GitLab

• Gajim git master 6ef3b8e01b2b01
• OMEMO plugin 2.6.52

Follow-up for #399 (comment 192608), related: #278 (closed), #389 (closed)

I've investigated this issue during last weeks and it looks like it is not a bug in python or sqlite but a bug in the OMEMO plugin, which shows in its python2 flavor.

Here are test scripts:

test-blob-write-buggy-using-py2.py

test-blob-read-buggy-using-py2.py

This illustrates the issue, when keys have been written using 0.16.x Gajim (python2) and then accessed with 1.x Gajim (python3):

$ rm -f test.db
$ python ./test-blob-write-buggy-using-py2.py 
DONE
$ python ./test-blob-read-buggy-using-py2.py 
OK
$ python3 ./test-blob-read-buggy-using-py2.py 
FAILED

To override a run-time failure running in python2:

sqlite3.ProgrammingError: You must not use 8-bit bytestrings unless you use a text_factory that can interpret 8-bit bytestrings (like text_factory = str). It is highly recommended that you instead just switch your application to Unicode strings.

the following has been added:

con.text_factory = bytes

Instead: in python2 the variable holding BLOB data should be passed as:

buffer(public_key)

Then the content is saved properly as BLOB not as string.

test-blob-write.py

test-blob-read.py

Writing using python2:

$ rm -f test.db
$ python ./test-blob-write.py 
DONE
$ python ./test-blob-read.py 
OK
$ python3 ./test-blob-read.py 
OK

and writing using python3:

$ python3 ./test-blob-write.py 
DONE
$ python ./test-blob-read.py 
OK
$ python3 ./test-blob-read.py 
OK

This is an example how to sanitize the existent DB (fetching with explicit cast to BLOB) using python3:

import sqlite3

con = sqlite3.connect("test.db")

query = '''SELECT some_id, CAST(public_key as BLOB) FROM identities'''

results = con.execute(query).fetchall()

for result in results:
    some_id = result[0]
    public_key = result[1]
    con.execute('UPDATE identities SET public_key = ? WHERE some_id = ?', (public_key, some_id))

con.commit()
con.close()

Testing:

$ python ./test-blob-write-buggy-using-py2.py 
DONE
$ python3 ./test-blob-read.py 
FAILED
$ python3 ./test-blob-sanitize.py 
$ python3 ./test-blob-read.py 
OK

Before sanitizing:

$ echo ".dump identities" | sqlite3 -readonly test.db 
PRAGMA foreign_keys=OFF;
BEGIN TRANSACTION;
CREATE TABLE identities (some_id INTEGER, public_key BLOB);
INSERT INTO identities VALUES(123,replace(���!Np��:,�%\nU&�!(�	S�܉�ÿͽ�v �','\n',char(10)));
COMMIT;

after:

$ echo ".dump identities" | sqlite3 -readonly test.db 
PRAGMA foreign_keys=OFF;
BEGIN TRANSACTION;
CREATE TABLE identities (some_id INTEGER, public_key BLOB);
INSERT INTO identities VALUES(123,X'08a29cb1071221054e70a5e71a3a2cdf250a55268721281b8f9556a20953b9dc89c5c3bfcdbd99761a20880015acbfa1517b91044f4e260c38d62e900f2c7ed9a2dc76bd12da9d925070');
COMMIT;

It is obvious, that binary data should be saved indeed only as BLOB not as TEXT/string, otherwise it wouldn't be properly treated by conventional tools.