Remove unwanted spaces when looking up builtin types

Message ID CAGyQ6gyM5OXG35e+q+3TpTxB-NTjC7DqAhQKxk_ZZYZfKwnSMQ@mail.gmail.com
State New, archived
Headers

Commit Message

Siva Chandra Reddy Nov. 7, 2014, 1:49 p.m. UTC
  On Wed, Oct 29, 2014 at 11:50 AM, Siva Chandra <sivachandra@google.com> wrote:
> Currently, if we do something like gdb.lookup_type(" unsigned long "),
> GDB errors out because of the leading and trailing spaces.  The
> attached patch fixes this problem.
>
> A practical situation where this problem is hit is when invoking
> template methods. Its not uncommon to do things like this:
>
> (gdb) p foo.bar< unsigned long >()
>
> If "bar" happens to be an xmethod, then its implementation will
> typically need to parse the name of the method ("bar< unsigned long >"
> in the above example) to get the template argument and lookup the
> type. GDB currently fails for such cases. One could of course
> sanitize/fix such inputs in Python before calling lookup_type, but I
> think it is better done on the GDB side as having white spaces is
> valid syntax.
>
> For non-builtin types, lookup_type goes through the symbol lookup
> path. AFAIU, that path already ignores whitespaces.

Ping.

gdb/ChangeLog:

2014-11-07  Siva Chandra Reddy  <sivachandra@google.com>

        * language.c (language_lookup_primitive_type_by_name): Remove
        unwanted space in the type name before looking it up.
        (whitespace_p): New function.

gdb/testsuite/ChangeLog:

2014-11-07  Siva Chandra Reddy  <sivachandra@google.com>

        * gdb.python/py-type.exp: Add new tests.
  

Comments

Siva Chandra Reddy Nov. 18, 2014, 11:59 p.m. UTC | #1
On Fri, Nov 7, 2014 at 5:49 AM, Siva Chandra <sivachandra@google.com> wrote:
> On Wed, Oct 29, 2014 at 11:50 AM, Siva Chandra <sivachandra@google.com> wrote:
>> Currently, if we do something like gdb.lookup_type(" unsigned long "),
>> GDB errors out because of the leading and trailing spaces.  The
>> attached patch fixes this problem.
>>
>> A practical situation where this problem is hit is when invoking
>> template methods. Its not uncommon to do things like this:
>>
>> (gdb) p foo.bar< unsigned long >()
>>
>> If "bar" happens to be an xmethod, then its implementation will
>> typically need to parse the name of the method ("bar< unsigned long >"
>> in the above example) to get the template argument and lookup the
>> type. GDB currently fails for such cases. One could of course
>> sanitize/fix such inputs in Python before calling lookup_type, but I
>> think it is better done on the GDB side as having white spaces is
>> valid syntax.
>>
>> For non-builtin types, lookup_type goes through the symbol lookup
>> path. AFAIU, that path already ignores whitespaces.
>

Ping.
This is not super critical, but I see that an xmethod test fails after
I have upgraded to gcc-4.8.2 from gcc-4.6. The essential difference I
have found is that looking up a symbol for "  unsigned int  " works
when the test is compiled with gcc-4.6 and it does not when compiled
with 4.8.2. And, when it does not, the type is looked up via
"language_lookup_primitive_type_by_name" which errors out as it does
an exact string match.

[I have not yet looked into why there is a difference between gcc-4.6
and gcc-4.8.2 as I think that the proposed fix is not inappropriate
and does fix the problem.]

>
> gdb/ChangeLog:
>
> 2014-11-07  Siva Chandra Reddy  <sivachandra@google.com>
>
>         * language.c (language_lookup_primitive_type_by_name): Remove
>         unwanted space in the type name before looking it up.
>         (whitespace_p): New function.
>
> gdb/testsuite/ChangeLog:
>
> 2014-11-07  Siva Chandra Reddy  <sivachandra@google.com>
>
>         * gdb.python/py-type.exp: Add new tests.
  

Patch

diff --git a/gdb/language.c b/gdb/language.c
index 034086d..abfed40 100644
--- a/gdb/language.c
+++ b/gdb/language.c
@@ -981,6 +981,17 @@  language_bool_type (const struct language_defn *la,
   return ld->arch_info[la->la_language].bool_type_default;
 }
 
+/* Return 1 if C is a whitespace character, 0 otherwise.  */
+
+static int
+whitespace_p (const char c)
+{
+  if (c == ' ' || c == '\n' || c == '\t')
+    return 1;
+  else
+    return 0;
+}
+
 struct type *
 language_lookup_primitive_type_by_name (const struct language_defn *la,
 					struct gdbarch *gdbarch,
@@ -989,14 +1000,44 @@  language_lookup_primitive_type_by_name (const struct language_defn *la,
   struct language_gdbarch *ld = gdbarch_data (gdbarch,
 					      language_gdbarch_data);
   struct type *const *p;
+  int len = strlen (name);
+  int i, j;
+  char *clean_name = (char *) xmalloc (sizeof (char) * len + 1);
+
+  /* Remove unwanted whitespace in the typename.  This could happen, for
+     example, happen if one does gdb.lookup_type(' unsigned long ') in
+     Python.  */
+  for (i = 0, j = 0; i < len; i++)
+    {
+      if (whitespace_p (name[i]))
+	{
+	  if (j == 0 || clean_name[j - 1] == ' ')
+	    continue;
+	}
+
+      if (whitespace_p (name[i]))
+	clean_name[j] = ' ';
+      else
+	clean_name[j] = name[i];
+
+      j++;
+    }
+  if (j > 0 && clean_name[j - 1] == ' ')
+    j--;
+  clean_name[j] = '\0';
 
   for (p = ld->arch_info[la->la_language].primitive_type_vector;
        (*p) != NULL;
        p++)
     {
-      if (strcmp (TYPE_NAME (*p), name) == 0)
-	return (*p);
+      if (strcmp (TYPE_NAME (*p), clean_name) == 0)
+	{
+	  xfree (clean_name);
+	  return (*p);
+	}
     }
+  xfree (clean_name);
+
   return (NULL);
 }
 
diff --git a/gdb/testsuite/gdb.python/py-type.exp b/gdb/testsuite/gdb.python/py-type.exp
index 6b61f48..c9388ed 100644
--- a/gdb/testsuite/gdb.python/py-type.exp
+++ b/gdb/testsuite/gdb.python/py-type.exp
@@ -264,3 +264,15 @@  with_test_prefix "lang_cpp" {
     test_template
     test_enums
 }
+
+# Tests to lookup builtin types
+gdb_test "python print gdb.lookup_type ('unsigned int')" "unsigned int" \
+    "lookup unsigned int"
+gdb_test "python print gdb.lookup_type (' unsigned long ')" "unsigned long" \
+    "lookup unsigned long"
+gdb_test "python print gdb.lookup_type (' unsigned  char ')" "unsigned char" \
+    "lookup unsigned char"
+gdb_test "python print gdb.lookup_type (' unsigned\\n char ')" "unsigned char" \
+    "lookup unsigned char"
+gdb_test "python print gdb.lookup_type (' unsigned\\tlong ')" "unsigned long" \
+    "lookup unsigned long"