From: Pavel Raiskup <praiskup@redhat.com>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Robert Haas <robertmhaas@gmail.com>,
 "pgsql-hackers@postgresql.org" <pgsql-hackers@postgresql.org>,
 Devrim =?ISO-8859-1?Q?G=DCND=DCZ?= <devrim@gunduz.org>
Subject: Re: plruby: rb_iterate symbol clash with libruby.so
Date: Wed, 07 Nov 2018 16:27:46 +0100
Message-ID: <1930142.Hb3bbeJPGd@nb.usersys.redhat.com>
In-Reply-To: <25954.1541600731@sss.pgh.pa.us>
References: <4738198.8KVIIDhgEB@nb.usersys.redhat.com>
 <5127122.UsUPqPCgps@nb.usersys.redhat.com> <25954.1541600731@sss.pgh.pa.us>
MIME-Version: 1.0
Content-Transfer-Encoding: 7Bit
Content-Type: text/plain; charset="us-ascii"
Precedence: bulk

On Wednesday, November 7, 2018 3:25:31 PM CET Tom Lane wrote:
> Pavel Raiskup <praiskup@redhat.com> writes:
> > On Tuesday, November 6, 2018 7:28:21 PM CET Tom Lane wrote:
> >> Done.  I realized that the immediate problem, rb_iterate(), was only
> >> added as of PG v10, which may explain why we hadn't heard complaints
> >> about this till now.  So I've made the change only as far back as v10.
> 
> > The 'rb_iterate' seems to exist at least in REL9_2_STABLE branch, so it is
> > probably much older.
> 
> Oh!  Hmm ... I think I jumped to conclusions when the part of my patch
> that touched struct RBTreeIterator failed to apply.  But you're right,
> rb_iterate has been there since 9.0 now that I look more carefully.
> So we really ought to back-patch further.  However:
> 
> > That
> > said, I'm still not sure how this could work before ...  Maybe it has not
> > been working for some time.
> 
> Yeah, I'm now mighty confused about this as well.  PL/Ruby is pretty old
> too, so how come nobody noticed this before?  Is its rb_iterate call in
> someplace that hardly gets any use?

I can not authoritatively answer this (getting familiar with the code) but
many of the rb_iterate() calls were replaced with alternative API (if
available), so you might be right.

That said, I reproduced this by trigger function from tests/plt test-case
on the first try, more info and related fixes in [1].  But it is truth
that we haven't run the test-suite for RPM builds so far.

Might the reason be that nobody used plruby at all for a very long time?

[1] https://github.com/devrimgunduz/postgresql-plruby/pull/3

Pavel