bug#33780: network-stream.el: network-stream-certificate always returns nil

[Robert Pluim]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: Robert Pluim <rpluim@gmail.com>
>> Cc: darael@dracon.is,  tzz@lifelogs.com,  33780@debbugs.gnu.org
>> Date: Mon, 14 Jan 2019 17:25:55 +0100
>> 
>> When running the test suite, weʼre in batch mode, so thereʼs no way to
>> answer that question, as far as I know, so turning off the nsm is the
>> only way to go.
>
> You could also override the nsm-query-user function, I think.
>

True.

>> This only fails for the existing tests with ':nowait t', since then I
>> suspect nsm gets called automatically, whilst in the ':nowait nil'
>> case nsm never gets called (the existing tests all use
>> 'make-network-process' directly, rather than 'open-network-stream', so
>> they bypass nsm).
>
> I don't think I understood why nsm gets called only in the ":nowait t"
> case.  What did I miss?

process.c has:

#ifdef HAVE_GNUTLS
static void
finish_after_tls_connection (Lisp_Object proc)
{
  struct Lisp_Process *p = XPROCESS (proc);
  Lisp_Object contact = p->childp;
  Lisp_Object result = Qt;

  if (!NILP (Ffboundp (Qnsm_verify_connection)))
    result = call3 (Qnsm_verify_connection,
                    proc,
                    Fplist_get (contact, QChost),
                    Fplist_get (contact, QCservice));

so loading nsm.el causes nsm-verify-connection to get called in the
':nowait t' case. Presumably in the ':nowait nil' case gnutls-boot has
already completed the tls connection, and finish_after_tls_connection
never gets called (thatʼs speculation on my part). I donʼt know the
GnuTLS code well enough to know if this is a bug. Ted?

Robert