Georg Brandl wrote:
I've become cautious of labeling patches as "trivial". Some may really be, e.g. typos and the like, but those are almost always dealt with quickly. Others may seem trivial, as in "add that line here", but there is often a problem associated -- like the question of portability, or backwards compatibility. In a few cases, we can see that as committing the fix leads to some complaint, and it is backed out again. But there might be others where the problem is overlooked and only noticed after some time in a more public fashion.
And other times something that seems to have a simple fix turns out to be a symptom of a deeper problem (there was one along those lines recently where there was an underlying issue with the changes to __hash__ inheritance in Py3k that surfaced as an apparent misbehaviour of hashing of range() instances - the problem was actually in PyObject_Hash(), range() just happened to trigger it).
Deciding when to commit a fix directly and when to use the tracker (or even a branch) to get additional input on a change is actually one of the more interesting judgment calls that comes with commit privileges.