RFA: fix PR gdb/2489

RFA: fix PR gdb/2489

[Tom Tromey]

This patch fixes PR gdb/2489.

The bug is that field name completion does not consider methods.

The fix is to also examine method names when completing; but not to
consider constructor names.

Built and regression tested on x86-64 (compile farm).
New test case included.

Ok?

Tom

b/gdb/ChangeLog:
2008-08-01  Tom Tromey  <tromey@redhat.com>

	PR gdb/2489:
	* completer.c (count_struct_fields): Count method names.
	(add_struct_fields): Add matching method names.

b/gdb/testsuite/ChangeLog:
2008-08-01  Tom Tromey  <tromey@redhat.com>

	* gdb.cp/pr2489.cc: New file.
	* gdb.cp/cpcompletion.exp: New file.

diff --git a/gdb/completer.c b/gdb/completer.c
index e7ee817..e42d8a3 100644
--- a/gdb/completer.c
+++ b/gdb/completer.c
@@ -339,7 +339,7 @@ location_completer (char *text, char *word)
 }
 
 /* Helper for expression_completer which recursively counts the number
-   of named fields in a structure or union type.  */
+   of named fields and methods in a structure or union type.  */
 static int
 count_struct_fields (struct type *type)
 {
@@ -353,17 +353,25 @@ count_struct_fields (struct type *type)
       else if (TYPE_FIELD_NAME (type, i))
 	++result;
     }
+
+  for (i = TYPE_NFN_FIELDS (type) - 1; i >=0; --i)
+    {
+      if (TYPE_FN_FIELDLIST_NAME (type, i))
+	++result;
+    }
+
   return result;
 }
 
-/* Helper for expression_completer which recursively adds field names
-   from TYPE, a struct or union type, to the array OUTPUT.  This
-   function assumes that OUTPUT is correctly-sized.  */
+/* Helper for expression_completer which recursively adds field and
+   method names from TYPE, a struct or union type, to the array
+   OUTPUT.  This function assumes that OUTPUT is correctly-sized.  */
 static void
 add_struct_fields (struct type *type, int *nextp, char **output,
 		   char *fieldname, int namelen)
 {
   int i;
+  char *type_name = NULL;
 
   CHECK_TYPEDEF (type);
   for (i = 0; i < TYPE_NFIELDS (type); ++i)
@@ -378,6 +386,22 @@ add_struct_fields (struct type *type, int *nextp, char **output,
 	  ++*nextp;
 	}
     }
+
+  for (i = TYPE_NFN_FIELDS (type) - 1; i >=0; --i)
+    {
+      char *name = TYPE_FN_FIELDLIST_NAME (type, i);
+      if (name && ! strncmp (name, fieldname, namelen))
+	{
+	  if (!type_name)
+	    type_name = type_name_no_tag (type);
+	  /* Omit constructors from the completion list.  */
+	  if (strcmp (type_name, name))
+	    {
+	      output[*nextp] = xstrdup (name);
+	      ++*nextp;
+	    }
+	}
+    }
 }
 
