On Fri, Aug 16, 2019 at 11:26 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> I wrote:
> > [ thinks some more... ] Actually, I'm beginning to recall that
> > we made changes here because v11 plpgsql is capable of actually
> > returning "trigtuple" where before it would always have made a copy.
> > If that's accurate, then very likely the bug exists further back
> > but requires some other PL than plpgsql to manifest.
>
> As I suspected ... the attached test case crashes 9.4 through 11.
> We already had some problems in this area, which is why a suitable
> trigger is already at hand in regress.c.
Ah, I see. I had written a test patch that uses plpgsql (attached for
posterity) but yours is better because it crashes more releases. I
will now get out of your way :-)
--
Thomas Munro
https://enterprisedb.com