Message-ID: <3D5DD399.7090200@cogeco.ca>
Date: Sat, 17 Aug 2002 00:39:53 -0400
From: "Warren W. Gay VE3WWG" <ve3wwg@cogeco.ca>
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US;
	rv:1.0rc2) Gecko/20020618 Netscape/7.0b1
MIME-Version: 1.0
To: PostgreSQL mailing list <pgsql-interfaces@postgresql.org>
Subject: libpq bug handling signals
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Transfer-Encoding: 7bit

I am experiencing problems with lo_write() (mostly), WRT to
life after a signal has been handled.

Socket I/O (read(2) for example) will return -1 and have errno = EINTR
if it was blocked reading, and a signal was handled successfully while
it was blocked. It appears that lo_write() should do something similar,
but doesn't in practice ...

If I clear errno=0 and then perform lo_write()s during which
time a signal is received and handled, lo_write() will return -1
indicating an error. However, upon checking the
errno vallue, it indicates that no error has occurred (errno=0).

However, if I set errno=EINTR prior to calling lo_write() each
time, the same -1 is returned (after a signal) but now the
errno = EINTR (as I had originally expected it should have been).

It appears that EINTR processing is not correct or is incomplete in
the libpq code. I have experienced similar weirdness in the
operation of lo_lseek().

Do I need to subsequently submit this as a bug somewhere?

Thanks, Warren.
-- 

Warren W. Gay VE3WWG
http://home.cogeco.ca/~ve3wwg