 /* Complete on expressions.  Often this means completing on symbol
diff --git a/gdb/testsuite/gdb.cp/cpcompletion.exp b/gdb/testsuite/gdb.cp/cpcompletion.exp
new file mode 100644
index 0000000..1f9caa6
--- /dev/null
+++ b/gdb/testsuite/gdb.cp/cpcompletion.exp
@@ -0,0 +1,78 @@
+# Copyright 2008 Free Software Foundation, Inc.
+
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+# Please email any bugs, comments, and/or additions to this file to:
+# bug-gdb@prep.ai.mit.edu
+
+# This file is part of the gdb testsuite.
+
+if $tracelevel then {
+	strace $tracelevel
+}
+
+if { [skip_cplus_tests] } { continue }
+
+set testfile pr2489
+set binfile ${objdir}/${subdir}/${testfile}
+
+if {[gdb_compile "${srcdir}/${subdir}/${testfile}.cc" "${testfile}.o" object {c++ debug}] != ""} {
+    untested completion.exp
+    return -1
+}
+
+if {[gdb_compile "${testfile}.o" ${binfile} executable {c++ debug}] != "" } {
+     untested completion.exp
+     return -1
+}
+
+gdb_exit
+
+# Don't let a .inputrc file or an existing setting of INPUTRC mess up
+# the test results.  Even if /dev/null doesn't exist on the particular
+# platform, the readline library will use the default setting just by
+# failing to open the file.  OTOH, opening /dev/null successfully will
+# also result in the default settings being used since nothing will be
+# read from this file.
+global env
+if [info exists env(INPUTRC)] {
+    set old_inputrc $env(INPUTRC)
+}
+set env(INPUTRC) "/dev/null"
+
+gdb_start
+gdb_reinitialize_dir $srcdir/$subdir
+gdb_load ${binfile}
+
+set bp_location [gdb_get_line_number "Set breakpoint here" ${testfile}.cc]
+
+if {![runto $bp_location]} {
+    perror "test suppressed"
+}
+
+send_gdb "p foo1.g\t"
+gdb_expect  {
+        -re "^p foo1\\.get_foo $"\
+            { send_gdb "()\n"
+              gdb_expect {
+                      -re "^.* = 0.*$gdb_prompt $"\
+                                        { pass "complete 'p foo1.g'"}
+                      -re ".*$gdb_prompt $" { fail "complete 'p foo1.g'"}
+                      timeout           {fail "(timeout) complete 'p foo1.g'"}
+                     }
+            }
+        -re ".*$gdb_prompt $"       { fail "complete 'p foo1.g'" }
+        timeout         { fail "(timeout) complete 'p foo1.g' 2" }
+        }
+
diff --git a/gdb/testsuite/gdb.cp/pr2489.cc b/gdb/testsuite/gdb.cp/pr2489.cc
new file mode 100644
index 0000000..f735205
--- /dev/null
+++ b/gdb/testsuite/gdb.cp/pr2489.cc
@@ -0,0 +1,34 @@
+
+class Foo    
+{
+
+private:
+  int foo_value;
+
+public:
+  Foo () { foo_value = 0;}
+  Foo (int i) { foo_value = i;}
+  ~Foo () { }
+  void set_foo (int value);
+  int get_foo ();
+
+  bool operator== (const Foo &other) { return foo_value == other.foo_value; }
+};
+ 
+void Foo::set_foo (int value)
+{
+  foo_value = value;
+}
+
+int Foo::get_foo ()
+{
+  return foo_value;
+}
+
+int main ()
+{
+  Foo foo1;
+  foo1.set_foo (42);		// Set breakpoint here.
+  return 0;
+}
+

Re: RFA: fix PR gdb/2489

[Pedro Alves]

On Friday 01 August 2008 20:14:50, Tom Tromey wrote:
> This patch fixes PR gdb/2489.
> 
> The bug is that field name completion does not consider methods.
> 
> The fix is to also examine method names when completing; but not to
> consider constructor names.
> 

> b/gdb/ChangeLog:
> 2008-08-01  Tom Tromey  <tromey@redhat.com>
> 
> 	PR gdb/2489:
> 	* completer.c (count_struct_fields): Count method names.
> 	(add_struct_fields): Add matching method names.
> 
> b/gdb/testsuite/ChangeLog:
> 2008-08-01  Tom Tromey  <tromey@redhat.com>
> 
> 	* gdb.cp/pr2489.cc: New file.
> 	* gdb.cp/cpcompletion.exp: New file.
> 

This looks mostly OK to me.  This patch is pending for a few months
now, so it shouldn't hurt to wait a couple of days more to give the
C++ maintainer a chance to comment.  :-)

> +  for (i = TYPE_NFN_FIELDS (type) - 1; i >=0; --i)
                                              ^ missing space

We have this description in gdbtypes.c:

  /* Return a typename for a struct/union/enum type without "struct ",
     "union ", or "enum ".  If the type has a NULL name, return NULL.  */

  char *
  type_name_no_tag (const struct type *type)
  {


+  for (i = TYPE_NFN_FIELDS (type) - 1; i >=0; --i)
+    {
+      char *name = TYPE_FN_FIELDLIST_NAME (type, i);
+      if (name && ! strncmp (name, fieldname, namelen))
+       {
+         if (!type_name)
+           type_name = type_name_no_tag (type);
+         /* Omit constructors from the completion list.  */
+         if (strcmp (type_name, name))
+           {

Can type_name ever be NULL here then?  I'm no guru in this
area, but I think not.

