On Tue, Feb 25, 2003 at 08:55:29AM +0100, martin.regner@xxxxxxxxx wrote:
>
> Brad Hards wrote:
> > Except, that both of them produce:
> > GLib-CRITICAL **: file ghash.c: line 138 (g_hash_table_lookup): assertion `hash_table != NULL' failed.
> >
>
>
> The problem seems to be that ncp_req_eid_hash is NULL.
> I added some g_warning printouts in ncp_eid_hash_lookup
> and ncp_hash_lookup (packet-ncp2222.inc)
>
> When I opened the file I got the following output:
>
> ncp_req_eip_hash_lookup ncp_req_eid_hash=0224D098
> ncp_req_eip_hash_lookup ncp_req_eid_hash=0224D098
> ncp_req_eip_hash_lookup ncp_req_eid_hash=00000000
>
> One thing to avoid the GTK-Critical printout is to add check if the pointer is null, and then return NULL. However I'm not sure if that
> is the best solution.
It appears that "dissect_nds_request()" can call "ncp_eid_hash_lookup()"
even if "pinfo->fd->flags.visited" is TRUE, but that means it's called
after the first sequential pass - but the hash tables are freed after
that pass.
If so, then either
1) the hash tables need *NOT* to be freed after the first
sequential pass
or
2) the call to "ncp_eid_hash_lookup()" has to be done only if
"pinfo->fd->flags.visited" is false.
(BTW, in "dissect_ncp_reply()", "resolve_eid" is initially set to FALSE,
and never set to TRUE, so the "if (resolve_eid) { ... }" code is dead.)
