Re: 9.3.9 and pg_multixact corruption

On 25/09 09.37, Andreas Seltenreich wrote:
> [ adding Bjorn Munch to Cc ]

Oh. I am on the -hackers list but usually just scan for any subject
mentioning Solaris and this one did not. :-)

> Jim Nasby writes:
> > On 9/20/15 9:23 AM, Christoph Berg wrote:
> >> a short update here: the customer updated the compiler to a newer
> >> version, is now compiling using -O2 instead of -O3, and the code
> >> generated now looks sane, so this turned out to be a compiler issue.
> >> (Though it's unclear if the upgrade fixed it, or the different -O
> >> level.)
> >
> > Do we officially not support anything > -O2? If so it'd be nice if
> > configure threw at least a warning (if not an error that you had to
> > explicitly over-ride).
> 
> At least the solaris binaries distributed via postgresql.org[1] have
> been compiled with -xO3 according to pg_config.  And their code for
> multixact.c looks inconspicuous.  To recap the data points:
> 
> | compiler                                          | flags | multixact.o |
> |---------------------------------------------------+-------+-------------|
> | Sun C 5.12 SunOS_sparc Patch 148917-07 2013/10/18 | -xO3  | bad         |
> | Sun C 5.13 SunOS_Sparc 2014/10/20                 | -xO2  | good        |
> | Sun C 5.8 Patch 121015-04 2007/01/10              | -xO3  | good        |

All the binaries I have compiled for distribution via postgresql.org
have been built with Sun Studio 11, aka Sun C 5.8 as listed on the
bottom here. Except the 9.5 Alpha where I finally had to upgrade to
version 12.1 (aka Sun C 5.10).

I am also pretty sure they have always been compiled with -O3. At
least that's what the build script is set to and I don't think that
has even been changed.

- Bjorn Munch