I work on a codebase that uses a strange dialect of C++ --- it includes many new
"keywords", annotations, and so on that are #defined away when we compile so
that we produce working code. cc-mode doesn't benefit from these saving
preprocessor macros, however, and becomes confused by these annotations.

e.g.,

_Success_(return == TRUE)
BOOL
Foo(
    _In_ DWORD Blah
)
{
    ...
}

Until now, I've hacked up my copy of cc-mode to understand our syntax, but this
approach isn't sustainable because I inevitably fall behind cc-mode trunk. I
have an idea for an alternative that might be generally useful instead:

1. We can define a text property, cc-ignore, that cc-mode will look for when
it's about to read a symbol from the buffer. When it finds this property, it'll
skip over all characters until the text property disappears again. Then, a minor
mode can make itself responsible for identifying all non-standard constructs in
a cc-mode buffer and mark them with cc-ignore so that cc-mode's parsing engine
handles the syntactically-valid remainder of the buffer.

2. Create a cc-mode synonym table. My team (much to my chagrin) insists on
#define struct interface, then using interface IFoo { ... }. Because "interface"
behaves identically to "struct", I just tried it as "struct" in my cc-mode. I
wonder whether this mechanism can be made more general.

With these changes, I think cc-mode could be made to work for a greater variety
of C dialects. Any thoughts?