Re: Delay time to spawn new threads?

[David Buchan <pdbuchan yahoo com>]

David Buchan <pdbuchan yahoo com> wrote:
> Hi Michael,
>
> My 32-bit, GTK+2 version does
>
>   // Secure glib
>   if (!g_thread_supported ()) {
>     g_thread_init (NULL);
>   }
>
> at the beginning, and then the thread is spawned via:
 on_button1_clicked (GtkButton *button1, MyData *data)
> {
>   GThread *thread;
>   GError *error = NULL;
>
>   thread = g_thread_create ((GThreadFunc) my_function, data, FALSE,
> &error); if (! thread) {
>       g_print ("Error: Unable to create new thread for my_function()
> in on_button1_clicked().%s\n", error->message); exit (EXIT_FAILURE);
>     }
>
> My 64-bit, GTK+3 versions does not do the g_thread_init() call.
>
> It spawns a new thread via:
>
> int
> on_button1_clicked (GtkButton *button1, MyData *data)
> {
>   GThread *thread;
>
>   thread = g_thread_new ("my_function", (GThreadFunc) my_function,
> data); if (! thread) {
>       fprintf (stderr, "Error: Unable to create new thread for
> my_function() in on_button1_clicked().\n"); exit (EXIT_FAILURE);
>     }

Show us your my_function(): you are almost certainly doing something
wrong. Best of all, provide a complete compilable example which
demonstrates the problem. And why are you casting the function pointer
to GThreadFunc?  You do not need to call g_thread_init() with glib >=
2.32, and you do with earlier versions. Prior to version 2.24
g_thread_init() had to be the first glib call. Between 2.24 and 2.30 it
had to be the first
 call relevant to threads.

Also, please don't top post.

Chris
==========================================
Hi Chris,


I removed the call to g_thread_init() and it still works fine! Great catch there.

I had put the cast to GThreadFunc there because otherwise I get the following error:

callbacks.c: In function ‘on_button1_clicked’:
callbacks.c:3829: warning: passing argument 1 of ‘g_thread_create_full’ from incompatible pointer type
/usr/include/glib-2.0/glib/gthread.h:225: note: expected ‘GThreadFunc’ but argument is of type ‘int 
(*)(struct MyData *)’

I saw other people on the web do the cast as well to remove that error. Is there a mistake in how I use 
g_thread_create()? I'd like to drop the cast if I can. It seemed strange at the time I put it in.

Top-posting is due to the way Yahoo mail handles replies.
Maybe this
 is better, but it didn't put the >'s in, thus the "====..." line. I dunno. Sorry about that.

I'm still investigating the delay. The my_function() code is hundreds of lines, so I'd need to create a small 
example to post.
Let me continue to investigate.

Dave

==========================
... and now I see what's going on.

My newly spawned thread calls a function which starts a messaging idle-function.

[my main program] ---> [my new thread] ----> [post_message function] ===> [update GUI idle function]

In my new thread, I call:

post_message (text, data);

and that looks like:

// Start an instance of activity_message() idle function.
int
post_message (char *message, MyData *data)
{
  // Allocate memory for string; Idle function will free it.
  data->activity_message = allocate_strmem (TEXTLEN);
  // Copy message.
  strncpy (data->activity_message, message, TEXTLEN - 1);  // Minus 1 for string termination.

  g_idle_add ((GSourceFunc) activity_message, data);
  sleep (1);
  return (EXIT_SUCCESS);
}

// Idle function to add a message to the activity textview
// This idle function returns 0 in order to stop.
int
activity_message (MyData *data)
{
  ...
  Update a textview with our message ...

  return (0)
}

I set it up this way so the GUI can be updated.

The call to sleep() protects me against the situation in which my thread makes a second call to 
post_message() when activity_message() hasn't finished updating the GUI from the first call. 

Is there a way to have my thread call post_message() whenever it needs to, and have the messages queue-up?

Dave