Hi

I have an application with a main form, and several others whuich are only ever used in modal form.

The application is slow to terminate, and I cannot see why. The shutdown procedure, simplified a little looks like this: (datamodified a global Boolean; savedata another procedure in the same unit; the SQLcomponents are in a a data module)

procedure FinishUp;
begin
  if datamodified then
    savedata;
 sqltransaction1.commit;
 sqlconnection.close(True);

Application.Terminate {line A}

end;

The debugger shows me that everything is fine until line A, when the program hangs: often for 15 seconds, sometimes for ever.

XFCE on Linux complains that the window containing the main form cannot be closed because 'the application may be busy'.

I have tried to replace line A with 'halt' or 'close' without any difference. The connection to the database is indeed closed.

I am using Lazarus 1.4.2, 64-bit, Linux XFCE desktop, GTK+2 widget set.

Any suggestions about where to look next will be very welcome. -------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus.freepascal.org/pipermail/lazarus/attachments/20150930/0549f749/attachment.html>

1. Are you using threads ?
2. Do you create a lot of components dynamically that are freed in the destructor of the form ?

Both can cause significant delays.

I am using Lazarus 1.4.2, 64-bit, Linux XFCE desktop, GTK+2 widget set. Any suggestions about where to look next will be very welcome.

What I usually do in such case is not use the debugger, but check using the strace command-line utility if there is still some activity in the application:

Put a writeln('something') before and after the Application.Terminate, and check what happens between the write() commands in strace's output.

Michael.

You said Linux with xcfe, but which version of LInux? Some of them
behave quite differently from others.

Gordon.

Hi Gordon - thanks for replying

OpenSuse, 13.2, with XFCE 4.10

Susie
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus.freepascal.org/pipermail/lazarus/attachments/20150930/a4196f56/attachment.html>

On Wed, Sep 30, 2015 at 7:50 PM, Michael Van Canneyt <

Hi Michael.

I'm not using threads.

Very few overall, and I thought they were safely freed. But heaprtc shows 66 unfreed memory blocks, so that is something I shall check tomorrow (it's late in NZ)

Thanks for that ingenious suggestion. A quick check shows nothing happens between the two write()s, so Application.Terminate is returning immediately. There follows 96 thousand lines of output before the program gets SIGKILL, so I guess that it was pretty busy!

Something for me to get my teeth
Susie
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus.freepascal.org/pipermail/lazarus/attachments/20150930/db567d2f/attachment-0001.html>

Very few overall, and I thought they were safely freed. But heaprtc shows 66 unfreed memory blocks, so that is something I shall check tomorrow (it's late in NZ)

?
Resolve this errors and then try without heaptrc, it is expected that memory analysis take at lot of time when the program exit.

Patrick

Hello Patrick,

Patrick Chevalley wrote

Resolve this errors and then try without heaptrc, it is expected that memory analysis take at lot of time when the program exit.

TIP:
Also develop your programs with -gh -gl (memory leak detection) enabled, and show console output. It is MUCH easier to fix a memory leak the minute you introduced it, compared to months down the line when you have stacks of memory leaks.

Regards,
- Graeme -

fpGUI Toolkit - a cross-platform GUI toolkit using Free Pascal http://fpgui.sourceforge.net/

My public PGP key: http://tinyurl.com/graeme-pgp

Hi,

I am using windows32 and I want to hnow something about the Heaptrc for windows. Can you send me where I can read about?

Thanks in advance,

Ing. H?ctor Fiandor

Hello H?ctor,

H?ctor Fiandor wrote

Can you send me where I can read about?

http://www.freepascal.org/docs-html/3.0.0/rtl/heaptrc/index.html

http://www.freepascal.org/docs-html/rtl/heaptrc/usage.html

Regards,
- Graeme -

fpGUI Toolkit - a cross-platform GUI toolkit using Free Pascal http://fpgui.sourceforge.net/

My public PGP key: http://tinyurl.com/graeme-pgp

Hi Graeme,

TIP: Also develop your programs with -gh -gl (memory leak detection)

I agree and always use heaptrc for my development projects.

My suggestion to try without heaptrc is only to check if this is the reason for "slow shutting down". In this case the problem is solved because a release build without this option will not be affected.

Patrick

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus.freepascal.org/pipermail/lazarus/attachments/20150930/79d0cf42/attachment.html>

Hello Patrick,

Patrick Chevalley wrote

My suggestion to try without heaptrc is only to check if this is the reason for "slow shutting down". In this case the problem is solved because a release build without this option will not be affected.

Run your program from the command line, and don't compile it with -WG (ie: tell the compiler it is a console program, not a GUI one - if you are using Windows).

If it takes so long to shut down, that means you have a heck of a lot of memory leaks. As you can guess, that is bad and should always be fix! Simply disabling heaptrc is not really a solution.

But yes I fully understand "releas builds" of your application shouldn't include heaptrc. But memory leaks should be fixed none the less.

Regards,
- Graeme -

fpGUI Toolkit - a cross-platform GUI toolkit using Free Pascal http://fpgui.sourceforge.net/

My public PGP key: http://tinyurl.com/graeme-pgp

Thanks all.

Once I tracked down the two components that had not been freed, the hanging stopped.

Some create tips in this thread. Thansk for all of them.

On Thu, Oct 1, 2015 at 6:31 AM, Graeme Geldenhuys <

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus.freepascal.org/pipermail/lazarus/attachments/20151001/1a92515e/attachment.html>

Sorry all: that last reply was from me (the OP).

I was on my friend's machine, and he is also on the list. Confusion is universal.

Susie

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus.freepascal.org/pipermail/lazarus/attachments/20151001/79000cff/attachment-0001.html>