Re: [PATCH] Fix buffer not null terminated on (ecpg lib)

Em ter., 15 de jun. de 2021 às 15:48, Andres Freund <andres(at)anarazel(dot)de>
escreveu:

> Hi,
>
> On 2021-06-15 13:53:08 -0400, Tom Lane wrote:
> > Andres Freund <andres(at)anarazel(dot)de> writes:
> > > On 2021-06-15 07:40:46 -0300, Ranier Vilela wrote:
> > >> memcpy would not suffer from it?
> >
> > > It'd not be correct for short sqlstates - you'd read beyond the end of
> > > the source buffer. There are cases of it in the ecpg code.
> >
> > What's a "short SQLSTATE"? They're all five characters by definition.
>
> I thought there were places that just dealt with "00" etc. And there are -
> but
> it's just comparisons.
>
> I still don't fully feel comfortable just using memcpy() though, given that
> the sqlstates originate remotely / from libpq, making it hard to rely on
> the
> fact that the buffer "ought to" always be at least 5 bytes long? As far as
> I
> can tell there's no enforcement of PQresultErrorField(...,
> PG_DIAG_SQLSTATE)
> being that long.
>
And replacing with snprintf, what do you guys think?

n = snprintf(sqlca->sqlstate, sizeof(sqlca->sqlstate), "%s", sqlstate);
Assert(n >= 0 && n < sizeof(sqlca->sqlstate));

regards,
Ranier Vilela