PATCH: Fix and simplify get_netid()

[Jules Colding]

Hi,

The linc2/src/linc-protocols.c:get_netid() function produced nonsense  
values for ORBNetID="ipaddr" on Mac OS X Leopard. Please see:

    http://bugzilla.gnome.org/show_bug.cgi?id=539510


The patch below fixes this and makes for a simpler get_netid() bu  
introducing a new helper function. The patch has been tested on  
Leopard and Ubuntu Gutsy. "make check" is also happy on both  
platforms. Please forgive me if the patch below is whitespace-messed-
up. My current mailer isn't too happy about inserted text...

May I commit? Comments?

Best regards,
   jules




Index: linc2/ChangeLog
===================================================================
--- linc2/ChangeLog (revision 2073)
+++ linc2/ChangeLog (working copy)
@@ -1,3 +1,13 @@
[phone]  Jules Colding  <colding@...>
+
+ * src/linc-protocols.c (get_netid): This function
+ failed on Mac OS X for (ORBNetID == "ipaddr"). It is
+ now fixed and simplified greatly by the use of the new
+ function get_first_non_local_ipaddr().
+ (get_first_non_local_ipaddr):  This function will
+ correctly deduce the first non-loopback interface
+ on the host.
+
  ============================== 2.14.13 ===========================

  2008-05-25  Tor Lillqvist  <tml@...>
Index: linc2/src/linc-protocols.c
===================================================================
--- linc2/src/linc-protocols.c (revision 2073)
+++ linc2/src/linc-protocols.c (working copy)
@@ -145,7 +145,49 @@
  #endif
  #define DEFAULT_LINC2_IFNUM (32)

+#ifndef G_OS_WIN32
  static char *
+get_first_non_local_ipaddr(char *buf,
+   size_t len)
+{
+ int sock;
+ const char *retv;
+ struct if_nameindex *ifname_idx_array;
+ struct if_nameindex *ifname_idx;
+ struct ifreq ifr;
+ struct sockaddr_in sin;
+
+ *buf = '\0';
+
+ sock = socket(AF_INET, SOCK_DGRAM, 0);
+ if (-1 == sock)
+ return;
+
+ ifname_idx_array = if_nameindex();
+ for (ifname_idx = ifname_idx_array; ifname_idx && ifname_idx-
 >if_name && ifname_idx->if_name[0]; ifname_idx++) {
+ strncpy(ifr.ifr_name, ifname_idx->if_name, IFNAMSIZ);
+ if (!ioctl(sock, SIOCGIFADDR, &ifr)) {
+ memcpy(&sin, &(ifr.ifr_addr), sizeof(struct sockaddr_in));
+ retv = (const char*)inet_ntoa(sin.sin_addr);
+ retv = strcmp("127.0.0.1", retv) ? retv : NULL;
+ if (retv) {
+ strncpy(buf, (const char*)inet_ntoa(sin.sin_addr), len);
+ break;
+ }
+ }
+ }
+
+ if (ifname_idx_array)
+ if_freenameindex(ifname_idx_array);
+
+ if (-1 != sock)
+ close(sock);
+
+ return buf;
+}
+#endif
+
+static char *
  get_netid(LinkNetIdType which,
   char *buf,
   size_t len)
@@ -161,58 +203,13 @@
  if ((LINK_NET_ID_IS_IPADDR == which)
     || (LINK_NET_ID_IS_CUSTOM == which)) {
  #ifndef G_OS_WIN32
- struct sockaddr_in *adr = NULL;
- struct ifreq *my_ifreqs = NULL;
- struct ifconf my_ifconf = { 0 };
- int sock = -1;
- int num = 0;
- int i = 0;
+ if (fixed_host_net_id)
+ strncpy(buf, fixed_host_net_id, len);
+ else
+ get_first_non_local_ipaddr(buf, len);
+ if ('\0' == *buf)
+ strncpy(buf, "127.0.0.1", len);

