Re: [Bacula-devel] [PATCH] ipv6 and multipe ipv4 binds

[Kern S. <ke...@si...>]

I'm copying the devel list, please do the same (thanks). Rather than
snipping, I'll top answer:

- I'm rethinking the "address" keyword issue. I had never informed users
of the new names ...  maybe now is the time to do it.

- I've reindented several of the files using my latest .indent.pro
  (a bit different from what I sent you). It still screws up a lot
  of things, but is better than what we had.

- I've created a branch so that we can synchronize -- see next email.

- Thanks for the free_resource patch -- applied, but there is still
  a problem in the FD.

Regards, Kern

On Sun, 2004-08-01 at 08:35, Meno Abels wrote:
> Kern Sibbald wrote:
> 
> >Hell Meno,
> >
> >On Sat, 2004-07-31 at 22:48, Meno Abels wrote:
> >  
> >
> >>Hello Kern,
> >>
> >>So I have a look to the problem it is pretty easy you modified my 
> >>orignal code
> >>to bstrncpy to prefend buffer overflows, but inet_ntop has also no 
> >>buffer limiting
> >>value -:(.
> >>You mix the parameter order
> >>from (char *, const char*, size_t) -> (char *, size_t, const char *) so 
> >>fix it -:)
> >>    
> >>
> >
> >Silly me -- sorry.
> >
> >  
> >
> >>In the standard config -Wall isn't enable which should?
> >>    
> >>
> >
> >I *always* add it to my ./configure line:
> >CFLAGS="-g -O2 -Wall" ./configure \
> >...
> >
> >
> >  
> >
> >>-Wall related thinks:
> >>* delete line 58 in src/store/stored_conf.h and src/store/stored_conf.c
> >>   it is a nested comment (it is really in both 58)
> >>    
> >>
> >
> >This is going to cause a problem that we need to fix. We cannot delete
> >Address. It will break too many .conf files, which is against my
> >philosophy of programming. It will probably also make my regression
> >scripts fail.
> >
> >  
> >
> You never used the address anywhere so i deleted it infact of nothings 
> happend. But your
> philosophy is welcome so we put these lines back:
> 
> int static RES_ITEM store_items[] = {
> 	{"address",               store_str,  ITEM(res_store.address),    0, 0, 0}, /* deprecated */
> 
> 
> in
> free_resource
>      if (res->res_dir.address) {
>         free(res->res_dir.address);
>      }
> 
> >>- To the indenting do you use tab or spaces and what are the rules?
> >>012456789012345
> >>void foo()
> >>{
> >>       i = 4;
> >>       if (foo) {
> >>            if (foo++) {
> >>                 call_to(foo, foo+1, foo++,
> >>                            moo, moo+1);
> >>            }
> >>       }
> >>}
> >>    
> >>
> >
> >The indenting is 3 spaces for everything. Multiple spaces can be
> >replaced by a tab using the normal standards for a tab -- i.e. the tab
> >inserts spaces up to the next multiple of 8.
> >  
> >
> thanks i will apply to it in future.
> 
> >  
> >
> >>- I compiled the stuff also now on linux it compiles fine-:) you need 
> >>this patch
> >>linux doesn't have sa_len in sockaddr.
> >>--- src/dird/ua_cmds.c.orig     2004-07-31 22:43:23.000000000 +0200
> >>+++ src/dird/ua_cmds.c  2004-07-31 22:44:02.000000000 +0200
> >>@@ -581,7 +581,9 @@
> >>    }
> >> #ifdef HAVE_INET_NTOP
> >>    inet_ntop(ua->UA_sock->client_addr.sa_family,
> >>-                &(ua->UA_sock->client_addr), buf, 
> >>ua->UA_sock->client_addr.sa_len);
> >>+                &(ua->UA_sock->client_addr), buf,
> >>+                ua->UA_sock->client_addr.sa_family == AF_INET ?
> >>+                sizeof(sockaddr_in) : sizeof(sockaddr_in6));
> >> #else
> >>    buf = inet_ntoa(((struct sockaddr_in 
> >>*)&(ua->UA_sock->client_addr))->sin_addr);
> >> #endif
> >>    
> >>
> >
> >Thanks.
> >
> >The was also a problem with inet_ntoa() in bnet_server.c.
> >
> >OK, now it compiles, but there are two problems:
> >
> >1. It doesn't connect here. I get:
> >
> >rufus-dir: ERROR in bnet_server.c:103 Cannot open stream socket: Address
> >family not supported by protocol. Retrying ...
> >  
> >
> my source is out of sync so there is no way to dig in that-:(
> 
> >This may be caused by a change I made to eliminate serious compiler
> >warning.  I'll try to look at this in the morning ...
> >
> >2. There are two buffers that are not released when the SD terminates:
> >Orphaned buffer:      24 bytes allocated at line 154 of rufus-sd
> >address_conf.c
> >Orphaned buffer:      16 bytes allocated at line 94 of rufus-sd
> >bnet_server.c
> >  
> >
> the line numbers are not matching to mine so it is difficult to me, but 
> one i found
> that is in store_conf.c in free_resource case R_STORAGE i miss to release:
>      if (res->res_store.sddaddrs) {
>         free_addresses(res->res_store.sddaddrs);
>      }
> 
> 
> morning
> 
> kern