[SOLVED]error when modifying main form from indy thread

[Michael Collier]

I've posted this here (linux forum) because this error only seems to happen on my linux machine and not on windows, but I'm not sure if it is related to linux per se. I'm using lazarus v1.0.10 wth fpc2.6.2 compililng with dwarf2 , not using cmem.

I have an indy server object that uses sendmessage() to update a memo on the main form. Very simple but on linux I get SIG type errors, the call stack shows functions relating to gtk e.g. gtk_text_layout_xxxxxxx

My understanding is that calls from threads to the main form are ok using sendmessage, there are no other updates to the main form (I don't even move the mouse over the application).

I even tried using a criticalsection, just to see if it made any difference (it didn't).

Could the problem be with gtk? Can I do anything about that?
Is the problem with lazarus 1.1.10?

[JuhaManninen]

Use TThread.Synchronize.

[balazsszekely]

Always update visual controls with TIdSync or TIdMessage. The first one is synchronous the later on the other hand is asynchronous.
Something like this:

Code: Pascal  [Select][+][-]

1. uses IdSync;
2. type
3.   TSync = class(TIdSync)
4.   protected
5.     procedure DoSynchronize; override;
6.   public
7.     FMessage: string;
8.   end;
9.  
10. procedure TSync.DoSynchronize;
11. begin
12.   Form1.Memo1.Lines.Add(FMessage)
13. end;
14.  
15. procedure TForm1.IdTCPServer1Execute(AContext: TIdContext);
16. var
17.   Sync: TSync;
18. begin
19.   //...
20.   if (SomeCondition) then
21.   begin
22.     Sync := TSync.Create;
23.     try
24.       Sync.FMessage := 'This is a test';
25.       Sync.Synchronize;
26.     finally
27.       Sync.Free;
28.     end;
29.   end;
30. end;                            
31.  

[Michael Collier]

Thank you, it is much better now

[taazz]

what happens if the postmessage is used instead? Is this behavior a bug (probably) or by design on GTK? Just to clarify a bit of what I'm asking here I'm quoting the help file

The difference between SendMessage and PostMessage is the way that they return control to the calling thread. With SendMessage control is not returned until the window that the message was sent to has completed processing the sent message, however with PostMessage control is returned immediately.

[Michael Collier]

Just tried postmessage() to find out what would happen, same errors as sendmessage().

[taazz]

Thank you for testing. At this point I think that this issue worth a bug report but since I haven't looked close enough on the message implementation in GTK I'll probably stay back on this one and let others decide. One last think can you create a simple test application that shows the error something along the lines of a thread sending a message every 10 seconds up to a number of times eg 10 messages. I would like to run it on QT as well to see how it fairs.

[Michael Collier]

Ok I'll get something done today, I can test on apple mac too if you like. Not sure what is meant by QT though, is it this http://wiki.freepascal.org/Qt_Interface

[zeljko]

Postmessage works fine with Qt (multithreading app).

[Michael Collier]

I have attached test project..

thread_sendmessage.zip

You can choose the number of messages, on linux I can do 3 messages without crashing, but higher numbers cause the error to happen.

You can also select a sleep period in milliseconds between messages, not sure if this has any affect on bug.

[Michael Collier]

I extended the test project (attached thread_sendmessage2.zip )to offer the option of performing update(),refresh() & processmessages() procedures and retested linux/windows and this time apple mac too.

The mac crashes if update(),refresh() & processmessages()  is called within sendmessage() but ok for postmessage() and  synchronise(). The mac is Version 10.5.8. Lazarus 1.0.12 fpc 2.6.2

Windows is fine, under all circumstances, so I guess an application tested on windows could fail on apple.

thread_sendmessage2.zip

[taazz]

I'll take a closer look on the demo later tonight. Thank you for your time, I appreciate it.

[Michael Collier]

I'm happy to help, any improvements to lazarus benefits us all 

[taazz]

I'm still having problems installing lazarus 1.4.4 on my linux. Testing is impossible at this time for me I have the test application though so at some point in the future I will run my tests.
Sorry for any and all delays on this .

[Michael Collier]

No worries, thanks for trying