This is more of a comment to Martin Scharrer's answer (Case 2 of tohecz's answer), but too long for a comment...
While I can't disagree with anything in Martin's answer (excluding the first part of the initial sentence), I personally would prefer that bugs be posted as a question here, even if they end up being closed. Many times it is not clear whether it is a bug or not. For example consider these two questions of mine:
One was a bug, the other was not.
Had someone else reported the same symptom I would have checked that first, and it would have saved me many hours to have to distill it down to that point.
I had no idea it was related to
enumitem until I was able to get it down to a small MWE. I just thought I was perhaps redefining something incorrectly that was causing this issue.
And then at that point, do I report this as a bug for
enumitem? Not sure how a user is supposed to decide.
Furthermore I was thinking that perhaps there is still some corruption in my TL2012 installation as per:
Once others can confirm that they see the same behavior, and not some option clash issue, only then would it be known as a bug (except of course to the pacakge authors), at which case I would have emailed the authors.
And even if the problem is fixed in a updated version, many people tend to stick with old versions given that we often get questions where the answer is to update their distribution. Here is one form today: problem with mdframed commands. If there were question here with the same problems that those people were experiencing, then they would know that they needed updated packages and/or release.
I am not suggesting that this site be used as a bug tracking system or that all bugs be reported here, just that if one person took quite some time to narrow down to a bug, then posting it here is useful. And if an answer shows up that provides a workaround until that bug is fixed, that might be also be useful to others who do not want to update their packages/releases.