When I run that code I see a window with 4 buttons and text propmting me to "Choose on option". If I click any button I get a smaller window containing the text

This is the second box
Index = 1
[download]

and a single button ("Shutdown and close"). At this point the console shows:

+MAIN
+do_button
do_button Array_Index: 1
-do_button
==MAIN - the first box has been closed
[download]

When I click the shutdown button the window closes and the following additional text is generated:

abort_routine Array_Index: 1
+abort_routine
abort_routine Array_Index: 1
-abort_routine
-abort_routine
-MAIN
[download]

No errors nor unusual behavior occur. $Tk::VERSION is 804.027 and Perl -v reports 'This is perl, v5.8.7 built for MSWin32-x86-multi-thread'.

Thanks for your comment....

When I click the shutdown button the window closes and the following additional text is generated:
abort_routine Array_Index: 1
+abort_routine
abort_routine Array_Index: 1
-abort_routine
-abort_routine
-MAIN

I get a different series of responses than you....

Firstly when I run the code and the first box opens...

+MAIN
[download]

Then when I click on something in the first box...

+do_button
do_button Array_Index: 1
-do_button
==MAIN - the first box has been closed
[download]

Then in the last box, when I click on the "shutdown and close" button...

+abort_routine
abort_routine Array_Index: 1
Segmentation fault
[download]

Alternately, if I slimply click the close box, I instead see...

Segmentation fault
[download]

You see this is quite different from your response, ending in a segfault. But I notice you're using windows and I'm using Linux.

I'm running Fedora 7, perl -v says "This is perl, v5.8.8 built for x86_64-linux-thread-multi", the Tk rpm which is installed is tk-8.4.13-5.fc7.src.rpm from the livna rpm repository.

Is there any other info from me that would help work out what is going on?

I suspect there is something nasty happening when you re-enter abort_routine. You could add:

my $inAbort;
[download]

at the top of the script and change abort_routine to:

sub abort_routine {
    return if $inAbort;
    
    $inAbort = 1;
    
    debug("+abort_routine");
    debug("abort_routine Array_Index: $Array_Index");

    # DO SOMETHING USEFUL HERE TO CLEAN UP THE PROCESS

    $MW->destroy();
    debug("-abort_routine");
}
[download]

so that abort_routine is not re-entered during the destroy or (probably better) change the button command to initiate a different sub that just performs $MW->destroy ();:

...

        $button = $MW->Button(
            -text    => "Shutdown and close",
            -command => [ \&do_close ] )->pack( -side => 'bottom' );

...

sub abort_routine {
    debug("+abort_routine");
    debug("abort_routine Array_Index: $Array_Index");

    # DO SOMETHING USEFUL HERE TO CLEAN UP THE PROCESS

    debug("-abort_routine");
}


sub do_close {
    $MW->destroy();
}
[download]

---

DWIM is Perl's answer to Gödel

NEVER write scripts that try to create-destroy-create a MainWindow. The MainWindow in Tk IS the eventloop. The eventloop MUST continue to run for proper operation, and trying to destroy it, with some code saying "restart yourself at death" is not going to be reliable.

The way you approach your type of script, is to make one mainwindow,and then some "toplevel" windows to present information. You can destroy a toplevel window, without it affecting the mainwindow's eventloop; but repeated create/destroy cycles of a toplevel window, may cause a memory gain, so you should always design scripts to reuse windows...... don't destroy them, just withdraw them, then repack them with different widgets.

Here is a way to use your code .

and a better way to do it

Arghhh. Thanks for the tip. It all makes sense now and indeed, the segmentation faults disappear when following your suggested strategy of avoiding destroys.

I note that if you ever destroy a TopLevel window, you eventually provoke another segmentation fault. However if you withdraw the TopLevel windows instead of destroying them, you can go on seemingly indefinitely by withdrawing the old TopLevel window and then defining a subsequent TopLevel window.

Then to exit the MainLoop gracefully, you destroy the original (hidden and withdrawn) MainWindow.

Thanks,
Stefan

I note that if you ever destroy a TopLevel window, you eventually provoke another segmentation fault.

That dosn't sound right. You may have a Tk version that needs patching, or you are doing something weird like trying to access a widget packed in the toplevel, after you destroy the toplevel.

If you want to be happy with Tk, stop using destroy to eliminate windows. To exit gracefully, use plain old "exit" or "Tk::exit", they clean up. If you use $mw->destroy, you may be leaving the calling shell alive. Of course, that may be useful in certain circumstances.

#!/usr/bin/perl -w
use Tk;

my $mw = MainWindow->new(-title=>"#1");

$mw->Button(-text=>"Destroy Me", 
         -command=> sub { $mw->destroy })->pack;

$mw->Button(-text=>"Exit Me", 
         -command=> sub { Tk::exit })->pack;
                 
MainLoop;

print "ha ha still going\n";
<>;
[download]

---

I'm not really a human, but I play one on earth. Cogito ergo sum a bum