[Quilt-dev] Re: remove GNU date dependency [Was: quilt mail broken with sed 3.02]

[Andreas Gruenbacher]

On Wednesday 01 February 2006 11:44, Gary V. Vaughan wrote:
> Andreas Gruenbacher wrote:
> > Gary,
>
> Morgen!
>
> > I've looked into this a little, and standard date(1) doesn't have the -d
> > option, and doesn't support `%s' and `%z'. We could get of the first use
> > of -d, but incrementing timestamps is harder, and I'd hate to have to
> > sleep a second between each date generated just to get a different
> > timestamp. We could put the same timestamp in each message in case we
> > can't increment, but then, we still can't generate `%z'. The Perl
> > workaround also relies on strftime(3), which probably is the platform's
> > native libc implementation, and so probably has the same limitations as
> > date(1).
>
> Apparently, perl doesn't use the native strftime... or at least on darwin
> %z works within perl, but isn't used in Apple's date implementation.

Okay, that's good for us.

> Here is a patch for a perl date wrapper (my first perl script!), I hope you
> can commit this in time for 0.43...  with it, I can `./configure
> --without-date' and have the testsuite complete successfully.

Looks very nice. I have only a few comments:

 * You ignore the result of GetOptions, which is only minor.
 * I don't see why we would need the pass_through option;
   we still fail further below, anyway.
 * parse_utc_secs() shouldn't affect the --utc option. Also, the seconds
   since the epoch are always relative to UTC.
 * In scalar contect, @ARGV evaluates to $#ARGV + 1.
 * When you use sub foo() { ... } instead of sub foo { }, the
   parentheses declare "no parameters". It should be sub
   parse_utc_secs($) { ... } instead.
 * "use" inside a subroutine??

Can you maybe give the checked-in version a quick check?

Thanks,
Andreas.