> +# Please email any bugs, comments, and/or additions to this file to:
> +# bug-gdb@prep.ai.mit.edu

Please remove this bit.  Jan Kratochvil has recently gone through removing
all references to this long gone address.  (It seems a couple have crept
in since though.)

> +
> +if {[gdb_compile "${srcdir}/${subdir}/${testfile}.cc" "${testfile}.o" object {c++ debug}] != ""} {
> +    untested completion.exp
> +    return -1
> +}
> +
> +if {[gdb_compile "${testfile}.o" ${binfile} executable {c++ debug}] != "" } {
> +     untested completion.exp
> +     return -1
> +}

s/completions.exp/cpcompletion.exp/

> +
> +send_gdb "p foo1.g\t"
> +gdb_expect  {
> +        -re "^p foo1\\.get_foo $"\
> +            { send_gdb "()\n"
> +              gdb_expect {
> +                      -re "^.* = 0.*$gdb_prompt $"\
> +                                        { pass "complete 'p foo1.g'"}
> +                      -re ".*$gdb_prompt $" { fail "complete 'p foo1.g'"}
> +                      timeout           {fail "(timeout) complete 'p foo1.g'"}
> +                     }
> +            }
> +        -re ".*$gdb_prompt $"       { fail "complete 'p foo1.g'" }
> +        timeout         { fail "(timeout) complete 'p foo1.g' 2" }
> +        }
> +

(I wish we had a function we could call that abstracted and made easier to
write/read these completion tests.)

What do you think about extending the test a little bit?

I think it would be nice to have,

Plain inheritance testing that the base class methods are completed.

--

Inheritance + masking, something like:

 class FooBase
 {
 public:
    int get_foo () { return 1; }
 }

 class Foo : public FooBase
 {
 public:
    int get_foo () { ... }
 }

 Foo foo1;

From my testing, foo1.g<tab> will make get_foo only show up once,
which is fine, IMO.

A possible *future* improvement would be go show:

foo1.<tab>
 foo1.FooBase::get_foo
 foo1.get_foo

 (foo1.Foo::get_foo too perhaps, not sure)

--

Masking, changing type:

 class FooBase
 {
 private:
   int get_foo;
 }

 class Foo : public FooBase
 {
 public:
    int get_foo () { ... }
 }

 Foo foo1;

This case, foo1.g<tab> completes to foo1.get_foo, but 'p foo1.get_foo'
prints FooBase::get_foo.  It's a bit confusing, especially since
get_foo is private in the base class --- not related to your patch,
just pointing it out.  This is somewhat related to the future improvement I
mentioned above.  That is, maybe showing:

 foo1.<tab>
 foo1.FooBase::get_foo
 foo1.get_foo

Or:

 foo1.g<tab>
 foo1.get_foo
 foo1.get_foo(

would make things clearer.

--

Anonymous struct with method,

struct
{
   int get_foo () { ... }
} a;

a.<tab>
 get_foo

--

Also, would it make sense to add a test that made sure the
ctors aren't completed?

--

Consider any test additions you make pre-approved.

-- 
Pedro Alves

Re: RFA: fix PR gdb/2489

[Tom Tromey]

>>>>> "Pedro" == Pedro Alves <pedro@codesourcery.com> writes:

Finally getting back to this old patch...

Tom> This patch fixes PR gdb/2489.
Tom> The bug is that field name completion does not consider methods.

Tom> +  for (i = TYPE_NFN_FIELDS (type) - 1; i >=0; --i)
Pedro>                                               ^ missing space

Fixed.

