On 07/22/2015 04:59 AM, Joe Orton wrote:

On Mon, Jul 13, 2015 at 10:26:36AM +0300, Alexander Todorov wrote: ...

...

...

data must be a byte string $ rpm -qf /usr/bin/bugzilla python-bugzilla-1.2.1-1.fc22.noarch

Any help appreciated!

Hi Joe, check out this bug: https://bugzilla.redhat.com/show_bug.cgi?id=1241561

It's not that. I'm confused though - is everybody else using python-bugzilla successfully on Fedora 22, or has nobody else tried it?

I tracked down the error, anyway.

The exception originates from pyOpenSSL's Connection.send(), which has:

    if not isinstance(buf, bytes):
        raise TypeError("data must be a byte string")

The reason is that the passed-in buffer is a unicode object rather than a byte string.

python-bugzilla's RequestsTransport.request() does this, which I don't understand:

    # Needed for python-requests < 2.0 with python3, otherwise we get
    # Content-Type error later for the POST request
    request_body = request_body.decode('utf-8')

Is the problem that some versions of python-requests want a unicode object and some want a UTF-8-encoded string (str object)? Flipping the decode to an encode fixes the problem for me, ensuring request_body is a str object not a unicode object. For reference, package versions:

python-requests-2.7.0-1.fc22.noarch python-bugzilla-1.2.1-1.fc22.noarch pyOpenSSL-0.14-3.fc22.noarch

$ rpm -q python-requests python-bugzilla pyOpenSSL python-requests-2.7.0-1.fc22.noarch python-bugzilla-1.2.1-1.fc22.noarch pyOpenSSL-0.14-3.fc22.noarch

$ bugzilla --debug login jorton foo [10:52:25] INFO (bugzilla:1160) Connecting to https://bugzilla.redhat.com/xmlrpc.cgi [10:52:25] INFO (bugzilla:1163) Autodetecting Bugzilla type [10:52:25] DEBUG (__init__:39) Detecting subclass for https://bugzilla.redhat.com/xmlrpc.cgi [10:52:25] INFO (__init__:46) Using RHBugzilla for URL containing bugzilla.redhat.com [10:52:25] INFO (__init__:114) Chose subclass RHBugzilla v0.1 [10:52:25] DEBUG (base:420) Using tokenfile=/home/crobinso/.bugzillatoken [10:52:25] DEBUG (base:537) Using cookiefile=/home/crobinso/.bugzillacookies [10:52:25] INFO (base:676) Using username/password for authentication [10:52:25] INFO (base:684) Logging in... Login failed: The username or password you entered is not valid.

I'm guessing you have some outdated python dep somewhere in your python path. Check python -c 'import sys; print sys.path' non-system paths for any stale installs.

- Cole

On 07/22/2015 02:06 PM, Joe Orton wrote:

On Wed, Jul 22, 2015 at 10:56:26AM -0400, Cole Robinson wrote:

...

I'm guessing you have some outdated python dep somewhere in your python path. Check python -c 'import sys; print sys.path' non-system paths for any stale installs.

Thanks for the suggestsion - I had checked that already... but you prompted me to confirm correct operation on another F22 machine.

I compared a strace from failing vs passing /usr/bin/bugzilla. The presence of the "python-ndg_httpsclient" package on the system breaks python-bugzilla. Remove it, bugzilla works, install it again, bugzilla fails. Can anybody else confirm that result?

Confirmed... that's interesting. I don't really have time to dig into it now. Can you file a bug against that package and CC me?

Thanks, Cole

https://github.com/python-bugzilla/python-bugzilla/pull/3

@Joe; this should fix the issue from the python-bugzilla perspective.

@Cole; I have justified dropping the workaround in the PR.

On Thu, Jul 23, 2015 at 1:10 PM Arun Babu Neelicattu < arun.neelicattu@gmail.com> wrote:

The issues seems to be caused due to:

https://github.com/kennethreitz/requests/blob/master/requests/__init__.py#L5...

That enables SNI on python 2 when ndg-httpsclient is available essentially. Haven't dug into this further. Probably see if we have a usable workaround (monkey patch a monkey patch?) or patch upstream in requests.

On Thu, Jul 23, 2015 at 4:17 AM Cole Robinson crobinso@redhat.com wrote:

...

On 07/22/2015 02:06 PM, Joe Orton wrote:

...

On Wed, Jul 22, 2015 at 10:56:26AM -0400, Cole Robinson wrote:

...

I'm guessing you have some outdated python dep somewhere in your

python path.

...

...

Check python -c 'import sys; print sys.path' non-system paths for any

stale

...

...

installs.

Thanks for the suggestsion - I had checked that already... but you prompted me to confirm correct operation on another F22 machine.

I compared a strace from failing vs passing /usr/bin/bugzilla. The presence of the "python-ndg_httpsclient" package on the system breaks python-bugzilla. Remove it, bugzilla works, install it again, bugzilla fails. Can anybody else confirm that result?

Confirmed... that's interesting. I don't really have time to dig into it now. Can you file a bug against that package and CC me?

Thanks, Cole

---

python-bugzilla mailing list python-bugzilla@lists.fedorahosted.org https://lists.fedorahosted.org/mailman/listinfo/python-bugzilla