Andreas Seltenreich wrote:
> Tom Lane <tgl@sss.pgh.pa.us> writes:
>
> > Andreas Seltenreich <andreas.seltenreich@credativ.de> writes:
> >> The scary one is due to an integer overflow the attached patch also
> >> fixes.
> >
> > s/int/Size/ doesn't fix anything on 32-bit machines.
>
> Well, it changes the signedness of the computation on 32-bit, and in
> combination with the fact that "len" is always smaller than 2^32, but
> may exceed 2^31-1, the change avoids the dependency on the undefined
> behavior of signed integer overflows in C on 32-bit as well.
Why not just use an unsigned 64 bit variable? Also, perhaps
palloc_huge() avoids the whole problem in the first place ... though it
might only move the issue around, if you cannot ship the longer-than-1GB
resulting escaped value. (Of course, if you try to allocate 2 GB in a
32 bit machine, you're going to be having quite some fun ...)
--
Álvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services