From: Aleksandar Ristovski <aristovski@qnx.com>
To: gdb-patches@sources.redhat.com
Cc: GDB Patches <gdb-patches@sourceware.org>, brobecker@adacore.com
Subject: Re: [rfc] get rid of redundant data in c++ and java
Date: Fri, 02 May 2008 13:43:00 -0000 [thread overview]
Message-ID: <481B198F.30701@qnx.com> (raw)
In-Reply-To: <20080501192532.GG22218@caradoc.them.org>
[-- Attachment #1: Type: text/plain, Size: 2766 bytes --]
Daniel Jacobowitz wrote:
> On Sat, Feb 09, 2008 at 12:21:02AM -0500, Aleksandar Ristovski wrote:
>> Hello,
>>
>> Currently we synthesize typedef info for java and c++ (and ada, but I am
>> not familiar with that language) for any symbol with STRUCT_DOMAIN domain.
>> This can add up to a lot of redundant data. For partial symbols, we
>> generate two identical names for the same symbol, once as struct second as
>> typedef. The same happens with symbols, except we do not duplicate symbol
>> name, but we do allocate another symbol struture.
>>
>> Instead of generating redundant data for symbols with STRUCT_DOMAIN domain
>> in java and c++ languages, treat STRUCT_DOMAIN as equal to VAR_DOMAIN for
>> these two languages when looking up symbols.
>
> Sorry for taking so long to look at this. It's a great idea and the
> patch seems correct to me. Just some cosmetic issues, and then it can
> go in.
>
>> - if (cu->language == language_cplus
>> - || cu->language == language_java
>> - || cu->language == language_ada)
>> + if (cu->language == language_ada)
>> {
>> - /* For C++ and Java, these implicitly act as typedefs as well. */
>> + /* FIXME: Check if Ada really
>> + needs to implicitly set typedef. */
>
> If this works for C++ and Java, it will work for Ada too. If you
> can't test Ada, please post a final version of the patch and ask Joel
> to run tests for you - I'm sure he won't mind getting all that memory
> back for Ada too.
Ok, here is the revised patch. Cc-ing Joel to take a look if it affects ada.
>> -
>> - /* The semantics of C++ state that "struct foo { ... }" also
>> - defines a typedef for "foo". A Java class declaration also
>> - defines a typedef for the class. Synthesize a typedef symbol
>> - so that "ptype foo" works as expected. */
>> +
>
> Since we still need to set the typedef name here, leave the first two
> sentences of the comment.
Sentences put back.
>
>> + if (symbol_language == language_cplus
>> + || symbol_language == language_java)
>
> Only need one space there.
>
One space.
Waiting for an OK to commit.
Thanks,
Aleksandar
ChangeLog:
* dwarf2read.c (add_partial_symbol): Do not add new psym for
STRUCT_DOMAIN. Make sure you recognize c++ struct and java and ada
class as typedefs. See lookup_partial_symbol function.
(new_symbol): Similar to add_partial_symbol, do not create
symbol for the typedef. See lookup_block_symbol.
* symtab.c (symbol_matches_domain): New function, takes care
of dual meaning of STRUCT_DOMAIN symbol for c++, ada and java.
(lookup_partial_symbol): Use symbol_matches_domain to see if the
found psym domain matches the given domain.
(lookup_block_symbol): Likewise.
[-- Attachment #2: duplicatesymbolfix20080502.diff --]
[-- Type: text/plain, Size: 4692 bytes --]
Index: gdb/dwarf2read.c
===================================================================
RCS file: /cvs/src/src/gdb/dwarf2read.c,v
retrieving revision 1.258
diff -u -p -r1.258 dwarf2read.c
--- gdb/dwarf2read.c 25 Apr 2008 18:45:23 -0000 1.258
+++ gdb/dwarf2read.c 2 May 2008 13:36:52 -0000
@@ -2060,16 +2060,6 @@ add_partial_symbol (struct partial_die_i
: &objfile->static_psymbols,
0, (CORE_ADDR) 0, cu->language, objfile);
- if (cu->language == language_cplus
- || cu->language == language_java
- || cu->language == language_ada)
- {
- /* For C++ and Java, these implicitly act as typedefs as well. */
- add_psymbol_to_list (actual_name, strlen (actual_name),
- VAR_DOMAIN, LOC_TYPEDEF,
- &objfile->global_psymbols,
- 0, (CORE_ADDR) 0, cu->language, objfile);
- }
break;
case DW_TAG_enumerator:
add_psymbol_to_list (actual_name, strlen (actual_name),
@@ -7598,23 +7588,16 @@ new_symbol (struct die_info *die, struct
