Actually, the docs say SSL_read/write can error needing READ or WRITE:
http://www.openssl.org/docs/ssl/SSL_read.html#
The SSL_write docs are the same. I have applied the following patch to
handle both new cases. Does this help the SSL test program you have?
---------------------------------------------------------------------------
Tom Lane wrote:
> Bruce Momjian <pgman@candle.pha.pa.us> writes:
> > Yep, typo. Patched to CVS current and backpatched to 7.3.X.
>
> I think this fix is exactly backward. Why would SSL_write need to
> return ERROR_WANT_WRITE? It couldn't. The correct fix is that
> SSL_write might return ERROR_WANT_READ, for which reading would be
> the right response.
>
> BTW the real problem, both here and elsewhere in this file, is the
> lack of a "default: elog-out" case in the switch statements. This
> code will simply break if any unexpected case occurs.
>
> regards, tom lane
>
--
Bruce Momjian | http://candle.pha.pa.us
pgman@candle.pha.pa.us | (610) 359-1001
+ If your life is a hard drive, | 13 Roberts Road
+ Christ can be your backup. | Newtown Square, Pennsylvania 19073
Index: src/backend/libpq/be-secure.c
===================================================================
RCS file: /cvsroot/pgsql-server/src/backend/libpq/be-secure.c,v
retrieving revision 1.27
diff -c -c -r1.27 be-secure.c
*** src/backend/libpq/be-secure.c 29 Mar 2003 03:56:44 -0000 1.27
--- src/backend/libpq/be-secure.c 29 Mar 2003 04:59:01 -0000
***************
*** 285,290 ****
--- 285,293 ----
case SSL_ERROR_WANT_READ:
n = secure_read(port, ptr, len);
break;
+ case SSL_ERROR_WANT_WRITE:
+ n = secure_write(port, ptr, len);
+ break;
case SSL_ERROR_SYSCALL:
if (n == -1)
elog(COMMERROR, "SSL SYSCALL error: %s", strerror(errno));
***************
*** 336,341 ****
--- 339,347 ----
{
case SSL_ERROR_NONE:
port->count += n;
+ break;
+ case SSL_ERROR_WANT_READ:
+ n = secure_read(port, ptr, len);
break;
case SSL_ERROR_WANT_WRITE:
n = secure_write(port, ptr, len);
