Problem with RHBugzilla and _add_field_alias

[Jason L Tibbitts III]

I'm kind of stumped on this one.  I have a bit of code in Fedora
infrastructure (so running on RHEL6, python 2.6, in case that's
relevant) which works with python-bugzilla 1.0 but fails with 1.1.
Unfortunately it doesn't fail at all for me on my F20 dev machine.

Here's the backtrace:

Traceback (most recent call last):
  File "/usr/local/bin/review-stats.py", line 612, in <module>
    bz = bugzilla.RHBugzilla(url=config['url'], cookiefile=None,
user=config['username'], password=config['password'])
  File "/usr/lib/python2.6/site-packages/bugzilla/rhbugzilla.py", line 59, in
__init__
    _add_both_alias('fixed_in', 'cf_fixed_in')
  File "/usr/lib/python2.6/site-packages/bugzilla/rhbugzilla.py", line 56, in
_add_both_alias
    self._add_field_alias(newname, origname, is_api=False)
AttributeError: 'RHBugzilla' object has no attribute '_add_field_alias'

This looks like it's down in the bowels of python-bugzilla itself and
not some trivial issue with my code, so I'm at a loss.  Just in case, if
I ask for a Bugzilla object instead of a RHBugzilla one, I get a
different backtrace:

Traceback (most recent call last):
  File "./review-stats.py", line 614, in <module>
    (bugs, bugdata, usermap) = run_query(bz)
  File "./review-stats.py", line 199, in run_query
    bugs = filter(lambda b: b.id not in trackers, bz.query(querydata))
  File "/usr/lib/python2.6/site-packages/bugzilla/base.py", line 995, in query
    return [_Bug(bugzilla=self, dict=b) for b in r['bugs']]
  File "/usr/lib/python2.6/site-packages/bugzilla/bug.py", line 42, in __init__
    self._update_dict(dict)
  File "/usr/lib/python2.6/site-packages/bugzilla/bug.py", line 129, in
_update_dict
    aliases = self.bugzilla._get_bug_aliases()
AttributeError: 'Bugzilla' object has no attribute '_get_bug_aliases'

You can see the code at
https://git.fedorahosted.org/cgit/fedora-infrastructure.git/tree/scripts/...
if that would be useful.

Any ideas would be greatly appreciated, because I'm fresh out.

 - J<

[Cole Robinson]

On 11/28/2014 09:22 PM, Jason L Tibbitts III wrote:
>  I'm kind of stumped on this one.  I have a bit of code in Fedora
>  infrastructure (so running on RHEL6, python 2.6, in case that's
>  relevant) which works with python-bugzilla 1.0 but fails with 1.1.
>  Unfortunately it doesn't fail at all for me on my F20 dev machine.
>
>  Here's the backtrace:
>
>  Traceback (most recent call last):
>     File "/usr/local/bin/review-stats.py", line 612, in <module>
>       bz = bugzilla.RHBugzilla(url=config['url'], cookiefile=None,
> user=config['username'], password=config['password'])
>     File "/usr/lib/python2.6/site-packages/bugzilla/rhbugzilla.py", line 59, in
> __init__
>       _add_both_alias('fixed_in', 'cf_fixed_in')
>     File "/usr/lib/python2.6/site-packages/bugzilla/rhbugzilla.py", line 56, in
> _add_both_alias
>       self._add_field_alias(newname, origname, is_api=False)
>  AttributeError: 'RHBugzilla' object has no attribute '_add_field_alias'
>
>  This looks like it's down in the bowels of python-bugzilla itself and
>  not some trivial issue with my code, so I'm at a loss.  Just in case, if
>  I ask for a Bugzilla object instead of a RHBugzilla one, I get a
>  different backtrace:
>
>  Traceback (most recent call last):
>     File "./review-stats.py", line 614, in <module>
>       (bugs, bugdata, usermap) = run_query(bz)
>     File "./review-stats.py", line 199, in run_query
>       bugs = filter(lambda b: b.id not in trackers, bz.query(querydata))
>     File "/usr/lib/python2.6/site-packages/bugzilla/base.py", line 995, in
> query
>       return [_Bug(bugzilla=self, dict=b) for b in r['bugs']]
>     File "/usr/lib/python2.6/site-packages/bugzilla/bug.py", line 42, in
> __init__
>       self._update_dict(dict)
>     File "/usr/lib/python2.6/site-packages/bugzilla/bug.py", line 129, in
> _update_dict
>       aliases = self.bugzilla._get_bug_aliases()
>  AttributeError: 'Bugzilla' object has no attribute '_get_bug_aliases'
>
>  You can see the code at
>  https://git.fedorahosted.org/cgit/fedora-infrastructure.git/tree/scripts/...
>  if that would be useful.
>
>  Any ideas would be greatly appreciated, because I'm fresh out.
>  
Is this python-bugzilla installed from EPEL?

In a brief test just trying to instantiate RHBugzilla on RHEL6, I didn't see 
any backtrace like the first example.

Any chance you manually installed a previous version of bugzilla and maybe the 
imports are getting messed up? Check /usr/local/... for example. Try fully 
removing all traces of python-bugzilla and reinstalling.

