Looking for comments on changing strptime(3) %p behavior

[Brian Ginsbach <ginsbach%netbsd.org@localhost>]

The current behavior of strptime(3) and the %p format specifier is
a bit inconsistent/counter-intuitive/surprising.  The results aren't
always what one would expect.  Nor what is seen from other strptime(3)
implementations.  Looking for feedback on the following.

a. strptime(str, "%H %p", &tm)

   Currently if str contains 0-11, for the hour, and a am/pm string
   the time will be adjusted accordingly.  However if the str
   contains 12-23, for the hour, and a am/pm string strptime(3)
   returns NULL (error).  This seems inconsistent if not a bug.

   Does it make sense to apply %p to any hour parsed by %H (or %k)
   as %H implies a 24-hour clock value?  The current code is
   inconsistent; %p is valid with 0-11 while %p is invalid with
   12-23 (strptime(3) returns NULL).  The tm_hour is only adjusted
   when %pm is PM and < 11.

   Would a  more consistent approach would be to not apply %p to
   any 24-hour clock value?  It doesn't necessarily make sense that
   %H (24-hour clock) with %p should be an error regardless of the
   value parsed.  The %p would be parsed but ignored when combined
   with a 24-hour clock.

b. strptime(str, "%p", &tm)

   Currently if str contains a am/pm string and no hour is specified
   then tm_hour is adjusted as per above or an error is returned
   based on tm_hour.  This would seem to be a bug on both counts.

   The %p should be parsed and only applied to a valid 12-hour
   clock value obtained via the same parse.

   Additionally what should happen if no hour is parsed, either a
   12-hour or 24-hour clock?  Should the tm_hour be set to a default
   AM (0) or PM (12) or just left alone?

c. strptime("12", "%I", &tm)

   Currently the 12 is converted tm_hour = 0.  This seems rather
   odd since %I must be in the range of 1-12.  If this isn't a bug
   then clearly it violates POLA having 12 == 0.  It makes more
   sense, in the absence of %p, to have 12 == 12.

Changes are worked up for all of these.  Just waiting on feedback
about some of the above questions before committing.

Additionally, the changes will also allow for parsing (and applying)
%p before %I (or %H if the current behavior is kept).  Currently
a %p must be specified after the hour for it to be applied.

Thanks,
Brian