Control: severity 896234 normal Control: severity 896242 normal Control: severity 896272 normal Control: severity 896306 normal Control: severity 896307 normal Control: severity 896328 normal Control: severity 896378 normal Control: severity 896396 normal Control: severity 896429 normal
On Sat, Apr 21, 2018 at 09:47:59AM +1000, Brian May wrote: > Helmut Grohne <hel...@subdivi.de> writes: > > > django.core.exceptions.ImproperlyConfigured: Requested setting > > DEFAULT_INDEX_TABLESPACE, but settings are not configured. You must > > either define the environment variable DJANGO_SETTINGS_MODULE or call > > settings.configure() before accessing settings. > > I believe this bug report, and several others you filled recently that > contain this same text are false. I went passed the bug list to a few others for review and posted the full list to d-devel (including all tracebacks). Nobody spoke up and Chris Lamb vaguely said that the django ones looked legit to him. I have lowered the severity of the relevant bugs (matching "ImproperlyConfigured") to prevent issues with testing migration. > Like it or not, it is just not possible to import Django libraries > without providing a valid django settings file. This is not a sign that > something is broken. I wonder whether we can draw anything useful from these bugs before closing them. For one thing, you cannot use autopkgtest-pkg-python on these modules as is. Then having them not importable means that e.g, pydoc. That's unfortunate. Often times, modules with non-trivial impact on their environment do not do so at import time, but provide something like an install function such that the user makes a conscious choice. An example would be gbulb.install(). So yeah, for django this may make sense, but this behaviour is still unfortunate from a qa pov. I'd like to hear your opinion on this matter. After the dust has settled, I can follow up on d-devel with a summary that suggests filtering this particular django exception. Helmut
