[ruby-core:32908] Re: WIN32OLE VT_DATE handling bug

On Oct 26, 2010, at 10:10 AM, Chuck Remes wrote:

> I have found a problem with some date handling via the WIN32OLE API. Unfortunately I don't know how to provide a reproducible case because the COM program is proprietary, so I'm looking for guidance on how to file this bug.
> 
> Description
> The problem is exposed at the turnover from 23:59:00 to the start/midnight of the next day. First the code, then the output:
> 
>  # requested objects for a time range "2010/10/25 23:59" to "2010/10/26 00:01"
>  objects.each do |obj|
>    puts "#{obj.timestamp}, #{obj.timestamp.class}"
>  end
> 
> MRI 1.9.1-p430 (via RubyInstaller) produces:
> 
> 2010/10/25 23:59:00, String
> 2010/10/25 00:00:00, String
> 2010/10/26 00:01:00, String
> 
> 
> Under JRuby 1.5.1 using the very same code and the new jruby-win32ole support, it produces the correct output.
> 
> Mon Oct 25 23:59:00 -0500 2010, Time
> Tue Oct 26 00:00:00 -0500 2010, Time
> Wed Oct 26 00:01:00 -0500 2010, Time
> 
> 
> It looks to me like the MRI WIN32OLE code is mishandling the COM VT_DATE time conversion.
> 
> In the absence of a reproducible case, how should I write this bug report?

I apologize for the noise on this list. I installed MRI 1.9.2p0 and reran the test. This bug was apparently fixed in the 1.9.2 branch. It also returns Time objects now instead of Strings.

cr