Re: emulating tls for systems without

[Geoffrey Keating <geoffk at apple dot com>]

On 06/10/2006, at 3:53 PM, Paolo Carlini wrote:

> Geoffrey Keating wrote:
>
> > > About this, I would be curious to know whether this enhancement
> > > benefits libstdc++/24025, which, if I remember correctly, only  
> > > affects
> > > systems not defining _GLIBCXX_HAVE_TLS...
> > It might make it worse.  The thread emulation code abort()s when it
> > runs out of memory.
> Humm, unfortunately I'm missing too many details about the  
> emulation, sorry about that. Do you mean, it essentially allocates  
> on the heap behind the scenes, similarly to the current ! 
> _GLIBCXX_HAVE_TLS case? Thanks in advance for any clarification,

I only looked at it very briefly, but it does:

+static void
+emutls_init (void)
+{
+#ifndef __GTHREAD_MUTEX_INIT
+  __GTHREAD_MUTEX_INIT_FUNCTION (&emutls_mutex);
+#endif
+  if (__gthread_key_create (&emutls_key, emutls_destroy) != 0)
+    abort ();
+}
+#endif

and

+      ptr = malloc (obj->size + sizeof (void *));
+      if (ptr == NULL)
+       abort ();

and so on.

(The big benefit of the non-emulated TLS is that memory is allocated  
on pthread_create(), so you know that either memory gets allocated or  
the thread doesn't get created, and either way you don't get random  
crashes...)

Attachment: smime.p7s
Description: S/MIME cryptographic signature