[bug-gnulib] Re: same, utimecmp should depend on minmax

[Bruno Haible]

Paul Eggert wrote:
> More and more this is starting to sound like a loser situation.
> Perhaps we should give up on these predecessor #includes entirely, and
> warn users that they must include minmax.h after they include any file
> that might possibly include a system file.

I don't wish to go back to the days where we had to think about order of
#includes. Like "we need <sys/types.h> before <regex.h>" and similar.

The following heuristics works very well in practice:
  - First comes the #include "..." specifying the module being implemented.
  - Then come all the #include <...> of system or system-replacement headers,
    in arbitrary order.
  - Then come all the #include "..." of gnulib and private headers, in
    arbitrary order.

This little problem with minmax.h is now fixed, and not enough of a problem
to depart from this practice.

Bruno