Linus Torvalds writes: (Summary)
Maybe "make __UNIQUE_ID easier to use and encourage that model" is the
right answer.
right answer.
Right now "__UNIQUE_ID" is actually really nasty in several ways: Right now "__UNIQUE_ID" is actually really nasty in several ways: (1) the already mentioned "the fallback is broken for same-line use" (1) the already mentioned "the fallback is broken for same-line use" This doesn't really matter because both gcc and clang have _true_ unique macros, but we should probably remove the fallback as "know broken and not really guaranteed to give a unique ID" broken and not really guaranteed to give a unique ID" (2) The argument you give to __UNIQUE_ID() is pointless.
right answer.
Right now "__UNIQUE_ID" is actually really nasty in several ways: Right now "__UNIQUE_ID" is actually really nasty in several ways: (1) the already mentioned "the fallback is broken for same-line use" (1) the already mentioned "the fallback is broken for same-line use" This doesn't really matter because both gcc and clang have _true_ unique macros, but we should probably remove the fallback as "know broken and not really guaranteed to give a unique ID" broken and not really guaranteed to give a unique ID" (2) The argument you give to __UNIQUE_ID() is pointless.