Additionally try latest git, maybe it helps:

git clone git://git.fedorahosted.org/python-bugzilla.git
cd python-bugzilla
[run your script]

- Cole

[Jason L Tibbitts III]

> >>>> "CR" == Cole Robinson
> <crobinso(a)redhat.com&gt; writes: 
CR> Is this python-bugzilla installed from EPEL?

I believe it is.  This is running on Fedora infrastructure so I don't
have complete visibility, but the exact package is
python-bugzilla-1.1.0-1.el6.noarch

It looks like it's installed properly, but:

[tibbs(a)app01.stg ~]$ rpm -V python-bugzilla
S.5....T.    /usr/lib/python2.6/site-packages/bugzilla/base.py

Which is... interesting.  I wonder if there's some hotfix being applied
outside of the rpm which is breaking things.  Let me follow up with the
infra folks.

 - J<

[Jason L Tibbitts III]

OK, so it turns out that there is a hotfix in Fedora infra which is
overwriting base.py with a slightly modified 1.0 version.  Below is the
patch that's applied.  I note that the second hunk seems to be in 1.1,
but not the first.  I'm not really familiar with this patch or the
reasons it was applied, but maybe someone here has some idea of whether
or not it's still necessary.

commit d3c6ed32625f68fa7aeb4d1d62a012aa15ad1bcd
Author: Toshio くらとみ <toshio(a)lockbox01.phx2.fedoraproject.org&gt;
Date:   Wed May 21 16:50:56 2014 +0000

    If we specify None for cookiefile, then we don't want cookies saved to disk.  They
still have to be used within the process, though.

diff --git a/modules/hotfix/files/python-bugzilla/base.py
b/modules/hotfix/files/python-bugzilla/base.py
index db4cd86..6a862ff 100644
--- a/modules/hotfix/files/python-bugzilla/base.py
+++ b/modules/hotfix/files/python-bugzilla/base.py
@@ -116,6 +116,8 @@ class RequestsTransport(Transport):
             Transport.__init__(self, use_datetime=False)
 
         self.verbose = debug
+        if cookiejar is None:
+            cookiejar = _build_cookiejar(None)
         self._cookiejar = cookiejar
 
         # transport constructor needs full url too, as xmlrpc does not pass
@@ -128,7 +130,7 @@ class RequestsTransport(Transport):
 
         self.request_defaults = {
             'cert': sslcafile if self.use_https else None,
-            'cookies': cookiejar if cookiejar else None,
+            'cookies': cookiejar,
             'verify': sslverify,
             'headers': {
                 'Content-Type': 'text/xml',


-- 
 Jason L Tibbitts III - tibbs(a)math.uh.edu - 713/743-3486 - 660PGH
 System Manager:  University of Houston Department of Mathematics

[Cole Robinson]

On 11/29/2014 03:32 PM, Jason L Tibbitts III wrote:
>  OK, so it turns out that there is a hotfix in Fedora infra which is
>  overwriting base.py with a slightly modified 1.0 version.  Below is the
>  patch that's applied.  I note that the second hunk seems to be in 1.1,
>  but not the first.  I'm not really familiar with this patch or the
>  reasons it was applied, but maybe someone here has some idea of whether
>  or not it's still necessary.
>
>  commit d3c6ed32625f68fa7aeb4d1d62a012aa15ad1bcd
>  Author: Toshio くらとみ <toshio(a)lockbox01.phx2.fedoraproject.org&gt;
>  Date:   Wed May 21 16:50:56 2014 +0000
>
>       If we specify None for cookiefile, then we don't want cookies saved to disk. 
> They still have to be used within the process, though.
>
>  diff --git a/modules/hotfix/files/python-bugzilla/base.py
> b/modules/hotfix/files/python-bugzilla/base.py
>  index db4cd86..6a862ff 100644
>  --- a/modules/hotfix/files/python-bugzilla/base.py
>  +++ b/modules/hotfix/files/python-bugzilla/base.py
>  @@ -116,6 +116,8 @@ class RequestsTransport(Transport):
>                Transport.__init__(self, use_datetime=False)
>
>            self.verbose = debug
>  +        if cookiejar is None:
>  +            cookiejar = _build_cookiejar(None)
>            self._cookiejar = cookiejar
>
>            # transport constructor needs full url too, as xmlrpc does not pass
>  @@ -128,7 +130,7 @@ class RequestsTransport(Transport):
>
>            self.request_defaults = {
>                'cert': sslcafile if self.use_https else None,
>  -            'cookies': cookiejar if cookiejar else None,
>  +            'cookies': cookiejar,
>                'verify': sslverify,
>                'headers': {
>                    'Content-Type': 'text/xml',
>
>  
I think python-bugzilla 1.1.0 should be fixed in this respect, toshio reported 
the root issue and some patches went in:

commit fbd0b048198cde065687b69e5daa08ac6656964e
Author: Toshio Kuratomi <toshio(a)fedoraproject.org&gt;
Date:   Sun Jun 1 12:57:29 2014 -0400

     transport: Still store cookies in memory if cookiefile=None

So maybe ask infrastructure to drop their patch and give it a test?

- Cole