OK, so I decided to try to learn some XML related stuff and figured the best place to start would be actually working with documents in a real programming language. IIRC, I actually initially searched for XPath and tinyxpath caught my attention so I figured it was a good place to start. When I saw that it made use of tinyxml I decided to start with tinyxml instead.

I started by downloading tinyxml, tinyxpath, and tinyxml++ source to the computer sciences server at the college and compiled all three. I read through some tinyxml documentation and started following this tutorial.

It's not even really so much a tutorial as it is example code in unstructured order. Anyway, it's enough to get the point across. However, it doesn't appear that the author of the tutorial (at least as far as I am in the tutorial) is concerning himself with memory management. For example, we're instantiating numerous objects and attaching them to stack objects, but we never destroy them. In particular, we're creating TiXmlComments, TiXmlDeclarations, TiXmlElements, and TiXmlTexts; and generally LinkEndChild'ing them to a TiXmlDocument on the stack (or to other heap objects, like TiXmlElements).

I'm wondering if anybody knows whether the TiXmlDocument is going to manage these children or whether I'm supposed to destroy them after (which seems more likely). I did a little reading through the documentation and checked the source and didn't see anything to suggest that the TiXmlDocument would be managing the memory of children so I wrote a simple macro to check a pointer, delete it, and set it to 0 and started implementing it. However, for some reason one of my pointers is causing a SEGFAULT in (IIRC) build_simple_doc on line 101.
DELETE(text);
Does anybody see what might be causing this? I ran it through gdb and printed the value of text and it has a non-zero value so I assume it was allocated successfully...

** EDIT **

I added a line just before the 3 DELETE macros where the SEGFAULT occurs and accessed a member of text and it seemingly worked fine:
printf("text->CDATA() = %d\n", text->CDATA());

Ah, I see... I was looking at tinyxml.cpp from the Online Documentation and all of the destructors seem to be empty... Maybe that documentation is out of date. So I guess I don't have to worry about it while using a static parent... Thanks, ReyBrujo.

delete and free will already no-op when given NULL. You don't need to check them. And when doing a block command with a macro, you should use do{}while(0) like:

#define MYMACRO() do { \
  ..stuff.. \ 
} while(0) 
...
MYMACRO();

..so you won't have any nasty side effects with the semicolon or if/for/while commands.

An if statement is more costly than an int/pointer assignment. As long as you always set a pointer to 0/NULL after you free or delete it, you never have to check before you free or delete it.

http://www.noveltheory.com/TechPapers/while.htm

I suppose that makes sense...

Yes, well, as long as you free memory resources after you're done with them there will be no need for garbage collectors, but modern language designers still thought they were useful. I see what you're saying though. Thank you.

The problem with an inline function is that I would need to pass a pointer to my pointer to assign it a NULL value, which is less pretty (because of the address-of operator) and more error prone. A macro avoids those subtleties.

Ah, I see... As you might have guessed, I don't use macros very often.

OMG! I never thought of that! Thanks!!

** EDIT **

I'm having problems calling it though. They might be to do with my project structure or Makefile, but I'm not sure.

I'm getting a linker error:

undefined reference to `void bam_delete<TiXmlElement>(TiXmlElement*&)'

Am I calling it wrongly?

I don't formally know how to use templates... I've gotten by a little bit, but don't actually know the rules so I could easily be doing something wrong... According to this, template function specialization is used to handle specific cases that otherwise wouldn't work (i.e. a max function using the > operator won't work on char * without a specialized implementation).

I get the same result. I don't think that matters.

That's the ugly part... I'm using a Makefile and I don't really know what I'm doing with them... I'll just .tar.gz the whole project directory (AFAIK, you can freely redistribute tinyxml without modifying the license information in the header)... It's a small project so don't let that scare you. Let me know if you're unable to open .tar.gz, although 7-Zip for Windows can do it and is better than any other software I've used in Windows.

tinyxml-tutorial.tar.gz

The implementation is in include/bam and src/bam and everything is compiled with a very manual Makefile because I haven't figured out Makefiles yet. Don't ask me what is or isn't compatible with Windows.

src/bam/bam.hpp:

#ifndef BAM_HPP
  #define BAM_HPP

        /* bam/bam.h (and bam/bam.c) is the macro implementation... */
  #include <bam/bam.h>

template<typename T>
inline void bam_delete(T *&);

template<typename T>
inline void bam_free(T *&);

#endif

src/bam/bam.cpp:

src/main.cpp:

#include <bam/bam.hpp>

.
.
.

bam_delete(window);

Makefile:

** EDIT **

Unless the #include <bam/bam.hpp> should be using double-quotes (""), even though I added a -Iinclude option to g++.

I moved the template function definitions into include/bam/bam.hpp and it still doesn't work... Same linking error.

It may not matter, but did you comment out the declarations, and just leave the definitions? You could try using template <class T> instead of template <typename T>. Don't forget -Wall in your makefile's CPPFLAGS.

Edit
It could be old .o files causing it - try cleaning them out and rebuilding it.

If your makefile doesn't consider headers (for example, file.o : file.cpp file.h as rule) it won't care about the headers.