/* The semantics of C++ state that "struct foo { ... }" also
defines a typedef for "foo". A Java class declaration also
- defines a typedef for the class. Synthesize a typedef symbol
- so that "ptype foo" works as expected. */
+ defines a typedef for the class. */
if (cu->language == language_cplus
|| cu->language == language_java
|| cu->language == language_ada)
{
- struct symbol *typedef_sym = (struct symbol *)
- obstack_alloc (&objfile->objfile_obstack,
- sizeof (struct symbol));
- *typedef_sym = *sym;
- SYMBOL_DOMAIN (typedef_sym) = VAR_DOMAIN;
/* The symbol's name is already allocated along with
this objfile, so we don't need to duplicate it for
the type. */
if (TYPE_NAME (SYMBOL_TYPE (sym)) == 0)
TYPE_NAME (SYMBOL_TYPE (sym)) = SYMBOL_SEARCH_NAME (sym);
- add_symbol_to_list (typedef_sym, list_to_add);
}
}
break;
Index: gdb/symtab.c
===================================================================
RCS file: /cvs/src/src/gdb/symtab.c,v
retrieving revision 1.177
diff -u -p -r1.177 symtab.c
--- gdb/symtab.c 19 Apr 2008 11:39:50 -0000 1.177
+++ gdb/symtab.c 2 May 2008 13:36:52 -0000
@@ -1627,6 +1627,26 @@ lookup_symbol_global (const char *name,
domain, symtab);
}
+static int
+symbol_matches_domain (enum language symbol_language,
+ domain_enum symbol_domain,
+ domain_enum domain)
+{
+ /* For c++ "struct foo { ... }" also defines a typedef for "foo".
+ A Java class declaration also defines a typedef for the class.
+ */
+ if (symbol_language == language_cplus
+ || symbol_language == language_java
+ || symbol_language == language_ada)
+ {
+ if ((domain == VAR_DOMAIN || domain == STRUCT_DOMAIN)
+ && symbol_domain == STRUCT_DOMAIN)
+ return 1;
+ }
+ /* For all other languages, strict match is required. */
+ return (symbol_domain == domain);
+}
+
/* Look, in partial_symtab PST, for symbol whose natural name is NAME.
If LINKAGE_NAME is non-NULL, check in addition that the symbol's
linkage name matches it. Check the global symbols if GLOBAL, the
@@ -1691,10 +1711,9 @@ lookup_partial_symbol (struct partial_sy
? strcmp (SYMBOL_LINKAGE_NAME (*top), linkage_name) == 0
: SYMBOL_MATCHES_SEARCH_NAME (*top,name)))
{
- if (SYMBOL_DOMAIN (*top) == domain)
- {
- return (*top);
- }
+ if (symbol_matches_domain (SYMBOL_LANGUAGE (*top),
+ SYMBOL_DOMAIN (*top), domain))
+ return (*top);
top++;
}
}
@@ -1706,7 +1725,8 @@ lookup_partial_symbol (struct partial_sy
{
for (psym = start; psym < start + length; psym++)
{
- if (domain == SYMBOL_DOMAIN (*psym))
+ if (symbol_matches_domain (SYMBOL_LANGUAGE (*psym),
+ SYMBOL_DOMAIN (*psym), domain))
{
if (linkage_name != NULL
? strcmp (SYMBOL_LINKAGE_NAME (*psym), linkage_name) == 0
@@ -1891,7 +1911,8 @@ lookup_block_symbol (const struct block
sym != NULL;
sym = dict_iter_name_next (name, &iter))
{
- if (SYMBOL_DOMAIN (sym) == domain
+ if (symbol_matches_domain (SYMBOL_LANGUAGE (sym),
+ SYMBOL_DOMAIN (sym), domain)
&& (linkage_name != NULL
? strcmp (SYMBOL_LINKAGE_NAME (sym), linkage_name) == 0 : 1))
return sym;
@@ -1912,7 +1933,8 @@ lookup_block_symbol (const struct block
sym != NULL;
sym = dict_iter_name_next (name, &iter))
{
- if (SYMBOL_DOMAIN (sym) == domain
+ if (symbol_matches_domain (SYMBOL_LANGUAGE (sym),
+ SYMBOL_DOMAIN (sym), domain)
&& (linkage_name != NULL
? strcmp (SYMBOL_LINKAGE_NAME (sym), linkage_name) == 0 : 1))
{
WARNING: multiple messages have this Message-ID
From: Aleksandar Ristovski <aristovski@qnx.com>
To: Daniel Jacobowitz <drow@false.org>
Cc: GDB Patches <gdb-patches@sourceware.org>, brobecker@adacore.com
Subject: Re: [rfc] get rid of redundant data in c++ and java
Date: Fri, 02 May 2008 13:39:00 -0000 [thread overview]
Message-ID: <481B198F.30701@qnx.com> (raw)
Message-ID: <20080502133900.V528MfpvXxic-jbGuaLJi8wHL8d77BT-aOH9s6jM450@z> (raw)
In-Reply-To: <20080501192532.GG22218@caradoc.them.org>
[-- Attachment #1: Type: text/plain, Size: 2766 bytes --]
Daniel Jacobowitz wrote:
> On Sat, Feb 09, 2008 at 12:21:02AM -0500, Aleksandar Ristovski wrote:
>> Hello,
>>
>> Currently we synthesize typedef info for java and c++ (and ada, but I am
>> not familiar with that language) for any symbol with STRUCT_DOMAIN domain.
