Re: round-triping bytea/buffers

On 11/01/2008, Phil Mayers <p.mayers-AQ/[email protected]> wrote:
> The following script fails with a segfault:
>
> ----
> import psycopg2
>
> conn = psycopg2.connect(DSN)
> cur = conn.cursor()
>
> cur.execute("create temp table foo (id integer, data bytea)")
>
> data = '\x00\x01\x02'
>
> cur.execute("insert into foo (id, data) values (%s,%s)", (1, data))
> cur.execute("select data from foo where id=1")
> data_fromsql = cur.fetchone()[0]
>
> print "data from sql", repr(data_fromsql), repr(str(data_fromsql))
>
> cur.execute("insert into foo (id, data) values (%s,%s)", (1, data_fromsql))
> ----
>
> Like so:
>
> 2.0.6 (dec dt ext pq3)
> data from sql <read-only buffer for 0x2aaaaab12270, size 0, offset 0 at
> 0x2aaaae7bf1f0> ''
> Segmentation fault
>
> However, if I instead do this this:
>
> data_fromsql = cur.fetchone()[0]
> data_fromsql = buffer(str(data_fromsql))
>
> ..the script works. It seems the buffer objects that psycopg returns
> cannot be passed back to psycopg.
>
> Version as above is 2.0.6

This is probably http://initd.org/tracker/psycopg/ticket/209

James.