> Is the kernel using this sort of non-integer-constant-expression in > bit-field widths as well? Sizes of file-scope arrays (OK, a special case > for that was added for WINE)? Null pointer constants (that's much more > problematic to add special cases for)? __builtin_choose_expr conditions? > It seems to be making rather a lot of use of an old obscure undocumented > extension.
It is possible that the kernel uses this sort of expression in other places. I can not say for sure. At the moment I found only these two.