Pedro> +  for (i = TYPE_NFN_FIELDS (type) - 1; i >=0; --i)
Pedro> +    {
Pedro> +      char *name = TYPE_FN_FIELDLIST_NAME (type, i);
Pedro> +      if (name && ! strncmp (name, fieldname, namelen))
Pedro> +       {
Pedro> +         if (!type_name)
Pedro> +           type_name = type_name_no_tag (type);
Pedro> +         /* Omit constructors from the completion list.  */
Pedro> +         if (strcmp (type_name, name))
Pedro> +           {

Pedro> Can type_name ever be NULL here then?

The reason for the check here is that we compute type_name the first
time we need it.  It is initialized to NULL, but only set once.

Tom> +# Please email any bugs, comments, and/or additions to this file to:
Tom> +# bug-gdb@prep.ai.mit.edu

Pedro> Please remove this bit.

Whoops, fixed.

Tom> +     untested completion.exp

Pedro> s/completions.exp/cpcompletion.exp/

Thanks -- fixed.

Pedro> (I wish we had a function we could call that abstracted and
Pedro> made easier to write/read these completion tests.)

I rewrote the tests to use the "complete" command rather than sending
a TAB.  This makes them much simpler.

Pedro> What do you think about extending the test a little bit?

Pedro> I think it would be nice to have,
Pedro> Plain inheritance testing that the base class methods are completed.
[...]
Pedro> Inheritance + masking, something like:
[...]
Pedro> Anonymous struct with method,
[...]
Pedro> Also, would it make sense to add a test that made sure the
Pedro> ctors aren't completed?

I added these, thanks for the suggestions.

I've appended the final patch.  I am checking this in.

Tom

2009-02-02  Tom Tromey  <tromey@redhat.com>

	PR gdb/2489:
	* completer.c (count_struct_fields): Count method names.
	(add_struct_fields): Add matching method names.

2009-02-02  Tom Tromey  <tromey@redhat.com>

	* gdb.cp/Makefile.in (EXECUTABLES): Add pr2489.
	* gdb.cp/pr2489.cc: New file.
	* gdb.cp/cpcompletion.exp: New file.

Index: completer.c
===================================================================
RCS file: /cvs/src/src/gdb/completer.c,v
retrieving revision 1.30
diff -u -r1.30 completer.c
--- completer.c	3 Jan 2009 05:57:51 -0000	1.30
+++ completer.c	3 Feb 2009 00:56:41 -0000
@@ -339,7 +339,7 @@
 }
 
 /* Helper for expression_completer which recursively counts the number
-   of named fields in a structure or union type.  */
+   of named fields and methods in a structure or union type.  */
 static int
 count_struct_fields (struct type *type)
 {
@@ -353,17 +353,25 @@
       else if (TYPE_FIELD_NAME (type, i))
 	++result;
     }
+
+  for (i = TYPE_NFN_FIELDS (type) - 1; i >= 0; --i)
+    {
+      if (TYPE_FN_FIELDLIST_NAME (type, i))
+	++result;
+    }
+
   return result;
 }
 
-/* Helper for expression_completer which recursively adds field names
-   from TYPE, a struct or union type, to the array OUTPUT.  This
-   function assumes that OUTPUT is correctly-sized.  */
+/* Helper for expression_completer which recursively adds field and
+   method names from TYPE, a struct or union type, to the array
+   OUTPUT.  This function assumes that OUTPUT is correctly-sized.  */
 static void
 add_struct_fields (struct type *type, int *nextp, char **output,
 		   char *fieldname, int namelen)
 {
   int i;
+  char *type_name = NULL;
 
