Segmentation fault on updating a BYTEA field [psycopg2]

[George Sakkis]

I have a simple DB table that stores md5 signature pairs:

Table "public.duplicate"
Column | Type | Modifiers
----------+-------+-----------
sig | bytea | not null
orig_sig | bytea | not null
Indexes:
"duplicate_pkey" PRIMARY KEY, btree (sig)
"ix_duplicate_orig_sig" btree (orig_sig)

I use SqlAlchemy to interact with it and inserting works fine; update
however crashes hard with a segfault. Sure enough, the crash is
reproducible when using the underlying psycopg2 directly:

>>import psycopg2
connect_string = ...
conn = psycopg2.connect(connect_string)
cur = conn.cursor()
cur.execute('SELECT sig,orig_sig from duplicate limit 1')
d = cur.fetchone()
d

(<read-only buffer for 0x40209100, size 16, offset 0 at 0x403054e0>,
<read-only buffer for 0x402090f0, size 16, offset 0 at 0x40305580>)

>>map(str,d)

["\x02#qO\xb0\xcc\xfcx\xb9u\xa5\x83)\xc4'@", '\xa1\xf22\xf6y\xd0\xbc
\xea6\xf0Y\xf1"\xc9(\n']

>>cur.execute('UPDATE duplicate SET orig_sig=%(orig_sig)s WHERE duplicate.sig = %(duplicate_sig)s',

.... dict(orig_sig=d[0], duplicate_sig=d[1]))
Segmentation fault

Am I (and SqlAlchemy) doing something silly or is this a bug in
psycopg2 ?

George

[George Sakkis]

On May 21, 6:32 pm, George Sakkis <[e-mail]:

I have a simple DB table that stores md5 signature pairs:

Table "public.duplicate"
Column | Type | Modifiers
----------+-------+-----------
sig | bytea | not null
orig_sig | bytea | not null
Indexes:
"duplicate_pkey" PRIMARY KEY, btree (sig)
"ix_duplicate_orig_sig" btree (orig_sig)

I use SqlAlchemy to interact with it and inserting works fine; update
however crashes hard with a segfault. Sure enough, the crash is
reproducible when using the underlying psycopg2 directly:

>import psycopg2
connect_string = ...
conn = psycopg2.connect(connect_string)
cur = conn.cursor()
cur.execute('SELECT sig,orig_sig from duplicate limit 1')
d = cur.fetchone()
d

(<read-only buffer for 0x40209100, size 16, offset 0 at 0x403054e0>,
<read-only buffer for 0x402090f0, size 16, offset 0 at 0x40305580>)>>map(str,d)

["\x02#qO\xb0\xcc\xfcx\xb9u\xa5\x83)\xc4'@", '\xa1\xf22\xf6y\xd0\xbc
\xea6\xf0Y\xf1"\xc9(\n']>>cur.execute('UPDATE duplicate SET orig_sig=%(orig_sig)s WHERE duplicate.sig = %(duplicate_sig)s',

... dict(orig_sig=d[0], duplicate_sig=d[1]))
Segmentation fault

Am I (and SqlAlchemy) doing something silly or is this a bug in
psycopg2 ?

George

I installed the latest release (psycopg2-2.0.7) and it seems the
problem is gone. Sorry for the noise.

George