On Sat, 5 Jun 2021, Jason A. Donenfeld wrote:
> Thanks! That seems to be it indeed. Interestingly, including the
> actual path to the real headers with -I didn't fix the issue.
Most of the time, Coccinelle doesn't use macro definitions. The -I could
have been useful for figuring out some types, but probably wasn't needed
in this case.
> I tried
> adding other options passed as -D to the compiler in a macro file, in
> case that was influencing the parsing, to no avail. So in the end I
> indeed went with just defining those to int. Seems to work well! Last
> minor cosmetic issue is that it turns `struct a
> [space][space][space][space][space][space][space] x` into `struct b
> [space] x`, which is easy enough to fix up by hand.
That's beyond the degree of cleverness we try t oprovide. but in this
particular case, it would probably not have been needed if the rule for
changing the type name had only changed the name, and not the whole member
> (The BSDs do weird
> and terrible things with whitespace sometimes.)
Cocci mailing list