   CHECK_TYPEDEF (type);
   for (i = 0; i < TYPE_NFIELDS (type); ++i)
@@ -378,6 +386,22 @@
 	  ++*nextp;
 	}
     }
+
+  for (i = TYPE_NFN_FIELDS (type) - 1; i >= 0; --i)
+    {
+      char *name = TYPE_FN_FIELDLIST_NAME (type, i);
+      if (name && ! strncmp (name, fieldname, namelen))
+	{
+	  if (!type_name)
+	    type_name = type_name_no_tag (type);
+	  /* Omit constructors from the completion list.  */
+	  if (strcmp (type_name, name))
+	    {
+	      output[*nextp] = xstrdup (name);
+	      ++*nextp;
+	    }
+	}
+    }
 }
 
 /* Complete on expressions.  Often this means completing on symbol
Index: testsuite/gdb.cp/Makefile.in
===================================================================
RCS file: /cvs/src/src/gdb/testsuite/gdb.cp/Makefile.in,v
retrieving revision 1.3
diff -u -r1.3 Makefile.in
--- testsuite/gdb.cp/Makefile.in	28 Mar 2007 00:32:41 -0000	1.3
+++ testsuite/gdb.cp/Makefile.in	3 Feb 2009 00:56:43 -0000
@@ -4,7 +4,7 @@
 EXECUTABLES = ambiguous annota2 anon-union cplusfuncs cttiadd \
 	derivation inherit local member-ptr method misc \
         overload ovldbreak ref-typ ref-typ2 templates userdef virtfunc namespace \
-	ref-types ref-params method2
+	ref-types ref-params method2 pr2489
 
 all info install-info dvi install uninstall installcheck check:
 	@echo "Nothing to be done for $@..."
Index: testsuite/gdb.cp/cpcompletion.exp
===================================================================
RCS file: testsuite/gdb.cp/cpcompletion.exp
diff -N testsuite/gdb.cp/cpcompletion.exp
--- /dev/null	1 Jan 1970 00:00:00 -0000
+++ testsuite/gdb.cp/cpcompletion.exp	3 Feb 2009 00:56:43 -0000
@@ -0,0 +1,72 @@
+# Copyright 2009 Free Software Foundation, Inc.
+
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+# This file is part of the gdb testsuite.
+
+if $tracelevel then {
+	strace $tracelevel
+}
+
+if { [skip_cplus_tests] } { continue }
+
+set testfile pr2489
+set binfile ${objdir}/${subdir}/${testfile}
+
+if {[gdb_compile "${srcdir}/${subdir}/${testfile}.cc" "${testfile}.o" object {c++ debug}] != ""} {
+    untested cpcompletion.exp
+    return -1
+}
+
+if {[gdb_compile "${testfile}.o" ${binfile} executable {c++ debug}] != "" } {
+     untested cpcompletion.exp
+     return -1
+}
+
+gdb_exit
+
+# Don't let a .inputrc file or an existing setting of INPUTRC mess up
+# the test results.  Even if /dev/null doesn't exist on the particular
+# platform, the readline library will use the default setting just by
+# failing to open the file.  OTOH, opening /dev/null successfully will
+# also result in the default settings being used since nothing will be
+# read from this file.
+global env
+if [info exists env(INPUTRC)] {
+    set old_inputrc $env(INPUTRC)
+}
+set env(INPUTRC) "/dev/null"
+
+gdb_start
+gdb_reinitialize_dir $srcdir/$subdir
+gdb_load ${binfile}
+
+set bp_location [gdb_get_line_number "Set breakpoint here" ${testfile}.cc]
+
+if {![runto $bp_location]} {
+    perror "test suppressed"
+}
+
+# This also tests inheritance -- completion should only see a single
+# "get_foo".
+gdb_test "complete p foo1.g" "p foo1\\.get_foo"
+
+# Test inheritance without overriding.
+gdb_test "complete p foo1.base" "p foo1\\.base_function_only"
+
+# Test non-completion of constructor names.
+gdb_test "complete p foo1.Fo" "p foo1\\.Foofoo"
+
+# Test completion with an anonymous struct.
+gdb_test "complete p a.g" "p a\\.get"
Index: testsuite/gdb.cp/pr2489.cc
===================================================================
RCS file: testsuite/gdb.cp/pr2489.cc
diff -N testsuite/gdb.cp/pr2489.cc
--- /dev/null	1 Jan 1970 00:00:00 -0000
+++ testsuite/gdb.cp/pr2489.cc	3 Feb 2009 00:56:43 -0000
@@ -0,0 +1,51 @@
+
+class Base
+{
+public:
+  virtual int get_foo () { return 1; }
+  int base_function_only () { return 2; }
+};
+
+class Foo : public Base
+{
+
+private:
+  int foo_value;
+
+public:
+  Foo () { foo_value = 0;}
+  Foo (int i) { foo_value = i;}
+  ~Foo () { }
+  void set_foo (int value);
+  int get_foo ();
+
+  // Something similar to a constructor name.
+  void Foofoo ();
+
+  bool operator== (const Foo &other) { return foo_value == other.foo_value; }
+};
+ 
+void Foo::set_foo (int value)
+{
+  foo_value = value;
+}
+
+int Foo::get_foo ()
+{
+  return foo_value;
+}
+
+void Foo::Foofoo ()
+{
+}
+
+int main ()
+{
+  // Anonymous struct with method.
+  struct {
+    int get() { return 5; }
+  } a;
+  Foo foo1;
+  foo1.set_foo (42);		// Set breakpoint here.
+  return 0;
+}

