time: bug: must subtract gmtoff, not add.

This works fine:

  1> (time-struct-utc 0)
  #S(time year 1970 month 1 day 1 hour 0 min 0 sec 0 wday 4 yday 0
          dst nil gmtoff 0 zone "GMT")
  2> *1.(time-utc)
  0
  3> *1.(time-local)
  28800

But we want the following to return the same results:

  1> (time-struct-local 0)
  #S(time year 1969 month 12 day 31 hour 16 min 0 sec 0 wday 3 yday 364
          dst nil gmtoff -28800 zone "PST")
  2> *1.(time-utc)
  -57600
  3> *1.(time-local)
  -28800

With the patch, we do:

  1> (time-struct-local 0)
  #S(time year 1969 month 12 day 31 hour 16 min 0 sec 0 wday 3 yday 364
          dst nil gmtoff -28800 zone "PST")
  2> *1.(time-utc)
  0
  3> *1.(time-local)
  28800

This is also broken:

  1> (time-parse-utc "%H:%M:%z" "00:00:-0800")
  -28800

It must return 28800.

* time.c (time_meth): This function, which is the
imlpementation of the time-utc and time-local methods, must
subtract the gmtoff field, not add it. This is so that
a UTC time expressed in a local time zone will convert
back to the correct UTC epoch.
(time_parse_local, time_parse_utc): Here we likewise
must subtract the gmtoff.

0 files changed, 0 insertions, 0 deletions