Re: PAR 0.85 bugs (Perl/Tk related)

The newest Tk is only a few months old.  I installed it recently.  A search
script to find any occurance of the string "TK_EXIT" in any .pm or .pl file
under c:\perl only came up with the line below.

C:\perl\site\lib\Tk.pm :  return if $mess =~/^_TK_EXIT_\(\d+\)/;

But the last underscore after "TK_EXIT_" is missing.   I am guessing you
mistyped it.  Anyway, here is the short subroutine the line is from:

------------------------paste
#
# This is a $SIG{__DIE__} handler which does not change the $@
# string in the way 'croak' does, but rather add to Tk's ErrorInfo.
# It stops at 1st enclosing eval on assumption that the eval
# is part of Tk call process and will add its own context to ErrorInfo
# and then pass on the error.
#
sub __DIE__
{
 my $mess = shift;
 my $w = $Tk::widget;
 # Note that if a __DIE__ handler returns it re-dies up the chain.
 return unless defined($w) && Exists($w);
 # This special message is for exit() as an exception see pTkCallback.c
 return if $mess =~/^_TK_EXIT_\(\d+\)/;
 return if $w->_backTrace;
 # Not in an eval - should not happen
}
-----------------------end paste

----- Original Message ----- 
From: "Alan Stewart" <astewart1@xxxxxxx>
To: "Owen Marshall" <omarshall@xxxxxxxxxxx>; <par@xxxxxxxx>
Sent: Friday, July 23, 2004 6:55 PM
Subject: Re: PAR 0.85 bugs (Perl/Tk related)


> On 13 Jul 2004 at 15:55, Owen Marshall wrote:
>
> > 2)  Closing a Tk window briefly prints the line "_TK_EXIT(0)" in the
> > cmd.exe window before it closes, when the application is not compiled
> > with the --gui switch (which I can't use in my case because of another
> > strange bug with stdout that I've previously reported) - this Tk exit
> > message is not normally part of stdout or stderr, and should not be
> > displayed.
> >
>
> Are you saying that this message only occurs when running a PAR packed
program and not
> from a command line script? This message is not from PAR itself and I also
can't find
> it anywhere in my ActivePerl distribution.
>
> Of course, the message won't print when compiled with --gui, because the
application
> doesn't have any stdout or stderr. A gui app is a "DETACHED_PROCESS" and
is expected to
> communicate solely thru it's windows. Is that the strange bug you've
encountered
> previously? I remember a discussion about how to re-instate the console a
while ago.
>
> Alan Stewart
>