>> This can add up to a lot of redundant data. For partial symbols, we
>> generate two identical names for the same symbol, once as struct second as
>> typedef. The same happens with symbols, except we do not duplicate symbol
>> name, but we do allocate another symbol struture.
>>
>> Instead of generating redundant data for symbols with STRUCT_DOMAIN domain
>> in java and c++ languages, treat STRUCT_DOMAIN as equal to VAR_DOMAIN for
>> these two languages when looking up symbols.
>
> Sorry for taking so long to look at this. It's a great idea and the
> patch seems correct to me. Just some cosmetic issues, and then it can
> go in.
>
>> - if (cu->language == language_cplus
>> - || cu->language == language_java
>> - || cu->language == language_ada)
>> + if (cu->language == language_ada)
>> {
>> - /* For C++ and Java, these implicitly act as typedefs as well. */
>> + /* FIXME: Check if Ada really
>> + needs to implicitly set typedef. */
>
> If this works for C++ and Java, it will work for Ada too. If you
> can't test Ada, please post a final version of the patch and ask Joel
> to run tests for you - I'm sure he won't mind getting all that memory
> back for Ada too.
Ok, here is the revised patch. Cc-ing Joel to take a look if it affects ada.
>> -
>> - /* The semantics of C++ state that "struct foo { ... }" also
>> - defines a typedef for "foo". A Java class declaration also
>> - defines a typedef for the class. Synthesize a typedef symbol
>> - so that "ptype foo" works as expected. */
>> +
>
> Since we still need to set the typedef name here, leave the first two
> sentences of the comment.
Sentences put back.
>
>> + if (symbol_language == language_cplus
>> + || symbol_language == language_java)
>
> Only need one space there.
>
One space.
Waiting for an OK to commit.
Thanks,
Aleksandar
ChangeLog:
* dwarf2read.c (add_partial_symbol): Do not add new psym for
STRUCT_DOMAIN. Make sure you recognize c++ struct and java and ada
class as typedefs. See lookup_partial_symbol function.
(new_symbol): Similar to add_partial_symbol, do not create
symbol for the typedef. See lookup_block_symbol.
* symtab.c (symbol_matches_domain): New function, takes care
of dual meaning of STRUCT_DOMAIN symbol for c++, ada and java.
(lookup_partial_symbol): Use symbol_matches_domain to see if the
found psym domain matches the given domain.
(lookup_block_symbol): Likewise.
[-- Attachment #2: duplicatesymbolfix20080502.diff --]
[-- Type: text/plain, Size: 4692 bytes --]
Index: gdb/dwarf2read.c
===================================================================
RCS file: /cvs/src/src/gdb/dwarf2read.c,v
retrieving revision 1.258
diff -u -p -r1.258 dwarf2read.c
--- gdb/dwarf2read.c 25 Apr 2008 18:45:23 -0000 1.258
+++ gdb/dwarf2read.c 2 May 2008 13:36:52 -0000
@@ -2060,16 +2060,6 @@ add_partial_symbol (struct partial_die_i
: &objfile->static_psymbols,
0, (CORE_ADDR) 0, cu->language, objfile);
- if (cu->language == language_cplus
- || cu->language == language_java
- || cu->language == language_ada)
- {
- /* For C++ and Java, these implicitly act as typedefs as well. */
- add_psymbol_to_list (actual_name, strlen (actual_name),
- VAR_DOMAIN, LOC_TYPEDEF,
- &objfile->global_psymbols,
- 0, (CORE_ADDR) 0, cu->language, objfile);
- }
break;
case DW_TAG_enumerator:
add_psymbol_to_list (actual_name, strlen (actual_name),
@@ -7598,23 +7588,16 @@ new_symbol (struct die_info *die, struct
/* The semantics of C++ state that "struct foo { ... }" also
defines a typedef for "foo". A Java class declaration also
- defines a typedef for the class. Synthesize a typedef symbol
- so that "ptype foo" works as expected. */
+ defines a typedef for the class. */
if (cu->language == language_cplus
|| cu->language == language_java
|| cu->language == language_ada)
{
- struct symbol *typedef_sym = (struct symbol *)
- obstack_alloc (&objfile->objfile_obstack,
- sizeof (struct symbol));
- *typedef_sym = *sym;
- SYMBOL_DOMAIN (typedef_sym) = VAR_DOMAIN;
/* The symbol's name is already allocated along with
this objfile, so we don't need to duplicate it for
the type. */
if (TYPE_NAME (SYMBOL_TYPE (sym)) == 0)
TYPE_NAME (SYMBOL_TYPE (sym)) = SYMBOL_SEARCH_NAME (sym);
- add_symbol_to_list (typedef_sym, list_to_add);
}
}
break;
Index: gdb/symtab.c
===================================================================
RCS file: /cvs/src/src/gdb/symtab.c,v
retrieving revision 1.177
diff -u -p -r1.177 symtab.c
--- gdb/symtab.c 19 Apr 2008 11:39:50 -0000 1.177
+++ gdb/symtab.c 2 May 2008 13:36:52 -0000
@@ -1627,6 +1627,26 @@ lookup_symbol_global (const char *name,
domain, symtab);
}
+static int
+symbol_matches_domain (enum language symbol_language,
+ domain_enum symbol_domain,
+ domain_enum domain)
+{
+ /* For c++ "struct foo { ... }" also defines a typedef for "foo".
