Re: BUG #13985: Segmentation fault on PREPARE TRANSACTION

On Thu, Feb 25, 2016 at 6:06 PM, Andres Freund <andres@anarazel.de> wrote:

> On 2016-02-25 09:02:07 +0100, Shulgin, Oleksandr wrote:
> > On Wed, Feb 24, 2016 at 10:52 PM, Andres Freund <andres@anarazel.de>
> wrote:
> >
> > Hm...  this is against my understanding of what a compiler could (or
> > should) do.   Do you have a documentation reference or other evidence?
>
> Which part does not conform to your expectations?  Moving stores/loads
> around from where they're apparently happening in the C program?
>


> Repeatedly reading from memory instead of storing something on the
> stack?
>

This one.  But now that I think about it, this can buy you a free register,
so yeah it can be an optimization in some contexts.

How does compiler know this is not multi-threaded program?  Only because we
don't specify any -pthread flags (assuming gcc)?  Or it doesn't know/care
at all and we have to specifically turn off thread-unsafe optimizations?

--
Alex