Re: pgsql: Use better comment marker in Autoconf input - Mailing list pgsql-committers

From Thomas Munro
Subject Re: pgsql: Use better comment marker in Autoconf input
Date
Msg-id CADLWmXXncVaDhBWJ0U16b+NtkPeYRR+Ntb=h3bkB+d3yZxUExA@mail.gmail.com
Whole thread Raw
In response to Re: pgsql: Use better comment marker in Autoconf input  (Andres Freund <andres@anarazel.de>)
Responses Re: pgsql: Use better comment marker in Autoconf input
List pgsql-committers
On Sun, 10 Feb 2019 at 19:22, Andres Freund <andres@anarazel.de> wrote:
> On 2019-02-10 01:45:58 -0500, Tom Lane wrote:
> > Andres Freund <andres@anarazel.de> writes:
> > > On 2019-02-09 14:57:27 +0000, Peter Eisentraut wrote:
> > > I'd for a minute just put this down
> > > to an independent upgrade on the animal, but the other branches still
> > > build fine.
> >
> > The error seems to be a library version mismatch, which sure looks
> > like a busted software upgrade.  But you have a darn good point:
> > if that's the problem, why didn't v11 break too?
>
> After a few coffees it's clear why: In v11, we have jit=0. As LLVM is
> loaded on-demand, we'll not see errors in v11.

It was using CC="ccache cc" for C, but no explicit CXX.  Somehow our
configure script flipped from finding c++ (the system C++ compiler,
which is Clang) to finding g++ (an old GCC port I had installed) after
that commit, and the resulting .so didn't work too well.  I did not
upgrade or change anything at that time.  I must admit that commit of
Peter's looks fairly innocent though.  Is it possible that an earlier
change caused it, but this change caused the "accache" to be cleared?
That accache seems a bit flimsy, I've had to rm it before.  Anyway,
now that I've explicitly set CXX="ccache c++" and nuked the accache,
elver is green again.


pgsql-committers by date:

Previous
From: Tom Lane
Date:
Subject: pgsql: Fix indexable-row-comparison logic to account for coveringindex
Next
From: Peter Eisentraut
Date:
Subject: pgsql: Remove unused macro