On Monday 08 October 2012 15:30:31 Peter Stuge wrote:
> libs...@git.stuge.se wrote:
> > +++ b/src/channel.c
> > @@ -1483,10 +1483,11 @@ libssh2_channel_get_exit_signal(LIBSSH2_CHANNEL
> > *channel, char **langtag,
> > size_t *langtag_len)
> > {
> > - LIBSSH2_SESSION *session = channel->session;
> > size_t namelen = 0;
> >
> > if (channel) {
> > + LIBSSH2_SESSION *session = channel->session;
> > +
> > if (channel->exit_signal) {
> > namelen = strlen(channel->exit_signal);
> > if (exitsignal) {
>
> I think this fix is wrong. Please look at what happens after the condition.
It does exactly what the comments above the functions suggests, doesn't it?
> Also, I don't think that libssh2 needs to validate programmer input.
> If someone passes a NULL pointer to a function that is really an
> error, and they will then have a problem sooner or later anyway.
>
> It is much better for libssh2 to crash fast and hard in this case, to
> have a higher chance that the programmer discovers the error.
I fully agree with your attitude on this. Then we should just the check and
update the comment above the function, right?
Kamil
_______________________________________________
libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel
