Use CPP?

[Mesabloo]

CPP (the C PreProcessor) is used to transform some code into some other depending on the preprocessor directives in the processed code. It is possible, for example, to define macros, or do conditional compilation.

Macros (#38) and targets (#37) are some of the features that CPP can cover. It also is more general, allowing for #erroring out even before parsing the source code, #includeing source files into others (using include guards with #ifndef and #define).

While this should work, and provides a handful lot of features, there are several drawbacks:

• how does it change the source code? can we still get the initial positions of tokens out of it?
• we would have to define so much variables each time (for each architecture, to allow for target code, but also for some features like DEBUG compilation, etc)
• this could be a real pain to integrate cpp into the compiler
• the compiler would then heavily depend on the users' environments, possibly erroring out for reasons that are actually not coming from NSC
• this preprocessor in particular is very verbose, but this is not such a problem at this point

We could consider using it, but we need to evaluate the cost of it first, checking whether we can preserve positions, so that the user does not get trash errors. Creating our own macro system may also yield better errors, for example allowing to report errors like "at line N: blablabla, coming from expansion of macro at line M: blablabla".

[Mesabloo]

Because of the way CPP works, it complicates the grammar quite a bit. Basically, we would have to skip all lines starting with # (if we were doing a tool CPP-alike), while still registering them (keeping the starting and ending positions, etc) and post-process them after parsing. If we were to use CPP, we would most likely lose some of the positions in the source code, which is a bit worrying.

The target and macro alternatives (#37 and #38) are much easier to parse, and provide the same basic set of functionalities when combined with #40.

And because of all the problems that may arise from the user's environment (e.g. invalid or broken environment), I don't think it is worth tackling this problem.