Allegro `const'-correctness




This is a short document about the introduction of `const'-correctness to Allegro. It details what changes have occurred to internal library code, the API changes (mainly transparent) and what you will need to do to adapt code to compile without warnings (again, mainly nothing).



Contents




Library changes

There are very few actual changes to the library code itself; only some symbol declarations and definitions have been altered to include AL_CONST. See below for a description of the AL_CONST preprocessor define. In a few places, some string was changed that should not have been - in these cases, the string is simply duplicated and then the duplicate is erased on exiting the function.

In all, there were very few changes to the library code.



The AL_CONST preprocessor define

In order to support compilers which don't know about the `const' keyword, or perhaps use a different keyword, the preprocessor symbol AL_CONST is used wherever `const' would normally be used. Note that in the documentation, I have used `const' for readability.



Allegro API changes

These are, generally speaking, totally transparent to the user. I did not change the behaviour of any function; only its parameter types. Basically, if you can pass it as type* ptr, then you can pass it as const type* ptr without any problem whatsoever. Note also that certain changes may remove warnings in your program as static strings, etc, are now treated as `const' by Allegro functions.

There are a few places, described below, where there will be an effect on existing code.

`const'-correctness is deemed important for two reasons. Firstly, it can increase code readability and comprehension of Allegro functions (for instance, you can see which parameters are altered and which are not). Secondly, it ensures that the Allegro code is not changing data which it should not be, and that client callback functions are not breaking Allegro by changing data they should not be.



Callback functions and Pointers to Pointers

Certain callback functions now have a different type - they take `const' pointers as opposed to non-`const' pointers. As far as I know, a compiler will issue a warning about incompatible pointer types. You should update your callback function to the new format (which will be listed in the main Allegro documentation).

Also, when passing a pointer to a pointer to an Allegro function which is declared as taking an AL_CONST type** ptr, you will need to cast your pointer to be `const' if it is not already. For instance:

int some_allegro_function(AL_CONST char** p);

void my_func(char** x) { some_allegro_function((AL_CONST char**) x); }

I realise that this is a change to the Allegro API, and that we are supposed to avoid those at all costs, but this is essentially fixing a bug in Allegro and changing behaviour. It also ensures that client-supplied callback functions are functioning correctly, and not altering data that they should not. Callback functions which do not treat relevant parameters as `const' are, in a small (but potentially signficant) way, broken.

Please note that for the Unicode function ugetx(), I have provided an alternative version ugetxc(), which takes a `const char**' parameter as opposed to a `char**' parameter. This is because it is valid to pass either a `char**' or a `const char**', but unfortunately there is no way to tell the compiler exactly what we mean.



BITMAP objects

Allegro represents both a screen bitmap and a memory bitmap by a single object; a BITMAP. Unfortunately, these two things can be very different. For instance, reading a pixel from a bitmap would not seem to change it, but if it is a screen bitmap we are reading from, then it is possible that some parameter of the video card is changed to select the correct line, etc.

Therefore, a const BITMAP parameter does not make sense, and is not used throughout the library. This is unfortunate, but I cannot see any way around it.



Finally...

I have only tested this myself on DJGPP 2.95.2, so if you have experiences of it not working (I have tried to get Linux and BeOS ports right, but I've no guarantee it worked...) email me the *exact* error message and I will fix it. You can also contact the Allegro conductors list; see `How to get help' in the Allegro documentation.

Email: lwithers@lwithers.demon.co.uk

Thanks for listening :-)