The `IteratorUtil` trait hits a bug in method resolve, but the basic trait has no issue.

It's still broken because `jemalloc`'s `configure` script doesn't work when the build triple doesn't match the target triple due to missing a way to get the page size. I'm not entirely sure how to pass a hardcoded value to autoconf/configure.

This at least gets it a bit further along.

As the comment said, the subtraction is bogus for multibyte characters.
Fortunately, we can just use last_pos instead of pos to get the correct
position without any subtraction hackery.

This is something that's only been briefly mentioned in the beginning of
the tutorial and all of the closure examples within this subsection
include only one expression between { and }.

r? anybody As per #4765

As the comment said, the subtraction is bogus for multibyte characters.
Fortunately, we can just use last_pos instead of pos to get the correct
position without any subtraction hackery.

This is something that's only been briefly mentioned in the beginning of
the tutorial and all of the closure examples within this subsection
include only one expression between { and }.

The "4.3 Loops" section only describes `while` and `loop`. We then see `for`
used in a code sample at the end of the "13. Vectors and strings" section,
but it's explained for the first time only in the next section --
"14. Closures".

It is worth mentioning it in "4.3 Loops".

---

Feel free to reword it and commit yourself if you don't like my wording.

As per #4765

The "4.3 Loops" section only describes `while` and `loop`. We then see `for`
used in a code sample at the end of the "13. Vectors and strings" section,
but it's explained for the first time only in the next section --
"14. Closures".

It is worth mentioning it in "4.3 Loops".

Fixes #6155

This fail! caused ICEs, but was trivial to convert to a useful error message.

Several minor changes:
 - The clean-up I mentioned in #6851 (moving functions from deriving/mod.rs to deriving/generic.rs)
 - Move `expand_generic_deriving` to a method
 - Reimplement `deriving(Ord)` with no dependence on `Eq`

lt and gt are implement directly in terms of the corresponding
method on their elements, and le and ge are the negations of these.

These are now only called in generic and can be private. This
includes manually inlining/merging some that are called once.

This fixes #6977. Negative counts don't make sense anyway.