See #609

@cpplearner: Thanks for the reference.

Closing, since there is nothing wrong with the example.

Since you got two reports on this where people think the capture is erroneously missing - it makes sense to add a comment to this snippet explaining the expected j scope.

wdyt?

The standard is not a tutorial?

@zygoloid: what do you think?

[expr.prim.lambda]'s examples are all over the place w/r/t scope. Many are scope-neutral, so to speak, but we do have stuff like this that only works at block scope, stuff like the one at issue here that only works at namespace scope, and then stuff like this that actually needs to be split up and put into two different scopes to compile. Meanwhile, we also have examples like this that introduces a function solely to get a block scope so that captures can be used. It might not be a bad idea to come up with some rule on scopes for examples and apply it uniformly.

In this particular case, I would argue that the example could use some improvement; there's no explicit marker that the declarations are at namespace scope, and I'd wager that local lambdas arise more frequently than global ones, so the "missing capture" gets distracting because it resembles a common programming error with local lambdas. But I dislike an explicit comment too, as that also distracts from the point of the example, which has nothing to do with scope or captures.

How about just adding a static before int j;?