Re: RFA: fix PR gdb/2489

[Tom Tromey]

>>>>> "Tom" == Tom Tromey <tromey@redhat.com> writes:

Tom> 2009-02-02  Tom Tromey  <tromey@redhat.com>
Tom> 	PR gdb/2489:
Tom> 	* completer.c (count_struct_fields): Count method names.
Tom> 	(add_struct_fields): Add matching method names.

Whoops!  After committing, I remembered that this patch predates the
Bugzilla move.

I am checking in this patch to rename the test case.  I'm also going
to fix gdb/ChangeLog to point to the correct bug.

Tom

2009-02-02  Tom Tromey  <tromey@redhat.com>

	* gdb.cp/cpcompletion.exp: Name the test "pr9594".
	* gdb.cp/pr2489.cc: Rename...
	* gdb.cp/pr9594.cc: ... to this.

Index: testsuite/gdb.cp/Makefile.in
===================================================================
RCS file: /cvs/src/src/gdb/testsuite/gdb.cp/Makefile.in,v
retrieving revision 1.4
diff -u -r1.4 Makefile.in
--- testsuite/gdb.cp/Makefile.in	3 Feb 2009 01:00:40 -0000	1.4
+++ testsuite/gdb.cp/Makefile.in	3 Feb 2009 01:06:12 -0000
@@ -4,7 +4,7 @@
 EXECUTABLES = ambiguous annota2 anon-union cplusfuncs cttiadd \
 	derivation inherit local member-ptr method misc \
         overload ovldbreak ref-typ ref-typ2 templates userdef virtfunc namespace \
-	ref-types ref-params method2 pr2489
+	ref-types ref-params method2 pr9594
 
 all info install-info dvi install uninstall installcheck check:
 	@echo "Nothing to be done for $@..."
Index: testsuite/gdb.cp/cpcompletion.exp
===================================================================
RCS file: /cvs/src/src/gdb/testsuite/gdb.cp/cpcompletion.exp,v
retrieving revision 1.1
diff -u -r1.1 cpcompletion.exp
--- testsuite/gdb.cp/cpcompletion.exp	3 Feb 2009 01:00:40 -0000	1.1
+++ testsuite/gdb.cp/cpcompletion.exp	3 Feb 2009 01:06:13 -0000
@@ -21,7 +21,7 @@
 
 if { [skip_cplus_tests] } { continue }
 
-set testfile pr2489
+set testfile pr9594
 set binfile ${objdir}/${subdir}/${testfile}
 
