Bug in C++ template parsing

[flanggut]

Bug in C++ template parsing

[flanggut]

Looks great! Thanks 👍

[flanggut]

As an example for this header file:

#pragma once

template <typename D> class Foo {
public:
  float bar(float a) { return a; }
};

And with the cursor on bar, I am getting documentation like this when I call generate:

#pragma once

/**
 * @brief 
 *
 * @tparam D 
 * @param a 
 */
template <typename D> class Foo {
public:
  /**
   * @brief 
   *
   * @param a 
   * @return 
   */
  float bar(float a) { return a; }
};

As a side note: I think the root node might not be considered at all, so I had to create a different root with #pragma once.

[danymat]

I don't think I understand correctly... In order for me to understand, can you do a snippet of what the current generation is doing, and what the annotation should look like ?

[flanggut]

I already did. If you take the snippet from my previous comment

#pragma once

template <typename D> class Foo {
public:
  float bar(float a) { return a; }
};

and you generate annotations for the function bar you get two annotations, one for the function and a wrong one for the class Foo.

[flanggut]

Specifically this is what I get

#pragma once

/**
 * @brief 
 *
 * @tparam D 
 * @param a 
 */
template <typename D> class Foo {
public:
  /**
   * @brief 
   *
   * @param a 
   * @return 
   */
  float bar(float a) { return a; }
};

And this is what I would expect

#pragma once

template <typename D> class Foo {
public:
  /**
   * @brief 
   *
   * @param a 
   * @return 
   */
  float bar(float a) { return a; }
};

[danymat]

Oh yeah right, it's a bug indeed

[danymat]

Hello once again, this should fix the bug that caused those multiple annotations. Can you check again ?

PS: this fix will make Neogen potentially faster than previous versions, because it only parses one tree and execute generation only on this... 🥂