Glenn Morris wrote:
> Your example doesn't correspond to how the code in Emacs used to be.
> Change it to:
> 
>    for (int pos = 0; gciphers[pos] != GNUTLS_CIPHER_NULL; pos++)
> 
> and it fails for me, just as Emacs used to:

Weirder and weirder. First, with that change the code still works for me (I 
tried it with GnuTLS 3.4.10). Second, if I read the GnuTLS 3.2.21 source code, 
the changed code should still work for you too. This is because the test 
gciphers[pos] != GNUTLS_CIPHER_NULL should succeed one iteration before the test 
gciphers[pos] != 0 succeeds.

> 
>    gciphers[17] == GNUTLS_CIPHER_UNKNOWN ?!?
> 
> As I said, it seems to be a happy accident that GNUTLS_CIPHER_UNKNOWN == 0.
> So your example as written happens to skip it.
> 
> I hope we have now established I'm not mad or lying... :)

Sorry, not established yet. :-) Quite possibly I'm the crazy one, of course. 
Still, there's something very odd going on here.

To make sure we're on the same page I am attaching the changed code, which still 
works for me.