 if {[gdb_compile "${srcdir}/${subdir}/${testfile}.cc" "${testfile}.o" object {c++ debug}] != ""} {
Index: testsuite/gdb.cp/pr2489.cc
===================================================================
RCS file: testsuite/gdb.cp/pr2489.cc
diff -N testsuite/gdb.cp/pr2489.cc
--- testsuite/gdb.cp/pr2489.cc	3 Feb 2009 01:00:40 -0000	1.1
+++ /dev/null	1 Jan 1970 00:00:00 -0000
@@ -1,51 +0,0 @@
-
-class Base
-{
-public:
-  virtual int get_foo () { return 1; }
-  int base_function_only () { return 2; }
-};
-
-class Foo : public Base
-{
-
-private:
-  int foo_value;
-
-public:
-  Foo () { foo_value = 0;}
-  Foo (int i) { foo_value = i;}
-  ~Foo () { }
-  void set_foo (int value);
-  int get_foo ();
-
-  // Something similar to a constructor name.
-  void Foofoo ();
-
-  bool operator== (const Foo &other) { return foo_value == other.foo_value; }
-};
- 
-void Foo::set_foo (int value)
-{
-  foo_value = value;
-}
-
-int Foo::get_foo ()
-{
-  return foo_value;
-}
-
-void Foo::Foofoo ()
-{
-}
-
-int main ()
-{
-  // Anonymous struct with method.
-  struct {
-    int get() { return 5; }
-  } a;
-  Foo foo1;
-  foo1.set_foo (42);		// Set breakpoint here.
-  return 0;
-}
Index: testsuite/gdb.cp/pr9594.cc
===================================================================
RCS file: testsuite/gdb.cp/pr9594.cc
diff -N testsuite/gdb.cp/pr9594.cc
--- /dev/null	1 Jan 1970 00:00:00 -0000
+++ testsuite/gdb.cp/pr9594.cc	3 Feb 2009 01:06:13 -0000
@@ -0,0 +1,51 @@
+
+class Base
+{
+public:
+  virtual int get_foo () { return 1; }
+  int base_function_only () { return 2; }
+};
+
+class Foo : public Base
+{
+
+private:
+  int foo_value;
+
+public:
+  Foo () { foo_value = 0;}
+  Foo (int i) { foo_value = i;}
+  ~Foo () { }
+  void set_foo (int value);
+  int get_foo ();
+
+  // Something similar to a constructor name.
+  void Foofoo ();
+
+  bool operator== (const Foo &other) { return foo_value == other.foo_value; }
+};
+ 
+void Foo::set_foo (int value)
+{
+  foo_value = value;
+}
+
+int Foo::get_foo ()
+{
+  return foo_value;
+}
+
+void Foo::Foofoo ()
+{
+}
+
+int main ()
+{
+  // Anonymous struct with method.
+  struct {
+    int get() { return 5; }
+  } a;
+  Foo foo1;
+  foo1.set_foo (42);		// Set breakpoint here.
+  return 0;
+}

Re: RFA: fix PR gdb/2489

[Pedro Alves]

On Tuesday 03 February 2009 00:59:46, Tom Tromey wrote:
> Pedro> +  for (i = TYPE_NFN_FIELDS (type) - 1; i >=0; --i)
> Pedro> +    {
> Pedro> +      char *name = TYPE_FN_FIELDLIST_NAME (type, i);
> Pedro> +      if (name && ! strncmp (name, fieldname, namelen))
> Pedro> +       {
> Pedro> +         if (!type_name)
> Pedro> +           type_name = type_name_no_tag (type);
> Pedro> +         /* Omit constructors from the completion list.  */
> Pedro> +         if (strcmp (type_name, name))
> Pedro> +           {
> 
> Pedro> Can type_name ever be NULL here then?
> 
> The reason for the check here is that we compute type_name the first
> time we need it.  It is initialized to NULL, but only set once.

Oh, sorry, I guess I wasn't clear.  I meant after the
type_name_no_tag call, at the strcmp line.

