Re: oops? read/write vs type of length parameter

[Ted Zlatanov <tzz@lifelogs.com>]

On Fri, 15 Apr 2011 09:34:58 -0700 Paul Eggert <eggert@cs.ucla.edu> wrote: 

PE> On 04/15/2011 06:13 AM, Ted Zlatanov wrote:
>> Paul, did you already do these fixes in revno 103919?

PE> Yes.  The APIs for emacs_gnutls_read and emacs_gnutls_write
PE> had already changed recently, in bzr 103878,
PE> and I figured that if merging will be needed anyway
PE> we might as well merge the desired API.

>> I hope rebasing Claudio Bley's GnuTLS patch on top of your fixes is
>> trivial.  If not I may have to ask you for help later.

PE> It should be trivial, and I'll be happy to help if it isn't.

I posted a rebased patch with an emacs_gnutls_write change required
because of your revisions.

I think the old code returned -1 for 0 bytes to indicate the read should
be retried, but the new code doesn't have that return.  Still it works,
since the GnuTLS-specific EAGAIN (GNUTLS_E_AGAIN) is its own error code.
So the old behavior of returning -1 for 0 bytes was actually unnecessary.

I am using the patch now and have had no problems.

Ted