On Fri, 30 Aug 2002 Msc...@ao... wrote:
> Using Log::Dispatch::Screen instead doesn't show the problem.
> Dave, is there any limitations to Log::Dispatch::File you can think of?
Not that I know of. This is the first I've heard of such a problem.
> My suspicion is that this is caused by our closures holding references to the
> Log::Dispatch::File object(s) which is/are using a "local"ized filehandle
> instead of a $fh scalar variable (open my $fh, ">blah") to keep track of the
> opened file.
Actually, the filehandle in Log::Dispatch::File is lexical. The 'local'
bit in the code is just a Perl trick to get a filehandle into a scalar in
Perl 5.005 and earlier.
I can't imagine why closures would be a problem.
> I don't have an easy-to-reproduce test case yet, just to give you guys a
> heads-up and maybe you already know where to look for?
Without a test case I'm at a loss. Log::Dispatch::File is barely 100
lines of code. Not that I can't manage to fit a few bugs in there, but it
really is simple stuff so I don't see how this could be happening.
-dave
/*==================
www.urth.org
we await the New Sun
==================*/
|