[DUG] Program crashes - anyone knows why?

[Gajo Csaba]

On Mon, 27 Nov 2006 00:14:39 +0100, Matthew Comb <matt at ferndigital.com>  
wrote:

> You are creating your critical section within your worker thread. This
> won't protect anything as all of your worker threads will have their own
> instance of the critical section.

Hmm, I do this at a few more places too. What should I do instead?
Actually, THIS may be what's causing the error. When I process a keyword,  
I put the result into the following structure:

   GoogleResult = ^ResultRec;
   ResultRec = packed record
     next: GoogleResult;
     srNo: integer;
     // etc...
   end;

This GoogleResult is then used as a global variable. The main form has a  
public property
ResultList: GoogleResult. Each thread has a pointer back to the main form.  
To ass the result to the list, I write

AddItem(mainForm.ResultList, newItem);

And this is the procedure:

procedure AddItem(var list: GoogleResult; const item: GoogleResult);
var
   tmp: GoogleResult;
   critical: _RTL_CRITICAL_SECTION;
begin
   try
     InitializeCriticalSection(critical);
     EnterCriticalSection(critical);
     if list = nil then
       list := item
     else begin
       tmp := list;
       while tmp^.next <> nil do
         tmp := tmp^.next;
       tmp^.next := item;
     end;
     LeaveCriticalSection(critical);
     DeleteCriticalSection(critical);
   except on Exception do { nothing }
   end
end;

You see I've put everything in a critical section? So this won't do  
anything because I'm calling the procedure from the thread? What should I  
do instead, use Synchronize?
Hmm or maybe I should just use a TThreadList here as well. I've left this  
old list for historical reasons (in the previous version the program was  
single threaded).

> You are creating your critical section to access a boolean ? This is not
> needed as you cannot have a conflict on a boolean type.

There's a checkbox on the main form which enables/disables caching. If I  
change it while a thread is accessing the variable then there can be a  
conflict.

> Are any of your worker threads passed the same file name ? This for sure
> would cause a problem.

No they each create their own random-named file, before making sure it  
doesn't exist.

Csaba