[Savannah-hackers] Re: lsh crashed - core available

[Niels Möller]

Loic Dachary <address@hidden> writes:

>       lsh crashed once again. This time it dumped a core. I hope
> you can cure it from its disease, it makes everyone a bit edgy :-}

> #4  0x8067797 in werror_vformat (f=0x808091e "\n", args=0xbfffb94c)
>     at werror.c:407
> #5  0x8067a32 in verbose (format=0x8080900 "Unknown publickey algorithm %a\n")
>     at werror.c:542
> #6  0x805be03 in do_authenticate (s=0x80a1470, connection=0x80d54b8, 
>     username=0x811cc58, service=35, args=0xbfffb9e0, c=0x811cc38, e=0x811b8a8)
>     at server_publickey.c:107

It seems that werror("...%a...") couldn't handle the atom 0, which is
used to represent any algorithm not present in lsh's list in atoms.in.

The patch below should fix that. This seems unrelated to the previous
crash in do_dsa_sign (if it really crashed in that function, it's hard
to say just looking at the log), so it would be good if you could keep
extended logging and the patch for more information from do_dsa_sign
for yet some time.

Thanks for your patience,
/Niels

--- werror.c    2001/07/04 18:37:56     1.61
+++ werror.c    2001/09/12 07:23:51
@@ -429,11 +429,11 @@
            case 'a':
              {
                int atom = va_arg(args, int);
-               
-               assert(atom);
 
-               werror_write(get_atom_length(atom), get_atom_name(atom));
-               
+               if (atom)
+                 werror_write(get_atom_length(atom), get_atom_name(atom));
+               else
+                 werror_write(9, "<unknown>");
                break;
              }
            case 's':