book: claim about segfaults happening only in unsafe blocks is not true

[ivan]

If a Rust program segfaults, you can be sure it’s somewhere in the sections marked unsafe.

https://doc.rust-lang.org/nightly/book/unsafe.html

But as far as I understand it, an unsafe block can instead break an invariant that causes a segfault further down the line in safe code.

Also, stack overflows can cause a segfault.

[huonw]

You are correct in your understanding: the text could be tweaked to "you can be sure its cause is somewhere in the sections marked unsafe".

Also, stack overflows can cause a segfault.

NB. this is an implementation bug in rustc (#16012), i.e. this doesn't happen in a "perfect" Rust compiler (the book is designed to document the language not the details of rustc).

[mahkoh]

This is not quite true: http://is.gd/B4wO9H

[durka]

The nomicon and Ralf disagree.

[huonw]

@durka it depends what is meant by "cause"; a segfault will only occur if the program executes an unsafe block at some point, and does some unsafe operation with incorrect arguments, the point the nomicon and Ralf make is that the incorrectness in these arguments may flow from some safe piece of code. In any case, Steve's rewording is better than my proposal.