checking if the memory was allocted previously

[mjoachimiak]

Hi,

I used smth like this (forgive me typos):

//////////////////////////
mystrucure *dynamic_element = NULL

void some_function()
{
if (dynamic_element == NULL)
dynamic_element = malloc(sizeofm(mystructure));
if (dynamic_element != NULL) do_smth(dynamic_element);
}

main()
{
call some_function() many times;
if (dynamic_element != NULL)
free(dynamic_element);
}
/////////////////////////////

there is a doc about GMutex
http://library.gnome.org/devel/glib/unstable/glib-Threads.html#GMutex
that says that this line

if (!mutex) mutex = g_mutex_new ();


can cause race condition.

what about my previous memory allocation ?
Can it cause race condition also?

BR

 

[Bart van Ingen Schenau]

Hi,

I used smth like this (forgive me typos):

//////////////////////////
mystrucure *dynamic_element = NULL

void some_function()
{
if (dynamic_element == NULL)
                 dynamic_element = malloc(sizeofm(mystructure));
if (dynamic_element != NULL) do_smth(dynamic_element);

}

main()
{
call some_function() many times;
if (dynamic_element != NULL)
   free(dynamic_element);}

/////////////////////////////

there is a doc about GMutexhttp://library.gnome.org/devel/glib/unstable/glib-Threads.html#GMutex
that says that this line

 if (!mutex) mutex = g_mutex_new ();

can cause race condition.

what about my previous memory allocation ?
Can it cause race condition also?

If some_function can be called from more than one thread, and you
can't guarantee that the memory has been allocated before those
threads are started, then your code also has a race condition, the
same as for the GMutex case.

If your program only has one thread (which is the default), then it is
absolutely impossible to have race conditions.

BR

Bart v Ingen Schenau

 

[Marcel Müller]

Hi!

//////////////////////////
mystrucure *dynamic_element = NULL

void some_function()
{
if (dynamic_element == NULL)
dynamic_element = malloc(sizeofm(mystructure));
if (dynamic_element != NULL) do_smth(dynamic_element);
}

If you use C++ you should prefer new/delete over malloc where possible.
This eliminates the need for the second NULL check since new will not
return NULL.

main()
{
call some_function() many times;
if (dynamic_element != NULL)
free(dynamic_element);
}
/////////////////////////////
what about my previous memory allocation ?
Can it cause race condition also?

Yes, if and only if "call some_function() many times" calls
some_function in parallel.

You should put the initialization of dynamic_element in main before the
other processing. This would completely eliminate the need for the NULL
checks in some_funtion.

However, the content of mystructure has to be accessed synchronized or
atomic or read only anyway if you intend to operate in parallel without
race conditions.


Marcel