Univention Bugzilla – Bug 33627
AppCenter: init() seems to fail or not yet called
Last modified: 2015-05-28 16:49:42 CEST
We received two tracebacks: Traceback (most recent call last): File "/usr/lib/pymodules/python2.6/univention/management/console/modules/__init__.py", line 204, in execute func( request ) File "/usr/lib/pymodules/python2.6/univention/management/console/modules/decorators.py", line 305, in _response result = _multi_response(self, request) File "/usr/lib/pymodules/python2.6/univention/management/console/modules/decorators.py", line 432, in _response for res in function(self, iterator, *nones): File "/usr/lib/pymodules/python2.6/univention/management/console/modules/decorators.py", line 271, in _fake_func yield function(self, *args) File "/usr/lib/pymodules/python2.6/univention/management/console/modules/appcenter/__init__.py", line 469, in components_query self.uu.ucr_reinit() AttributeError: 'Instance' object has no attribute 'uu' and Traceback (most recent call last): File "/usr/lib/pymodules/python2.6/univention/management/console/modules/__init__.py", line 204, in execute func( request ) File "/usr/lib/pymodules/python2.6/univention/management/console/modules/decorators.py", line 305, in _response result = _multi_response(self, request) File "/usr/lib/pymodules/python2.6/univention/management/console/modules/decorators.py", line 432, in _response for res in function(self, iterator, *nones): File "/usr/lib/pymodules/python2.6/univention/management/console/modules/decorators.py", line 271, in _fake_func yield function(self, *args) File "/usr/lib/pymodules/python2.6/univention/management/console/modules/appcenter/__init__.py", line 105, in query self.package_manager.reopen_cache() AttributeError: 'Instance' object has no attribute 'package_manager' I would have thought this is impossible. self.uu and self.package_manager are both initialized in init(). Seems that this command either fails silently with a traceback or that it is not yet finished. If I time.sleep() in init() all other commands wait as well so this cannot be the reason. I checked the code before self.package_manager = [...] and could not identify any line that could cause a traceback. But: If it fails for whatever reason this goes unnoticed. I did not find any log information anywhere about a traceback in init(). However, it would be helpful if we: * put all the init() stuff in __init__() -> guaranteed to be executed * put all the __init__() stuff in try/except and log it somewhere Maybe log tracebacks in init() (would require a patch against umc-server)?
(In reply to Dirk Wiesenthal from comment #0) > ... > I would have thought this is impossible. In fact, quite some errors reported via Piwik happen when opening the app center UMC module. These seem probably related to the tracebacks you mentioned. > self.uu and self.package_manager are both initialized in init(). Seems that > this command either fails silently with a traceback or that it is not yet > finished. I also observed that tracebacks within theses methods are not written to the log files. > If I time.sleep() in init() all other commands wait as well so this cannot > be the reason. > > I checked the code before self.package_manager = [...] and could not > identify any line that could cause a traceback. But: If it fails for > whatever reason this goes unnoticed. I did not find any log information > anywhere about a traceback in init(). What about the following line, could it provoke an error? > util.install_opener(self.ucr) > However, it would be helpful if we: > * put all the init() stuff in __init__() -> guaranteed to be executed > * put all the __init__() stuff in try/except and log it somewhere Not sure whether this is a good idea. From the code documentation: > *init*: Is invoked after the module process has been initialised. At that > moment, the settings, like locale and username and password are available. > Maybe log tracebacks in init() (would require a patch against umc-server)? Hm… in modserver.py, the stacktrace is recorded, yet it is not printed out. I nvestigated the problem a bit and opened Bug 33673.
Maybe self.package_manager = PackageManager() failed. See Bug #33944 (Syntax error in /etc/apt/sources.list*)
Bug #33673 will improve tracebacks. Currently, I do not know what to fix. If Comment 2 is correct, then a fix is really hard. Actually I would leave it this way. What is the App Center supposed to do when apt.Cache() is unavailable? It cannot workaround this issue! (Even apt-get is unavailable in this case!) I would say always show an error. Just like it does now when init() fails... At best one could except SystemError as e: re.match(SOME_APT_ERROR_REGEX, str(e)) and raise UMC_Error('check your apt/sources.list!') Remove errata tag for now. Wait for more feedback, prepare to resolve as WONTFIX.
The very same both tracebacks have been reported again: 3.2-0 errata17 (Borgfeld) → Ticket #2014030321012411 3.2-0 errata0 (Borgfeld) → Ticket #2014021021003445
If this has nothing to do with the appcenter this bug can be marked as duplicate of Bug #34235.
Maybe related to Bug #34235 Bug #33848 Bug #37400
Fixed by Bug #33673?
The error message should be something else now, but the underlying problem still exists. We need an unobtrusive message that something went wrong during initializing. It is most probably a problem with the package manager and this is probably due to a broken sources.list. We have also seen "Not enough disk space"... I think you can except apt.SystemError when initializing the PackageManager
(In reply to Dirk Wiesenthal from comment #8) > The error message should be something else now, but the underlying problem > still exists. > > We need an unobtrusive message that something went wrong during > initializing. It is most probably a problem with the package manager and > this is probably due to a broken sources.list. We have also seen "Not enough > disk space"... > > I think you can except apt.SystemError when initializing the PackageManager Yes, you are right. I implemented the catch which shows a user friendly message now. Btw: apt.SystemError does not exists, it is the python SystemError! YAML: 2015-05-12-univention-management-console-module-appcenter.yaml Package: univention-management-console-module-appcenter Version: 4.1.20-26.353.201505131140 YAML: 2015-05-12-univention-lib.yaml Package: univention-lib Version: 4.0.5-16.300.201505131140
I think there might be a problem with the finished status: When the package manager can not be initialized while trying to uninstall an app, it is still in "working mode". Fixing the problem (which may be just waiting while the files are re-generated) does not really help, because from now on the package manager refuses to uninstall because it thinks it is still busy.
(In reply to Dirk Wiesenthal from comment #10) > I think there might be a problem with the finished status: > > When the package manager can not be initialized while trying to uninstall an > app, it is still in "working mode". Fixing the problem (which may be just > waiting while the files are re-generated) does not really help, because from > now on the package manager refuses to uninstall because it thinks it is > still busy. You are right. Fixed directly in package manger lib in svn r60852. The error should have been the same, before.
OK, works fine
<http://errata.univention.de/ucs/4.0/196.html>
<http://errata.univention.de/ucs/4.0/203.html>