stands a chance of looking it up.  "Unrecognised error" is always
disheartening.  :-)

Dominic Mitchell

Always create thread-specific variable.

global variables are problematic on this platform.  Simplest solution
seems to be to initialize pthread key variable to 0.  Also, rename this
variable and check_sigpipe_handler to something involving "pq" to
avoid gratuitous pollution of application namespace.

> user must initialize the lock. The problem are concurrent "first" users
> - the pthread_mutex_t initialization must be synchronized.
> The current implementation is broken, the attached patches fixes that:
> mutex_initlock is a spinlock. If the pthread_mutex_t mutex is not
> initialized, then the spinlock is acquired, if the pthread_mutex_t is
> initialized if it's not yet initialized and then the spinlock is dropped.

Manfred Spraul

user must initialize the lock. The problem are concurrent "first" users
- the pthread_mutex_t initialization must be synchronized.
The current implementation is broken, the attached patches fixes that:
mutex_initlock is a spinlock. If the pthread_mutex_t mutex is not
initialized, then the spinlock is acquired, if the pthread_mutex_t is
initialized if it's not yet initialized and then the spinlock is
dropped.

Manfred Spraul

Andreas Pflug

be interpreted just the other way round in initialize_SSL.

Andreas Pflug

conversion of basic ASCII letters.  Remove all uses of strcasecmp and
strncasecmp in favor of new functions pg_strcasecmp and pg_strncasecmp;
remove most but not all direct uses of toupper and tolower in favor of
pg_toupper and pg_tolower.  These functions use the same notions of
case folding already developed for identifier case conversion.  I left
the straight locale-based folding in place for situations where we are
just manipulating user data and not trying to match it to built-in
strings --- for example, the SQL upper() function is still locale
dependent.  Perhaps this will prove not to be what's wanted, but at
the moment we can initdb and pass regression tests in Turkish locale.

I have removed the docs mentioning that SSL and Kerberos are not
thread-safe.

Manfred Spraul

ignore SIGPIPE from send() in libpq, but terminate on any other SIGPIPE,
unless the user installs their own signal handler.

This is a minor fix because the only time you get SIGPIPE from libpq's
send() is when the backend dies.

Per Neil Conway.

$Header: -> $PostgreSQL Changes ...

and improve tests.

platform.

Andreas Pflug

Lee Kindness

protocol 3, then falls back to 2 if postmaster rejects the startup packet
with an old-format error message.  A side benefit of the rewrite is that
SSL-encrypted connections can now be made without blocking.  (I think,
anyway, but do not have a good way to test.)

Curious that gcc doesn't complain about unreferenced static variables.

believe I didn't notice this before -- once 64k was sent to/from the
server the client would crash.  Basicly, in 7.3 the server SSL code set
the initial state to "about to renegotiate" without actually starting
the renegotiation.  In addition, the server and client didn't properly
handle the SSL_ERROR_WANT_(READ|WRITE) error.  This is fixed in the
second patch.

Nathan Mueller

first, that I missed when checking over 7.3.1, was that the client
method was switched to SSLv23 along with the server.  The SSLv23 client
method does SSLv2 by default, but can also understand SSLv3.  In our
situation the SSLv2 backwords compatibility is really only needed on the
server.  This is the first patch.

The last thing is that I found a way for the server to understand SSLv2
HELLO messages (sent by pre-7.3 clients) but then get them to talk
SSLv3.  This is the last one.

Nathan Mueller

From the SSL_CTX_new man page:

"SSLv23_method(void), SSLv23_server_method(void), SSLv23_client_method(void)

 A TLS/SSL connection established with these methods will understand the SSLv2,
 SSLv3, and TLSv1 protocol. A client will send out SSLv2 client hello messages
 and will indicate that it also understands SSLv3 and TLSv1. A server will
 understand SSLv2, SSLv3, and TLSv1 client hello messages. This is the best
 choice when compatibility is a concern."

This will maintain backwards compatibility for those us that don't use
TLS connections ...

returned -1, per SSL_get_error() documentation.

Nathan Mueller

as well as unnecessary (and incorrect on Windows) assignments to
errno/SOCK_ERRNO.

support IPv6 anyway.