- sock = socket(AF_INET,SOCK_DGRAM,0);
- if (-1 == sock)
- goto out;
-#ifdef SIOCGIFNUM
- i = ioctl(sock, SIOCGIFNUM, &num);
- if ((0 > i) || !num) {
- close(sock);
- goto out;
- }
-#else
- num = DEFAULT_LINC2_IFNUM;
-#endif
- my_ifconf.ifc_len = sizeof(struct ifreq) * num;
- my_ifconf.ifc_req = (struct ifreq*)malloc(my_ifconf.ifc_len);
- if (!my_ifconf.ifc_req) {
- close(sock);
- goto out;
- }
- if (ioctl(sock,SIOCGIFCONF,&my_ifconf) < 0) {
- close(sock);
- free(my_ifconf.ifc_req);
- goto out;
- }
- close(sock);
-
- my_ifreqs = my_ifconf.ifc_req;
- for (i = 0; i < num; i++) {
- adr = (struct sockaddr_in *)&my_ifreqs[i].ifr_ifru.ifru_addr;
- if (fixed_host_net_id) {
- if (!strcmp(fixed_host_net_id, inet_ntoa(adr->sin_addr)))
- break;
- } else {
- if (strcmp("127.0.0.1", inet_ntoa(adr->sin_addr)))
- break;
- }
- if (!strcmp("0.0.0.0", inet_ntoa(adr->sin_addr))) {
- if (i)
- adr = (struct sockaddr_in *)&my_ifreqs[--i].ifr_ifru.ifru_addr;
- break;
- }
- }
- strncpy(buf, (const char*)inet_ntoa(adr->sin_addr), len);
- free(my_ifconf.ifc_req);
-
- /* will be 127.0.0.1 anyway, if no other address is defined... */
  return buf;
  #else
  SOCKET sock;

_______________________________________________
orbit-list mailing list
orbit-list@...
http://mail.gnome.org/mailman/listinfo/orbit-list

[Jules Colding]

On 21/06/2008, at 21.29, Jules Colding wrote:

> Hi,
>
> The linc2/src/linc-protocols.c:get_netid() function produced  
> nonsense values for ORBNetID="ipaddr" on Mac OS X Leopard. Please see:
>
>   http://bugzilla.gnome.org/show_bug.cgi?id=539510
>
>
> The patch below fixes this and makes for a simpler get_netid() bu  
> introducing a new helper function. The patch has been tested on  
> Leopard and Ubuntu Gutsy. "make check" is also happy on both  
> platforms. Please forgive me if the patch below is whitespace-messed-
> up. My current mailer isn't too happy about inserted text...
>
> May I commit? Comments?

Committed. Hopefully people will yell at me if something breaks...


--
   jules




_______________________________________________
orbit-list mailing list
orbit-list@...
http://mail.gnome.org/mailman/listinfo/orbit-list

[michael meeks]

Hi Jules,

On Sat, 2008-06-21 at 21:29 +0200, Jules Colding wrote:
> The linc2/src/linc-protocols.c:get_netid() function produced nonsense  
> values for ORBNetID="ipaddr" on Mac OS X Leopard. Please see:

        IPv4/6 is your baby ;-) go for it. OpenSolaris traditionally has
problems with changes here - but no doubt they can find those later
themselves.

        Regards,

                Michael.

--
 michael.meeks@...  <><, Pseudo Engineer, itinerant idiot


_______________________________________________
orbit-list mailing list
orbit-list@...
http://mail.gnome.org/mailman/listinfo/orbit-list

[Jules Colding]

Hi Michael,

On 23/06/2008, at 10.46, Michael Meeks wrote:
> Hi Jules,
>
> On Sat, 2008-06-21 at 21:29 +0200, Jules Colding wrote:
>> The linc2/src/linc-protocols.c:get_netid() function produced nonsense
>> values for ORBNetID="ipaddr" on Mac OS X Leopard. Please see:
>
> IPv4/6 is your baby ;-) go for it.

Thanks :-)


> OpenSolaris traditionally has
> problems with changes here - but no doubt they can find those later
> themselves.

Yes, I was worrying about Solaris myself. To bad that I don't have a  
box to test on before making changes...

Best regards,
   jules


_______________________________________________
orbit-list mailing list
orbit-list@...
http://mail.gnome.org/mailman/listinfo/orbit-list