I mainly asked due to this:

 /* Return a typename for a struct/union/enum type without "struct ",
    "union ", or "enum ".  If the type has a NULL name, return NULL.  */

                           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 char *
 type_name_no_tag (const struct type *type)
 {
   if (TYPE_TAG_NAME (type) != NULL)
     return TYPE_TAG_NAME (type);

   /* Is there code which expects this to return the name if there is
      no tag name?  My guess is that this is mainly used for C++ in
      cases where the two will always be the same.  */
   return TYPE_NAME (type);
 }

I don't know if that can happen here.  That was also the reason I
suggested an annonymous struct/class test.

On Tuesday 03 February 2009 00:59:46, Tom Tromey wrote:
> Pedro> (I wish we had a function we could call that abstracted and
> Pedro> made easier to write/read these completion tests.)
> 
> I rewrote the tests to use the "complete" command rather than sending
> a TAB.  This makes them much simpler.
> 

Indeed!  Much nicer.

-- 
Pedro Alves

Re: RFA: fix PR gdb/2489

[Tom Tromey]

>>>>> "Pedro" == Pedro Alves <pedro@codesourcery.com> writes:

Pedro> Oh, sorry, I guess I wasn't clear.  I meant after the
Pedro> type_name_no_tag call, at the strcmp line.

Oh, you were plenty clear -- I just misunderstood.

Pedro> I mainly asked due to this:
Pedro>  /* Return a typename for a struct/union/enum type without "struct ",
Pedro>     "union ", or "enum ".  If the type has a NULL name, return NULL.  */

Pedro>                            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Pedro>  char *
Pedro>  type_name_no_tag (const struct type *type)
[...]

Pedro> I don't know if that can happen here.  That was also the reason I
Pedro> suggested an annonymous struct/class test.

I looked at the anonymous test struct case and I see:

(top) p type_name
$2 = 0x8fa3f1e "<anonymous struct>"

However, this appears to come from g++:

 <2><361>: Abbrev Number: 33 (DW_TAG_structure_type)
    <362>   DW_AT_name        : (indirect string, offset: 0x172): <anonymous struct>    

I seem to recall that this is actually a change in g++.

I tried with an older g++ and the dwarf says:

 <2><58b>: Abbrev Number: 29 (DW_TAG_structure_type)
     DW_AT_sibling     : <5ab>  
     DW_AT_name        : ._0    

... which just seems weird.

Anyway, I suspect that you are correct, and we could see a NULL here,
but I don't know how to make it happen.  I will add a check for NULL
here tomorrow.

Tom

Re: RFA: fix PR gdb/2489

[Tom Tromey]

>>>>> "Tom" == Tom Tromey <tromey@redhat.com> writes:

Tom> Anyway, I suspect that you are correct, and we could see a NULL here,
Tom> but I don't know how to make it happen.  I will add a check for NULL
Tom> here tomorrow.

I am checking in the appended to address this potential problem.

Built and regtested on x86-64 (compile farm).

thanks,
Tom

2009-02-03  Tom Tromey  <tromey@redhat.com>

	* completer.c (add_struct_fields): Check type_name against NULL
	before use.

diff --git a/gdb/completer.c b/gdb/completer.c
index 5d7225f..298cdd0 100644
--- a/gdb/completer.c
+++ b/gdb/completer.c
@@ -371,6 +371,7 @@ add_struct_fields (struct type *type, int *nextp, char **output,
 		   char *fieldname, int namelen)
 {
   int i;
+  int computed_type_name = 0;
   char *type_name = NULL;
 
   CHECK_TYPEDEF (type);
@@ -392,10 +393,13 @@ add_struct_fields (struct type *type, int *nextp, char **output,
       char *name = TYPE_FN_FIELDLIST_NAME (type, i);
       if (name && ! strncmp (name, fieldname, namelen))
 	{
-	  if (!type_name)
-	    type_name = type_name_no_tag (type);
+	  if (!computed_type_name)
+	    {
+	      type_name = type_name_no_tag (type);
+	      computed_type_name = 1;
+	    }
 	  /* Omit constructors from the completion list.  */
-	  if (strcmp (type_name, name))
+	  if (type_name && strcmp (type_name, name))
 	    {
 	      output[*nextp] = xstrdup (name);
 	      ++*nextp;