From: Sam Roberts <sroberts@...>
Date: 2006-05-04T06:08:27+09:00
Subject: Re: parsedate doesn't work on Time.now.to_s

On Thu, May 04, 2006 at 04:37:56AM +0900, Austin Ziegler wrote:
> On 5/3/06, Berger, Daniel <Daniel.Berger@qwest.com> wrote:
> >Take a look at time_to_s in time.c and you'll see this line:
> >
> >len = strftime(buf, 128, "%a %b %d %H:%M:%S %Z %Y", &tobj->tm);
> >
> >The 'issue' is %Z.  Reading the man page on Solaris is merely says "time
> >zone name or abbreviation".  How it determines which to use, I'm not
> >sure.  On Windows it's apparently controlled by a registry setting
> >somewhere (Windows also supports %z).  So, my guess is most *nix systems
> >default to the abbreviation, while Windows defaults to the long name.
> >
> >I don't think we need to fix time_to_s, however.  We need to fix
> >ParseDate.
> 
> >>Time.now.strftime("%a %b %d %H:%M:%S %z %Y")
> => "Wed May 03 15:37:18 Eastern Daylight Time 2006"
> 
> Note the use of "%z" as opposed to "%Z". Windows is hosed.

The 1.8.4 source doesn't document %z as an allowed conversion. The linux
man page describes it as a GNU extension:

 %z     The time-zone as hour offset from GMT.  Required to emit RFC822-conformant dates (using "%a, %d %b %Y %H:%M:%S %z"). (GNU)

So while I think Windows is hosed as a general principle, I think its
within its rights in this case.

The ruby code isn't validating the string input, just passing it to the
strftime(), if ruby is supposed to support %z portably, it will have to
do it itself. Its kindof useful to be able to use the local extensions,
of course:

  ruby -e 'p Time.now.strftime("%P %r")'
  "pm 01:59:33 PM"

even if not very portable. Kind of a pain to have to reimplement library
functions, but when they give access to non-standard behaviour, what is
the right approach?

Cheers,
Sam