--=-Qy4k09ws5Lp3vFHlXVas
Content-Type: text/plain
Content-Transfer-Encoding: 7bit

Hi!

It seems that postgres allows writing to a read-only blob opened like:

fd = lo_open (cnc, oid, INV_READ);

I've attached a simple test case and the Makefile to build it.

I'm using postgresql 7.3.3 (7.3.3-1 is the debian package version).

Is that the intended behaviour or is it a bug?

-Gonzalo

P.S.: please, CC me as I'm not suscribed to this list. Thanks.

--
Gonzalo Paniagua Javier <gonzalo (AT) gnome-db (DOT) org>
http://www.gnome-db.org/~gonzalo/




--=-Qy4k09ws5Lp3vFHlXVas
Content-Disposition: attachment; filename=inversion-bug.c
Content-Type: text/x-c; name=inversion-bug.c; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit

#include <string.h>
#include <libpq-fe.h>
#include <libpq/libpq-fs.h>

static char *conn_string =
"dbname=test user=gonzalo password=password hostaddr=127.0.0.1";

int
main ()
{
PGconn *cnc;
PGresult *res;
int oid, fd;
char *the_string = "The String";
char *other_str = " ";
int written;
int result = 0;

cnc = PQconnectdb (conn_string);
if (PQstatus (cnc) != CONNECTION_OK) {
printf ("Error connecting: %s\n", PQerrorMessage (cnc));
return -1;
}

res = PQexec (cnc, "begin");
PQclear(res);

oid = lo_creat (cnc, INV_READ | INV_WRITE); /* Don't care if only one is set */
fd = lo_open (cnc, oid, INV_READ);
if (fd < 0) {
printf ("Error opening BLOB: %s\n", PQerrorMessage (cnc));
return -1;
}

written = lo_write (cnc, fd, the_string, strlen (the_string));
if (written >= 0) {
printf ("ERROR: I was able to write %d bytes.\n", written);
result = 1;
} else {
printf ("It worked! %s\n", PQerrorMessage (cnc));
}

lo_close (cnc, fd);
lo_unlink (cnc, oid);

res = PQexec (cnc, "end");
PQclear (res);
PQfinish (cnc);

return result;
}


--=-Qy4k09ws5Lp3vFHlXVas
Content-Disposition: attachment; filename=Makefile
Content-Type: text/x-makefile; name=Makefile; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit

CC=gcc
CFLAGS = -g -I/usr/include/postgresql
LDFLAGS = -lpq

inversion-bug: inversion-bug.o

inversion-bug.o: inversion-bug.c

clean:;
rm -f *.o inversion-bug *~


--=-Qy4k09ws5Lp3vFHlXVas
Content-Type: text/plain
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
MIME-Version: 1.0


---------------------------(end of broadcast)---------------------------
TIP 5: Have you checked our extensive FAQ?

http://www.postgresql.org/docs/faqs/FAQ.html

--=-Qy4k09ws5Lp3vFHlXVas--

Gonzalo Paniagua Javier <gonzalo (AT) gnome-db (DOT) org> writes:
AFAICS from the code, INV_READ/WRITE attached to lo_open only determine
the kind of lock taken on the LO (shared or exclusive). Not sure
whether it's a bug to prohibit the other kind of access or not.

Note that there's no "security" issue here, since anyone is allowed to
write any LO anyway.

regards, tom lane

---------------------------(end of broadcast)---------------------------
TIP 8: explain analyze is your friend