[erlang-questions] File creation times under Windows
Tue Oct 28 10:55:04 CET 2008
Dave Smith wrote:
> NTFS stores time properly in UTC. FAT/FAT32 stores in local time.
> Either way, time should be displayed in local time when
> file:read_file_info/1 is called.
> I did a little digging into the driver code an to appears the the
> FileTimeToLocalTime Win32API to convert the time. I suspect it is this
> API call that is the root of the problem. I don't have a Windows
> platform to test on, so I can't dig any further, but there is a
> discussion here <http://www.codeproject.com/KB/datetime/dstbugs.aspx>
> that might explain the problem more.
Looks like you're right. It's storing the timestamp in UTC, but then
destroys it when retrieving it, trying to make a local time of it.
Well, still a general bug in Windows, which don't have anything to do
with Erlang. In fact, from the way Windows seems to behave, you get a
better result (in this aspect) from FAT than NTFS file systems...
> 2008/10/27 Johnny Billquist <bqt@REDACTED <mailto:bqt@REDACTED>>
> John Hughes wrote:
> > I would have expected that, once a file has been created, then
> its creation
> > time should remain constant (unless explicitly
> modified)--wouldn't you? Not
> > so! At least, not under Windows. Under both XP and Vista, the
> file creation
> > time reported by file:read_file_info changes by an hour when
> daylight saving
> > time ends (and presumably also when it begins)!
> > File system design according to the Principle of Greatest
> > evidently.
> I believe that's a basic problem with windows, since it don't actually
> keep time in UTC, but local time, internally.
> In short: if you use Windows, that's the way it is. Don't have anything
> to do with Erlang.
> erlang-questions mailing list
> erlang-questions@REDACTED <mailto:erlang-questions@REDACTED>
More information about the erlang-questions