Re: [Evolution-hackers] Re: the bug about libsoup

[Sean Gao <sean gao sun com>]

JP Rosevear wrote:

> On Thu, 2003-04-17 at 06:23, Sean Gao wrote:
>  
>
> > Hi JP,
> >    I have tested as Joe suggested. The puzzle is that 
> > soup_ssl_hup_waitpid is not entered at all.
> > However, it does be really called on Linux Platform, that's why I say it 
> > could be the difference
> > of G_IO_XXX implementation or signal mechanism on different platforms 
> > causes such issue
> >    
>
> Are you getting reads of zero length instead?
>
> -JP
>  
Hi JP,
   I just added some code in soup_ssl_hup_waitpid() to test if I read 
zero length. It really gets zero bytes reading,
but after that  it complete the waitpid and following process and exit 
this function,  hydrogen hangs ... I do not know
if my method is right  or there may be some other problems casue the 
hanging. 
   Of course I add G_IO_IN as one of the condition to be watched.
  

Regards
-Sean


static gboolean
soup_ssl_hup_waitpid (GIOChannel *source, GIOCondition condition, 
gpointer user_data)
{

#ifndef RESPONSE_BLOCK_SIZE
#define RESPONSE_BLOCK_SIZE 8192
#endif
   SoupSSLInfo *ssl_info = user_data;

   gchar read_buf [RESPONSE_BLOCK_SIZE];
   gsize bytes_read = 0;
   GIOError error;

   if (condition&G_IO_IN) {
       error = g_io_channel_read (source,
                   read_buf,
                   sizeof (read_buf),
                   &bytes_read);           

       if  (!(error == G_IO_ERROR_NONE && bytes_read == 0))
           return TRUE;
   } 

   while (waitpid (ssl_info->ppid, NULL, 0) == -1 && errno == EINTR)
       ;
  
   g_io_channel_unref (ssl_info->real_sock);
   g_free (ssl_info);

   return FALSE;
}