From: Tom Lane <tgl@sss.pgh.pa.us>
To: exclusion@gmail.com
cc: pgsql-bugs@lists.postgresql.org
Subject: Re: BUG #16685: The ecpg thread/descriptor test fails sometimes on
 Windows
In-reply-to: <16685-d6cd241872c101d3@postgresql.org>
References: <16685-d6cd241872c101d3@postgresql.org>
Comments: In-reply-to PG Bug reporting form <noreply@postgresql.org>
	message dated "Sat, 24 Oct 2020 04:05:07 -0000"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <367781.1603554604.1@sss.pgh.pa.us>
Date: Sat, 24 Oct 2020 11:50:04 -0400
Message-ID: <367782.1603554604@sss.pgh.pa.us>
Precedence: bulk

PG Bug reporting form <noreply@postgresql.org> writes:
> I found that the get_descriptors() function called in ECPGdeallocate_desc
> sometimes can return null.

> The following debugging code inserted into the ECPGallocate_desc:
> ...
> shows on a failure:
> TlsGetValue() returned null on iteration 209, error: 0, descriptor_key: 28,
> initial descriptor_key: 0.
> or
> TlsGetValue() returned null on iteration: 369, error: 0, descriptor_key: 28,
> initial descriptor_key: 0

Hm.  I'm supposing that 0 isn't likely to be the correct TLS index,
and what this is showing us is that we got here before anyone had
done descriptor_key_init.

I'll bet that the correct fix is

 static void
 set_descriptors(struct descriptor *value)
 {
+	pthread_once(&descriptor_once, descriptor_key_init);
	pthread_setspecific(descriptor_key, value);
 }

and that you could probably reproduce this on non-Windows, too,
if you tried hard (sticking a delay in get_descriptors might do it).

			regards, tom lane