BUG: Auto-Completion with namespace

[darksquall57]

Hi,

I just found a bug that makes the auto-completion not working when using a namespace.
I'm using this base code :

Code Select Expand


namespace nBase {

class A
{
public:
    int ObjectA() const { return mObjectA; }
public:
    int mObjectA;
};

class B :
    public A
{
public:
    int mObjectB;
};

}

int main()
{
    return 0;
}


Here is what happens:

Code Select Expand


int main()
{
//----
    ::nBase:: //Autocompletion ok
//----
    ::nBase::B b;
    b. //Autocompletion ok
//----
    ::nBase::B:: //Here AutoCompletion doesn't popup
    return 0;
}


If I remove the namespace nBase, then everything works fine.

I'm running Codeblocks rev9594 64bits on Ubuntu 12.04, but I've noticed this bug a long time ago. It's only now that I've found what's really happening.

[ollydbg]

Hi, darksquall57, thanks for the report.
I can confirm this bug, I did a git blame, and found this bug was introduced in svn rev

Code Select Expand

Revision: 0da216bc5e6471b760be10b3e47b9445a981d37d
Author: mortenmacfly <mortenmacfly@2a5c6006-c6dd-42ca-98ab-0921f2732cef>
Date: 2010-8-16 12:42:07
Message:
* cc_branch: applied patch to fix global scope search failed

git-svn-id: http://svn.code.sf.net/p/codeblocks/code/branches/codecompletion_refactoring@6485 2a5c6006-c6dd-42ca-98ab-0921f2732cef
----
Modified: src/plugins/codecompletion/codecompletion.cpp
Modified: src/plugins/codecompletion/nativeparser.cpp

This line was added:

Code Select Expand


    else if (lineFirstChar == _T(':') && curChar == _T(':'))
        return;

in the function: void CodeCompletion::DoCodeComplete().

In you case:

Code Select Expand

::nBase::B::
It matches this condition, so this function just returned......

[MortenMacFly]

I can confirm this bug, I did a git blame, and found this bug was introduced in svn rev

Well but you also see why: Because otherwise global scope will not work. So be careful and do not just revert this change.

[darksquall57]

Thanks for your fast replies .

[ollydbg]

I can confirm this bug, I did a git blame, and found this bug was introduced in svn rev

Well but you also see why: Because otherwise global scope will not work. So be careful and do not just revert this change.

This is the change of that commit:

Code Select Expand

0da216bc5e6471b760be10b3e47b9445a981d37d
src/plugins/codecompletion/codecompletion.cpp | 9 +++++++--
src/plugins/codecompletion/nativeparser.cpp   | 4 ++--
2 files changed, 9 insertions(+), 4 deletions(-)

diff --git a/src/plugins/codecompletion/codecompletion.cpp b/src/plugins/codecompletion/codecompletion.cpp
index 2f41040..9709d6e 100644
--- a/src/plugins/codecompletion/codecompletion.cpp
+++ b/src/plugins/codecompletion/codecompletion.cpp
@@ -1397,7 +1397,10 @@ void CodeCompletion::DoCodeComplete()
    const int style = control->GetStyleAt(pos);
    const int lineIndentPos = control->GetLineIndentPosition(control->GetCurrentLine());

-    if (ed->GetControl()->GetCharAt(lineIndentPos) == _T('#'))
+    const wxChar lineFirstChar = ed->GetControl()->GetCharAt(lineIndentPos);
+    const wxChar curChar = ed->GetControl()->GetCharAt(pos - 1);
+
+    if (lineFirstChar == _T('#'))
    {
        const int start = control->WordStartPosition(lineIndentPos + 1, true);
        const int end = control->WordEndPosition(lineIndentPos + 1, true);
@@ -1409,7 +1412,9 @@ void CodeCompletion::DoCodeComplete()
            CodeCompletePreprocessor();
        return;
    }
-    else if (ed->GetControl()->GetCharAt(pos - 1) == _T('#'))
+    else if (curChar == _T('#'))
+        return;
+    else if (lineFirstChar == _T(':') && curChar == _T(':'))
        return;

    if (style != wxSCI_C_DEFAULT && style != wxSCI_C_OPERATOR && style != wxSCI_C_IDENTIFIER)
diff --git a/src/plugins/codecompletion/nativeparser.cpp b/src/plugins/codecompletion/nativeparser.cpp
index cc865b7..3a4c8b7 100644
--- a/src/plugins/codecompletion/nativeparser.cpp
+++ b/src/plugins/codecompletion/nativeparser.cpp
@@ -1917,8 +1917,8 @@ static bool IsOperatorEnd(int startAt, const wxString& line)
}
static bool IsOperatorBegin(int startAt, const wxString& line)
{
-    return (   (startAt > 0)
-            && ((size_t)startAt + 1< line.Len())
+    return (   (startAt >= 0)
+            && ((size_t)startAt < line.Len())
            && (   (   (line.GetChar(startAt ) == '-')
                    && (line.GetChar(startAt + 1) == '>') )
                || (   (line.GetChar(startAt) == ':')


I take some time to read the code and debug some examples, I don't see what exact bug this commit going to fix.

Here is the code I use:

Code Select Expand


void f1();
void f2();


namespace nBase {

   class A
   {
   public:
       int ObjectA() const { return mObjectA; }
   public:
       int mObjectA;
   };

   class B : public A
   {
   public:
       int mObjectB;
   };

}

int main()
{

   ::f                           // works fine here
   return 0;
}


I don't even understand what does "applied patch to fix global scope search failed" means.

What is the relation ship between the first char and the current char?

Code Select Expand


+    else if (lineFirstChar == _T(':') && curChar == _T(':'))
        return;

???

Maybe, that patch was created by Loaden? but I think Loaden is not active for a very long time.

[ollydbg]

Another change I think is wrong:

Code Select Expand


-- a/src/plugins/codecompletion/nativeparser.cpp
+++ b/src/plugins/codecompletion/nativeparser.cpp
@@ -1917,8 +1917,8 @@ static bool IsOperatorEnd(int startAt, const wxString& line)
}
static bool IsOperatorBegin(int startAt, const wxString& line)
{
-    return (   (startAt > 0)
-            && ((size_t)startAt + 1< line.Len())
+    return (   (startAt >= 0)
+            && ((size_t)startAt < line.Len())
             && (   (   (line.GetChar(startAt ) == '-')
                     && (line.GetChar(startAt + 1) == '>') )
                 || (   (line.GetChar(startAt) == ':')

startAt could be 0, that's Ok, but we later use line.GetChar(startAt + 1), so it should be the condition (size_t)startAt + 1< line.Len().

[ollydbg]

Code Select Expand

int main()
{

    ::                          // cc here
    return 0;
}


If cc after the double colon, then the code goes here:

Code Select Expand


    else if (lineFirstChar == _T(':') && curChar == _T(':'))
        return;

So, I guess this "else if" condition is going to skip this condition?
If true, I think the logic is not correct, the correct way should be:

Code Select Expand


    else if (lineFirstChar == _T(':') && curChar == _T(':') && "there is no extra char between linFirstChar and curChar")
        return;

What's your opinion?