I know questions that are bugs should be closed as off-topic with a comment explaining where to file a report. My question is, how do you tell when a question should be filed as a bug report?
For example I saw this question. It looks a lot like a bug to me, but how do I know for sure? Or is it even possible to know for sure?
It has an answer and the answer points to a user configuration problem. This is likely not a bug. How to tell if it's a bug in the future? If comments suggest it's a duplicate of a Launchpad bug, if the expected default behavior isn't working (and it's likely not a configuration issue), or it just smells like a bug. We're not here to dogpile people though, if you're not sure ask other users in chat, if there's no clear indication still then just leave it be. Someone else who is better versed in that software will come along and pass judgement on it. Don't act if you don't know or have all the facts and generally err on the side of caution/safety.
Like Jorge Castro says here, we usually don't know for sure.
Investigating bugs is what we're poorly equipped to do here. So, if the problem has to be investigated as a bug, it can usually be considered a bug.
For example:
Occasionally a program crashes for some reason other than a bug. But usually, a program crash has to be investigated as a bug--for example, producing and inspecting a stack trace--to figure this out.
Kernel panics can be caused by hardware problems, but even then it's hard to investigate them without the same mindset and techniques used to investigate a kernel bug. Especially if they are introduced by upgrading to a new kernel, or a new Ubuntu release, kernel panics are usually closed as bugs.
Sometimes the OP knows it's a bug. (That doesn't mean they're deliberately misusing the site.) Questions that say things like "can someone fix this bug please?" should almost always be closed as bugs.
Of course, there's no bright dividing line between investigating bugs and investigating configuration problems and the like, either. So outside those three cases, typically it's best to explain in a comment why you think it's a bug. Sometimes questions are wrongly closed. When they are, they can be reopened.
Usually if it's a bug you can find it listed already on one of the bug reporting sites like bugzilla or launchpad. While not everyone knows how to report a bug, many do and if it is a bug it is likely to be impacting a large enough audience that someone will have reported it properly. A simple google search is likely to turn up evidence which can help you determine if it might be a bug. As Eliah Kagan noted Jorge Castro provides some great advice here. Oli has also chimed in on this topic here.
In my opinion, a bug isn't a bug until it is proven to be a bug. Even then, I would be restraint to close it as a bug, unless the bug report offers a solution.
In many cases, bug or not, at least a temporary workaround can be found. I don't see why that would not be useful, or even inappropriate, to post on a site like this.