+ A Java class declaration also defines a typedef for the class.
+ */
+ if (symbol_language == language_cplus
+ || symbol_language == language_java
+ || symbol_language == language_ada)
+ {
+ if ((domain == VAR_DOMAIN || domain == STRUCT_DOMAIN)
+ && symbol_domain == STRUCT_DOMAIN)
+ return 1;
+ }
+ /* For all other languages, strict match is required. */
+ return (symbol_domain == domain);
+}
+
/* Look, in partial_symtab PST, for symbol whose natural name is NAME.
If LINKAGE_NAME is non-NULL, check in addition that the symbol's
linkage name matches it. Check the global symbols if GLOBAL, the
@@ -1691,10 +1711,9 @@ lookup_partial_symbol (struct partial_sy
? strcmp (SYMBOL_LINKAGE_NAME (*top), linkage_name) == 0
: SYMBOL_MATCHES_SEARCH_NAME (*top,name)))
{
- if (SYMBOL_DOMAIN (*top) == domain)
- {
- return (*top);
- }
+ if (symbol_matches_domain (SYMBOL_LANGUAGE (*top),
+ SYMBOL_DOMAIN (*top), domain))
+ return (*top);
top++;
}
}
@@ -1706,7 +1725,8 @@ lookup_partial_symbol (struct partial_sy
{
for (psym = start; psym < start + length; psym++)
{
- if (domain == SYMBOL_DOMAIN (*psym))
+ if (symbol_matches_domain (SYMBOL_LANGUAGE (*psym),
+ SYMBOL_DOMAIN (*psym), domain))
{
if (linkage_name != NULL
? strcmp (SYMBOL_LINKAGE_NAME (*psym), linkage_name) == 0
@@ -1891,7 +1911,8 @@ lookup_block_symbol (const struct block
sym != NULL;
sym = dict_iter_name_next (name, &iter))
{
- if (SYMBOL_DOMAIN (sym) == domain
+ if (symbol_matches_domain (SYMBOL_LANGUAGE (sym),
+ SYMBOL_DOMAIN (sym), domain)
&& (linkage_name != NULL
? strcmp (SYMBOL_LINKAGE_NAME (sym), linkage_name) == 0 : 1))
return sym;
@@ -1912,7 +1933,8 @@ lookup_block_symbol (const struct block
sym != NULL;
sym = dict_iter_name_next (name, &iter))
{
- if (SYMBOL_DOMAIN (sym) == domain
+ if (symbol_matches_domain (SYMBOL_LANGUAGE (sym),
+ SYMBOL_DOMAIN (sym), domain)
&& (linkage_name != NULL
? strcmp (SYMBOL_LINKAGE_NAME (sym), linkage_name) == 0 : 1))
{
next prev parent reply other threads:[~2008-05-02 13:40 UTC|newest]
Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top
2008-02-09 5:21 Aleksandar Ristovski
2008-02-14 0:16 ` Aleksandar Ristovski
2008-02-14 1:35 ` Daniel Jacobowitz
2008-05-01 19:25 ` Daniel Jacobowitz
2008-05-02 13:43 ` Aleksandar Ristovski [this message]
2008-05-02 13:39 ` Aleksandar Ristovski
2008-05-02 14:01 ` Daniel Jacobowitz
2008-05-03 0:01 ` Joel Brobecker
2008-05-05 16:15 ` Aleksandar Ristovski
2008-05-05 15:16 ` Aleksandar Ristovski
2008-04-22 16:46 Aleksandar Ristovski
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=481B198F.30701@qnx.com \
--to=aristovski@qnx.com \
--cc=brobecker@adacore.com \
--cc=gdb-patches@sources.redhat.com \
--cc=gdb-patches@sourceware.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox