rust-lang
Allow modifying macro tokens at parse time
This pull requests adds a callback to ParseCallbacks that enables users to mangle the tokens on a macro definition. The goal is to allow custom "sanitization" of macros to enable generation of rust code.
In my use case, I have a few header files that define (a few hundred) macros such as:
#define FLAG_BITS 8:10
#define FLAG_FOO 0x1
#define FLAG_BAR 0x2
#define FLAG_FOOBAR 0x4
Currently, bindgen, rightly, ignores FLAG_BITS since the meaning of that definition is application specific. In my case, the side effect is that I lose the information about how much to shift bits around.
This pull request allows programmers to define a modify_macro callback that enables inspecting the Tokens of a given macro definition and sanitize it for bindgen to be able to generate rust code for it.
In my use case, the callback translates the three tokens 8 : 10 into (8 << 16) | 10. A value I can use in my rust code to determine the location of the bits.
Err, sorry, should've read the PR message more carefully. Rather than modifying the tokens, would it be reasonable to pass the tokens to will_parse_macro (without modifying them)?
Passing tokens to will_parse_macro could help to identify problematic macros, but I believe that no other callback for that macro will be ever executed since the call to clang parse will fail (the macro is neither int_macro nor str_macro). Without modifying the tokens to make clang happy I fail to see how this will work. What do you have in mind?
:umbrella: The latest upstream changes (presumably 71fe5fdcb657470b7c2e1b841eaea4660d4ccc40) made this pull request unmergeable. Please resolve the merge conflicts.
