> It looks like these should be time_t.  As it happens, time_t is long,
> long is 32 bits, and these are {struct timespec.a_sec} or whatever, which
> is int32, but this gives compiler warnings when used in a ptr-to expression.
> 
> Shouldn't we have the types of these expressions come out as time_t, so
> code like ctime(&foo.st_mtime) works?

... actually...

in my opinion (and the opinion of several other well known kernel
hackers), many of the "explicit types" changes, while good as a first
pass, need to be rethought a bit.

In particular, the distinction between "on disk" and "in core"
structures is not as clear as it perhaps should be.



chris