4088 - Date parsing gets 12:30 AM wrong.

Status: VERIFIED FIXED

(Core :: JavaScript Engine, defect, P3)

Description

[Mike McCabe]

js> d = new Date('1/1/1999 12:30 AM')
Fri Jan 01 12:30:00 PST 1999
js>  d = new Date('1/1/1999 12:30 PM')
Sat Jan 02 00:30:00 PST 1999

... This also occurs in our Java implementation.  Looks like I copied the
parsing code a little too faithfully.

Comment 1

[Mike McCabe]

Checked in a fix to the tip, SpiderMonkeyDev_BRANCH, SpiderMonkey140_BRANCH, and
the java implementation.  The fix checks for 12:XX AM (subtract 12 for this
case, otherwise do nothing for AM) and checks for 12:XX PM (subtract 12 for any
hour except 12; leave 12:XX PM alone.)

This fix also applies the hour <= 12 check to the AM case that previously only
applied to the PM case (because no time adjustment was done for AM.)  This means
that the date string '1/1/1999 13:30 AM' no longer produces a valid date.  I'm
guessing that this will be ok, because '1/1/1999 13:30 PM' never did.

Old behavior:
js> d = new Date('1/1/1999 13:30 AM')
Fri Jan 01 13:30:00 GMT-0800 (PST) 1999
js> d = new Date('1/1/1999 13:30 PM')
Invalid Date
js> d = new Date('1/1/1999 12:30 AM')
Fri Jan 01 12:30:00 GMT-0800 (PST) 1999
js> d = new Date('1/1/1999 12:30 PM')
Sat Jan 02 00:30:00 GMT-0800 (PST) 1999

New behavior:
js> d = new Date('1/1/1999 13:30 AM')
Invalid Date
js> d = new Date('1/1/1999 13:30 PM')
Invalid Date
js> d = new Date('1/1/1999 12:30 AM')
Fri Jan 01 00:30:00 GMT-0800 (PST) 1999
js> d = new Date('1/1/1999 12:30 PM')
Fri Jan 01 12:30:00 GMT-0800 (PST) 1999

Comment 2

[Mike McCabe]

Checked in a fix to the tip, SpiderMonkeyDev_BRANCH, SpiderMonkey140_BRANCH, and
the java implementation.  The fix checks for 12:XX AM (subtract 12 for this
case, otherwise do nothing for AM) and checks for 12:XX PM (subtract 12 for any
hour except 12; leave 12:XX PM alone.)

This fix also applies the hour <= 12 check to the AM case that previously only
applied to the PM case (because no time adjustment was done for AM.)  This means
that the date string '1/1/1999 13:30 AM' no longer produces a valid date.  I'm
guessing that this will be ok, because '1/1/1999 13:30 PM' never did.

Old behavior:
js> d = new Date('1/1/1999 13:30 AM')
Fri Jan 01 13:30:00 GMT-0800 (PST) 1999
js> d = new Date('1/1/1999 13:30 PM')
Invalid Date
js> d = new Date('1/1/1999 12:30 AM')
Fri Jan 01 12:30:00 GMT-0800 (PST) 1999
js> d = new Date('1/1/1999 12:30 PM')
Sat Jan 02 00:30:00 GMT-0800 (PST) 1999

New behavior:
js> d = new Date('1/1/1999 13:30 AM')
Invalid Date
js> d = new Date('1/1/1999 13:30 PM')
Invalid Date
js> d = new Date('1/1/1999 12:30 AM')
Fri Jan 01 00:30:00 GMT-0800 (PST) 1999
js> d = new Date('1/1/1999 12:30 PM')
Fri Jan 01 12:30:00 GMT-0800 (PST) 1999

Comment 3

[cbegle]

so are you going to write some test cases for Date.parse.  :)

Comment 4

[leger]

Changing component to "Javascript Engine".  "Javascript" component is being
retired.

Comment 5

[cbegle]

checked in a regression test:  mozilla/js/tests/ecma/Date/15.9.4.2-1.js