, 34 insertions(+), 8 deletions(-)
>
> diff --git a/gdb/dwarf2/cooked-index.h b/gdb/dwarf2/cooked-index.h
> index f3c26480a81..2ea32781be5 100644
> --- a/gdb/dwarf2/cooked-index.h
> +++ b/gdb/dwarf2/cooked-index.h
> @@ -48,6 +48,9 @@ enum cooked_index_flag_enum : unsigned char
> IS_ENUM_CLASS = 4,
> /* True if this entry uses the linkage name. */
> IS_LINKAGE = 8,
> + /* True if this entry is just for the declaration of a type, not the
> + definition. */
> + IS_TYPE_DECLARATION = 16,
> };
> DEF_ENUM_FLAGS_TYPE (enum cooked_index_flag_enum, cooked_index_flag);
>
> @@ -76,6 +79,10 @@ struct cooked_index_entry : public allocate_on_obstack
> /* Return true if this entry matches SEARCH_FLAGS. */
> bool matches (block_search_flags search_flags) const
> {
> + /* Just reject type declarations. */
> + if ((flags & IS_TYPE_DECLARATION) != 0)
> + return false;
> +
> if ((search_flags & SEARCH_STATIC_BLOCK) != 0
> && (flags & IS_STATIC) != 0)
> return true;
> @@ -88,6 +95,10 @@ struct cooked_index_entry : public allocate_on_obstack
> /* Return true if this entry matches DOMAIN. */
> bool matches (domain_enum domain) const
> {
> + /* Just reject type declarations. */
> + if ((flags & IS_TYPE_DECLARATION) != 0)
> + return false;
> +
> switch (domain)
> {
> case LABEL_DOMAIN:
> @@ -106,6 +117,10 @@ struct cooked_index_entry : public allocate_on_obstack
> /* Return true if this entry matches KIND. */
> bool matches (enum search_domain kind) const
> {
> + /* Just reject type declarations. */
> + if ((flags & IS_TYPE_DECLARATION) != 0)
> + return false;
> +
> switch (kind)
> {
> case VARIABLES_DOMAIN:
> diff --git a/gdb/dwarf2/index-write.c b/gdb/dwarf2/index-write.c
> index f592734addc..3d215a6307b 100644
> --- a/gdb/dwarf2/index-write.c
> +++ b/gdb/dwarf2/index-write.c
> @@ -1167,6 +1167,11 @@ write_cooked_index (cooked_index_vector *table,
> be redundant are rare and not worth supporting. */
> continue;
> }
> + else if ((entry->flags & IS_TYPE_DECLARATION) != 0)
> + {
> + /* Don't add type declarations to the index. */
> + continue;
> + }
>
> gdb_index_symbol_kind kind;
> if (entry->tag == DW_TAG_subprogram)
> diff --git a/gdb/dwarf2/read.c b/gdb/dwarf2/read.c
> index b5efcb3cc09..e849f62576d 100644
> --- a/gdb/dwarf2/read.c
> +++ b/gdb/dwarf2/read.c
> @@ -18192,14 +18192,20 @@ cooked_indexer::scan_attributes (dwarf2_per_cu_data *scanning_per_cu,
> that is ok. Similarly, we allow an external variable without a
> location; those are resolved via minimal symbols. */
> if (is_declaration && !for_specification
> - && !(abbrev->tag == DW_TAG_variable && (*flags & IS_STATIC) == 0)
> - && !((abbrev->tag == DW_TAG_class_type
> - || abbrev->tag == DW_TAG_structure_type
> - || abbrev->tag == DW_TAG_union_type)
> - && abbrev->has_children))
> - {
> - *linkage_name = nullptr;
> - *name = nullptr;
> + && !(abbrev->tag == DW_TAG_variable && (*flags & IS_STATIC) == 0))
> + {
> + /* We always want to recurse into some types, but we may not
> + want to treat them as definitions. */
> + if ((abbrev->tag == DW_TAG_class_type
> + || abbrev->tag == DW_TAG_structure_type
> + || abbrev->tag == DW_TAG_union_type)
> + && abbrev->has_children)
> + *flags |= IS_TYPE_DECLARATION;
> + else
> + {
> + *linkage_name = nullptr;
> + *name = nullptr;
> + }
> }
> else if ((*name == nullptr
> || (*linkage_name == nullptr
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] Fix incorrect .gdb_index with new DWARF scanner
2022-10-18 13:41 ` Tom de Vries
@ 2022-10-21 16:13 ` Tom Tromey
0 siblings, 0 replies; 3+ messages in thread
From: Tom Tromey @ 2022-10-21 16:13 UTC (permalink / raw)
To: Tom de Vries via Gdb-patches; +Cc: Tom Tromey, Tom de Vries
>>>>> "Tom" == Tom de Vries via Gdb-patches <gdb-patches@sourceware.org> writes:
>> PR symtab/29694 points out a regression caused by the new DWARF
>> scanner when the cc-with-gdb-index target board is used.
[...]
Tom> Regression tested on openSUSE Tumbleweed x86_64 with native and
Tom> cc-with-gdb-index board, results looks good.
Tom> Patch LGTM.
Thank you. I'm going to check it in.
There's still one open regression for the new DWARF reader. I haven't
fully investigated it yet.
(There's also the TSan bug(s), but on my machine TSan reports a lot of
obviously untrue things, like captured_main being called from a signal
handler. So I haven't really looked at this at all.)
Tom
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2022-10-21 16:13 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-10-17 19:41 [PATCH] Fix incorrect .gdb_index with new DWARF scanner Tom Tromey
2022-10-18 13:41 ` Tom de Vries
2022-10-21 16:13 ` Tom Tromey
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).